2.1 Type I Proteins

Type I membrane proteins, like soluble proteins to be translocated into the ER lumen, contain an N-terminal ER-targeting signal sequence to be cleaved. The signal sequence is recognised by SRP, causing the ribosome-nascent peptide complex to be targeted to the Sec61 translocon. Sec61\(\alpha\) opens the lateral gate, and the signal peptide fits into the exposed hydrophobic bindng pocket. The sequence following the signal peptide displaces the plug and inserts into the central channel, resulting in the conformation shown in Figure 2.1. Signal peptidase cleaves off the signal peptide once it recognises a specific sequence on the C-terminal end of the signal peptide. After cleavage, translation continues and the newly synthesised sequence is threaded through the channel and enters the ER lumen, until another hydrophobic segment is encountered. This segment is the TM helix of the nascent protein, and it acts as a stop-transfer signal (or “stop-transfer anchor sequence”) by triggering the lateral opening of Sec61\(\alpha\) and thus allowing this helix to move into the membrane. The C-terminus continues to be synthesised and loops out on the cytosolic side of the membrane. Eventually, the hydrophic N- and C- termini are on the luminal and the cytosolic face, respectively.

Figure 2.1: Insertion of type I membrane proteins.

2.2 Type II and III Proteins

Type II and III proteins do not contain a cleavable signal sequence, and they have opposite orientations: the N-terminus is on the cytosolic and the luminal face in type II and type III proteins, respectively.

Type II and III proteins use the sequence of their TM helix as an ER-targeting signal (“start-transfer singal”, or “signal anchor sequence”), and the targeting is also mediated by SRP and SRP receptor. The distribution of positively charged residues around the helix dictates its orientation: the positively charged residues tend to remain on the cytosolic side of the membrane, thus in type II proteins where the N-terminal side has more positively charged residues, the helix adopts the orientation with the N-terminus facing the cytosolic side; and in type III proteins where the C-terminal side has more positively charged residues the opposite orientation is adopted (
Figure 2.2 ).

Figure 2.2: Insertion of type II and III membrane proteins. In type II proteins, the orientation of the signal anchor sequence in the hydrophibic binding pocket of the laterally opened Sec61\(\alpha\) results in the opening of the pore in the vertical direction (in the same way as type I proteins, but the signal peptide is not cleaved off), allowing the remaining C-terminal portion of the polypeptide to thread through the pore and enter the ER lumen. In type III proteins, the orientation of the signal anchor sequence causes the channel to close, and the C-terminal portion is synthesised in the cytosol.

4.1 Insertion of Tail-anchored Proteins by the Get system

Tail-anchored proteins also use an \(\alpha\)-helix to traverse the membrane, but this helix is located near the C-terminus. Because there are only few, if any, residues succeeding the C-terminal helix, translation terminates while the helix have not yet emerged from the exit tunnel, so recognition by SRP is not possible, and its insertion into ER membrane depends on a post-translational pathway involving Get1, Get2, and Get3 proteins.

In this pathway, the protein is completely synthesised and released into the cytosol. ATP-bound Get3 binds to the hydrophobic C-terminal tail of the protein (facilitated by some other proteins), and then docks onto the dimeric Get1/Get2 receptor on the ER membrane. Accompanying the hydrolysis of ATP by Get3, the Get1/Get2 complex facilitates the insertion of the tail into the ER membrane. Finally, ATP displaces ADP and Get3 is released back to the cytosol.

4.2 The Roles of YidC

YidC contains 5 TM helices arranged to form a partially hydrophilic groove that is open towards both the lipid bilayer and the cytosol (Figure 4.1). Cross-linking experiments suggest that this groove operates as a binding site for TM helices (of the protein to be inserted). YidC alone can mediate insertion of a small subset of small (single- or double-pass) proteins in either co-translationally or post-translationally, and it is also suggested that YidC may act as a chaperone in conjunction with the Sec translocon. Specifically, it may shield the hydrophilic surface of TM helice that line polar cavities/channels in the final structure.

Figure 4.1: The structure of Bacillus halodurans Yidc.

5.1 Minor Groove Narrowing

Rohs et al. (2009) reported that the binding of arginine residues to narrow (\(< 5\) angstroms) minor grooves is a common mode of protein-DNA recognition (Figure 5.1). These narrow groove are usually associates with A-tracts, which estabilish a connection between DNA sequence and shape. Arginines often insert into the minor groove as part of short sequence motifs that vary among different proteins, e.g. Arg-Gln-Arg in the Hox protein SCR, Arg-Lys-Lys-Arg in POU homeodomains, thus providing specificity.

The narrowing results in a more negative electrostatic potential in the minor groove, which promotes the binding of positively charged arginine. The preference of arginine over lysine can be explained by the greater energetic cost of removing a charged lysine from water due to lysine’s smaller radius of the charged group (or greater charge density) compared to arginine.

Figure 5.1: Amino acid frequencies in minor grooves. Taken from Rohs et al. (2009)

5.2 Kinks

YpR steps (especially TpA steps) have a strong tendency to form kinks that disrupt the linearity of the double helix. They can contribute to binding specificity by optimising protein-DNA contacts. The binding site of the catabolite activator protein (CAP), for example, shows dramatic kinks at two CpA steps, which along with two additional smaller kinks cause an overall bending of the DNA of about 90 degrees around the protein.

7.1 (a) bHLH Motif

The basic helix-loop-helix (bHLH) motif consists of two \(\alpha\)-helices separated by a loop. One of the helices (usually the longer one) contain basic residues (arginine and lysine) that bind to the major groove of DNA. The flexible loop and the other helix are involved in dimerisation. Many bHTH proteins occur as heterodimers (e.g. Myc/MAX), and their activity is hightly regulated by the dimerisation of subunits.

7.2 (b) Leucine Zipper

The leucine zipper motif consists of a dimer of \(\alpha\)-helices. The dimerisation is driven by interactions between the hydrophobic side chains that cover their inner surfaces. Specifically, each helix has a periodic repetition of leucine residues at every seventh position. Since every turn in an \(\alpha\)-helix contains 3.6 amino acids—two turns contain 7.2, which is slightly more than 7, the helices coil around one another in a left-handed sense. The N-terminal DNA-binding domains of each helix protrudes into the major groove of DNA, which together recognise a 8-bp long sequence.

7.3 (c) Zinc Finger

The zinc finger contains a short \(\alpha\)-helix, a two-stranded antiparallel \(\beta\)-sheet, and a Zn²⁺ ion coordinated by cysteine and histine residues. The zinc ion serves to stabilise the overall structure, while the helix make contact with DNA’s major groove. A typical zinc finger protein, e.g. Zif268, contains a chain of 3 zinc-finger modules that coil in a right-handed sense, so as to follow the curve within the major groove. Each finger recognises 3 base pairs, and thus a protein with 3 zinc finger domains recognise a continuous sequence of 9 base pairs.

7.4 (d) TATA box binding protein

TATA box binding proteins (TBP) use a ten-stranded \(\beta\)-sheet to recognise DNA by binding in the minor groove. Insertion of the concave \(\beta\)-sheet into the groove requires substantial DNA distortion. The flexibility intrinsic to TpA steps (Section 5) in the TATA sequence faciliates formation of kinks when TBP binds. In addition, yeast TBP-TATA structure shows that the kinks in the first and last base pair step (TATATAAA) are stabilised through intercalations with phenylalanine residues.

2.1 Reversibility

Protein-protein interactions can be stable (permanent) or transient. Stable interactions are involved in the assembly of proteins made of multi-subunit complexes such as haemoglobin, which non-reversible in normal physiological conditions. Transient interactions, on the other hand, are reversible, and it is this property that make them act like molecular switches that play versatile roles in controlling cellular processes.

2.2 Properties of the Binding Interfaces

Protein-protein interaction interfaces often have a large surface area (1000-2000 Ų) and are relatively flat compared to the deep cavities that typically bind small molecules. On a binding interface, some residues, known as “hotspots”, contribute to the overall affinity more than other residues.

2.3 Roles of Water Molecules

Crystal structures frequently reveal water molecules within PPI interfaces. These water molecules play multifaceted roles in the stability of PPI, e.g. offsetting unfavourable electrostatic interactions, bridging two distant residues via H-bonds.

2.4 Kinetics and Thermodynamics

Reversible PPIs have two important parameters: affinity and specificity. While affinity ranges from as low as millimolar to as high as femtomolar, it is important that the specificity, i.e. the relative affinity of a protein to its cognate binding partner compared to non-cognate ones, is high.

Reversible PPIs can the considered as a simple balance of association and dissociation reactions, with rate constants being \(k_{\text{on}}\) and \(k_{\text{off}}\).

\[ \text{A+B}\mathrel{\mathop{\rightleftarrows}^{k_{\text{on}}}_{k_{\text{off}}}} \text{AB} \]

The affinity is usually defined by the dissociation constant:

\[ K_\text{d} = \dfrac{\text{[A][B]}}{\text{[AB]}} = \dfrac{k_{\text{off}}}{k_{\text{on}}} \]

where [A], [B], and [AB] are the concentrations of each species at equilibrium.

\(K_\text{d}\) can be converted to \(\Delta G\) and vice versa:

\[ \Delta G = -RT\ln{(K_\text{d})} \]

In addition to the simple single-step model, Keeble and Kleanthous (2005) suggested that relatively low affinity PPIs may be better modelled with a two-step induced-fit mechanism involving an unstable intermediate, where electrostatics drives the fast first step (supported by strong dependence on ionic strength) and rigid body rotation occurs in the slow second step:

\[ \text{A+B} \mathrel{\mathop{\rightleftarrows}^{k_{1}}_{k_{-1}}} \text{AB}^\text{*} \mathrel{\mathop{\rightleftarrows}^{k_{2}}_{k_{-2}}} \text{AB} \]

3.1 Determining K_d (and rate constants)

Many experimental methods can be used used for studying thermodynamics and kinetics of PPIs (the frequent tasks are determining \(K_\text{d}\) and rate constants). Most of them assumes the simple single-step association-dissociation model (Section 2.4). Three methods are described in this essay.

3.1.1 Surface Plasmon resonance (SPR)

Surface plasmon resonance (Figure 3.1) can be used to measure both \(K_\text{d}\) and \(k_\text{on}\) and \(k_\text{off}\) rates. How \(K_\text{d}\) can be calculated is explained in Equation (3.1).

Figure 3.1: A surface plasmon is an electron oscillation generated at a surface interface between a metal and a dielectric. A plasmon resonance occurs when EM wave in visible light couples optimally with the oscillating electrons in the metal, and this results in a maximal reduction in the reflected light intensity. The resonance angle, \(\theta_\text{spr}\), is found by changing the angle of incidence of the light beam, giving a dip in a plot of intensity against angle. \(\Delta\theta_\text{spr}\) is sensitive to changes in the refractive index of the medium near the metal surface and this is a measure of the mass change at the sensor surface (in the evanescent region). In an SPR experiment, the protein acting as the “bait” is immobilised on the sensor surface, and a the analyte containing the other protein (acting as the “prey”) is passed through the cell. If binding occurs between the two proteins, \(\Delta\theta_\text{spr}\) would increase. Then, non-specific binding is washed off by buffer, and \(\Delta\theta_\text{spr}\) would decrease and \(\Delta\theta_\text{spr}\). Finally, regeneration solution is applied to remove all binding and reset \(\Delta\theta_\text{spr}\) to zero.

\[\begin{equation} \begin{split} \overbrace{k_{\text{on}}\text{[A][B]}}^\text{rate of association} & = \overbrace{k_{\text{off}}\text{[AB]}}^\text{rate of dissociation} \\ k_{\text{on}}\text{[A]([B]}_{\text{max}}-\text{[AB])} & = k_{\text{off}}\text{[AB]} \\ \text{[AB]} & = \dfrac{k_{\text{on}}\text{[A][B]}_{\text{max}}}{k_{\text{on}}\text{[A]} + k_{\text{off}}} \\ \text{[AB]} & = \dfrac{\text{[A][B]}_{\text{max}}}{\text{[A]} + K_{\text{d}}} \end{split} \tag{3.1} \end{equation}\]

In the equation, A is the protein in the analyte, whose concentration is kept constant, and B is the immobilised protein. Since the \(\Delta\theta_\text{spr} \propto \text{[AB]}\) (intensity of the signal is proportional to concentration of protein-protein complexes), we can work out \(K_\text{d}\) from our initial concentrations of A and B.

SPR was used in the early kinetic analysis of hGH binding (Wells (1996)).

3.1.3 Isothermal titration calorimetry (ITC)

ITC measures heat changes when a complex is formed at constant temperature. In ITC, an insulated reaction cell containing protein is kept at a temperature (usually 8\(^\circ\text{C}\) above the environment) which is equal to the temperature of a reference cell, and the reference cell is kept at a constant temperature by a thermostat. Then, increasing amounts of ligand is added into the chamber, and they form complexes with the protein, which can be exothermic or endothermic. The heat change is compensated by a power supply, which can be converted to \(\Delta H\) of the reaction. As more ligands are added, proteins become saturated and \(\Delta H\) approaches zero. The raw data obtained (power supplied to compensate the heat change caused by each addition of ligands) can be integrated and corrected to give a plot of \(\Delta H\) against the molar ratio of the ligand and the protein, and \(\Delta H\), K_d and stoichiometry can be inferred from the curve (Figure 3.2).

Figure 3.2: ITC data manipulation

3.2 Mechanistic Studies

The main theme of mechanistic studies is to find out the “hotspot” residues or regions that are main contributors to the affnity of PPI interfaces, and to attempt to generalise this knowledge in order to predict the affnity of any given PPI interfaces.

3.2.3 Directed Evolution

Directed evolution is an efficient approach to probe sequence and structure space in a PPI. Phage display is a specific implementation of it.

The phage display technique is summarised as follows and illustrated in Figure 3.3:

1. Generate a randomised library of mutants. This can be done using error-prone PCR (ep-PCR). In PPI studies usually the randomised mutations are generated only across small sections of protein sequence that are form of the PPI interface.
2. The randomised mutant DNA are ligated with a phage coat protein gene and the hybrid is then used to transduce E. coli. cells.
3. The phage library is amplified by E. coli. cells, and phage particles are produced.
4. The “bait” protein is immobilised, and the mixture of phage particles displaying different mutant proteins are added. Then, those with low binding affinity are washes away, and the remaining phages and collected, amplified, and used in the next round of selection.
5. Repeated cycles of selection (a.k.a. “panning”) will identify the mutant proteins with the highest binding affinity to the “bait” protein. The complexes formed by these proteins with the “bait” protein can then be used for detailed mechanistic studies of PPI.

Figure 3.3: Directed evolution with phage display.

3.2.5 Connectivity Map Reveals Modularity

Reichmann et al. (2005) analysed the TEM1–BLIP complex by drawing a connectivity map, which is build from the physical interactions between the proteins (hydrogen bonds, van der Waals interactions, etc.), and showed that the interface can be divided into 6 clusters. The change \(\Delta \Delta G\) on different clusters was found to be additive, whereas mutations within the same cluster caused complex energetic and structural consequences. Therefore, a PPI interface can be seen as a group of “hot regions”, where each region contribute relatively independently to the total binding affinity, but within each region the contributions from its component amino acids are cooperative.

Figure 3.4: Connectivity map of the TEM1–BLIP complex.

Ca²⁺ Binding Sites

Using the anomalous diffraction data, the F⁺_Ca - F^-_Ca anomalous difference map was calculated. Two strong peaks followed by a weaker peak were found along the ion-conduction pathway, which correspond to the three Ca²⁺ binding sites. They are designated site 1, 2 and 3 from the extracellular side to the intracellular side.

Site 2 is the site with the highest affinity for Ca²⁺. It is surrounded by a total of 8 oxygen atoms, 4 of which coming from the carboxylate of D177 above and the other 4 from the carbonyl of L176 below. Site 1 is coordinated by the plane of 4 carboxyl groups from D178, and site 3 by the plane of 4 carbonyls from T175 (Figure 2). Throughout the selectivity filter, the O-Ca²⁺ coordination distances are in the range of 4.0-5.0 Å, which is much longer than the ionic diameter of Ca²⁺ (2.28 Å), suggesting that the bound Ca²⁺ ion maintains its hydration shell while passing through the pore. Site 3 has the lowest affinity, consistent with its role in exit of Ca²⁺ from the selectivity filter into the central cavity. Figure 3 shows the molecular model and electron density near the selectivity fileter region of Ca_vAb (TLDDWSN + 15mM Ca²⁺) in COOT⁵ (contoured at r.m.s.d = \(3.02 \sigma\)).

Figure 2: A schemetic showing the Ca²⁺ ion coordination sites of Ca_vAb (TLDDWSD)

Figure 3: The selectivity filter region of 4MS2.

The relative affinities of the three sites were further confirmed by experiments on Ca_vAb (TLDDWSN) with varying Ca²⁺ concentrations (0.5, 2.5, 5, 10, 15mM): at low Ca²⁺ concentration, two strong peaks of approximately equal intensity are found at Site 1 and Site 2; at high concentration the electron density is significantly enhanced in Site 2 and decreased in Site 1. The electron density at Site 3 remains low in all concentrations.

Ion-Permeation Mechanism

Based on the properties of the three coordination sites, an ion-permeation mechanism can be deduced. The three coordination sites are separated by a distance of about 4.5 Å, which makes it energetically unfavourable for Ca²⁺ to occupy adjacent sites simutaneously. Thus the authors suggested that the selectivity filter oscillates between two states, in which either a single hydrated Ca²⁺ occupies Site 2, or two of them each occupies Site 1 and Site 3. The entry of Ca²⁺ into Site 1 is promoted by it high extracellular concentration, and the exit of Ca²⁺ is facilitated by the low affinity of Site 3.

Notably, this mechanism suggests that Ca²⁺ ions are kept hydrated during its passage through the selectivity filter. This is very different from the mechanism by which potassium channels achive selectivity, which requires K⁺ ions’ hydration shell to be removed.

Roles of Key Selectivity Filter Residues

The five variants of Ca_vAb the author produced have different Ca²⁺ selectivity ratios, which can be partially explained by directly comparing the difference in the arrangement of selectivity filter residues. This comparison is sometimes facilitated by superposition of one structure onto another. For example, the carboxyl group of D177 in TLDDWSD interacts with the Ca²⁺ ion, while the carboxyl group of E177 in TLEDWSD swings away from the selectivity filter and forms a hydrogen bond with D181 and the main-chain nitrogen atoms of S180.

2.1 R and R_free

R is a measure of the deviation of calculated intensities from models (details in Section 2.2) from the observed intensities in the diffraction pattern, defined by the following equation:

\[ R = \dfrac{\sum ||\mathbf{F}_\text{obs}| - |\mathbf{F}_\text{calc}||}{\mathbf{F}_\text{obs}} \]

Since bias can easily be introduced into the R value (especially by overparameterisation, see Section 2.2) and a reduction of R value sometimes does not improve the actual quality of structure (Kleywegt and Brünger (n.d.)), a small fraction (typically around 5%²) of randomly selected reflections are removed from the data used for refinement. These reflections can then be used to calculate an R factor, denoted as \(R_\text{free}\), whose reduction can be considered as an unbiased estimate of the improvement of the model.

Figure 2.1: R and R-free values decreases as refinement proceeds (left: Coot; right: phenix.refine).

Figure 2.1 shows the decrease of R and R_free during a refinement task conducted in Phenix.

2.2 Reciprocal-Space Refinement: Refinement by least squares

Reciprocal-space refinement involves computerised attempts to improve agreement between \(\mathbf{F}_\text{obs}\) and \(\mathbf{F}_\text{calc}\) by without consideration of the maps and models. Refinement by least squares is the earliest successful technique and is discussed here.³

The goal of refinement by least squares is, find \((x_j, y_j, z_j)\) for all atom \(j\) whose expected ( i.e. computed) structure factor amplitudes, \(|\mathbf{F}_\text{calc}|\) are as close as possible to observed structure factor amplitudes, \(|\mathbf{F}_\text{obs}|\). Specifically, this means minimising the function \(\Phi\):

\[\begin{equation} \Phi = \sum_{hkl}(w_{hkl}\mid\mathbf{F}_\text{obs}\mid - \mid\mathbf{F}_\text{calc}\mid)_{hkl}^{2} \end{equation}\]

where \(w_{hkl}\) is the weight term that depends on the reliability of the corresponding measured intensity and \(|\mathbf{F}_\text{calc}|\) is a variant form of Equation (2) that can include additional parameters such as B-factor (\(B_j\)) and occupancy \(n_j\). An equation with \(B_j\) and \(n_j\) included can be written as:

\[\begin{equation} \mathbf{F}_\text{calc} = \sum_{j}n_j f_j e^{2\pi i(hx_j +ky_j + lz_j) - B_j[(\sin\theta)/\lambda]^2} \end{equation}\]

Note that the equatin shows that the effect of B-factors depends on the angle of reflection \((\sin\theta)/\lambda\).

Solving the minimum of \(\Phi\) analytically is impractical, and instead numerical methods are used, which would lead to a minimum closest to the starting value.To prevent the refinement converging to a local minimum, it is important that the starting parameters be near the global minimum. Is also important not to include too many parameters (such as B-factor) at the initial stages of resolution when the resolution is low, as this would decrease the radius of convergence (Figure 2.2).

Figure 2.2: Adding number of parameters improves precision of refinement, but makes it more unlikely to reach the global minimum from a given point. Thus, refinement starts with a small number of parameters, and more parameters are only added after the success of previous lower-resolution refinement steps.

To minimise the number of parameters used during early stages of refinement (and thus to increase radius of convergence), individual \((x,y,z)\) coordinates are actually not used. Instead, only torional angles \(\psi\) and \(\phi\) are allowed to change, and all bond lengths and angles are fixed to their theoretical average, side chains are assumed to be in their preferred conformation, and peptide linkages are fixed to be planar. This strategy is known as restrained reciprocal space refinement. As refinement proceeds, more parameters, from individual \((x,y,z)\) coordinates to isotropic B-factors and finally anisotropic factors, can be added into calculation.

2.3 Real-Space Refinement: Map Fitting

Map fitting or model building entails building a molecular model that fits realistically into the current electron density contour map.

To reduce the bias (towards F_calc (F_c)) when constructing the electron density map, Fourier syntheses of F_obs and F_calc are used. A Fourier synthesis mF_o - nF_c is calculated as:

\[ \rho(x, y, z) = \dfrac{1}{V}\sum_{h}\sum_{k}\sum_{l}(m|\mathbf{F}_\text{o}|-|\mathbf{F}_\text{c}|)e^{-2\pi i(hx + ky + lz - \alpha^\prime_{hkl})} \]

and its corresponding electron density map is called an mF_o - nF_c map.

Simply put, the 2F_o - F_c map resembles a molecular surface, and a F_o - F_c map emphasises the error (positive density implies that the unit cell contains more electron density in this region than implied by the model (F_c). Near the end of refinement, the F_o - F_c map becomes rather empty except in problem areas, which may need to be corrected manually.

Fitting a molecular model into the electron density map depends on prior knowledge, such as average bond lengths and angles, the amino acid sequence of the protein, properties of peptide chains, etc. For example, we know that carboxyl oxygens in adjacent amino acid residues in a \(\beta\)-sheet point in opposite directions. Thus, once a \(\beta\)-sheet along with one or two carboxyl oxygen are discernible, we can make a sensible guess of the positions of all other carbonyl oxygens.

3.1 Density fit analysis and local geometry validation

Automatic model building and refinement use the decrease of R value as an indicator of progress and terminates when R is considered to be sufficiently low. This may lead to situations where the global R is favourable but local geometry can still be improved.

Local geometry validation programs, such as “Density fit analysis” in Coot (Figure 3.1), evaluate the model geomtry on a per-residue basis and flag outliers. These outliers can then be fixed manually. With the aid of electron density contour maps (where model atoms lie outside 2F_o - F_c contours, the F_o - F_c will often show the atoms with negativel contours, with nearby positive contours pointing to correct locations for these atoms).

Figure 3.1: Use the ‘Density fit analysis’ function to evaluate model geometry on a per-residue basis and plot a histogram that shows outliers, then use the ‘Rotamers’ tool to fix a side chain that’s pointing the wrong way.

REMARK   3  DEVIATIONS FROM IDEAL VALUES.                                       
REMARK   3                 RMSD          COUNT                                  
REMARK   3   BOND      :  0.003           1366                                  
REMARK   3   ANGLE     :  0.675           1846                                  
REMARK   3   CHIRALITY :  0.050            186                                  
REMARK   3   PLANARITY :  0.005            246                                  
REMARK   3   DIHEDRAL  : 15.473            459                                  

3.1.2 Torsional Angles and Ramachandran Plot

Torsional (dihedral) angles \(\psi\) and \(\phi\) are show much more variation than bond lengths and angles, but only a subset of all possible (\(\phi\), \(\psi\)) pairs are allowed so that adjacent amino acid side chains do not clash. Validation of torsional angle is achieved via a lookup table, where the keys are (\(\phi\), \(\psi\)) pairs and values are scores. A (\(\phi\), \(\psi\)) is considered preferred or allowed if its score is within certain thresholds. Otherwise, it is considered an outlier. Due to glycine’s small size and proline’s cyclic structure, the preferred/allowed regions of their torsinal angle pairs are defined differently, for example in Phenix/cctbx⁴. Torsional angle validation is often visualised with a Ramachandran plot, as shown in Figure 3.2.⁵

Figure 3.2: Validating dihedral angles with Ramachandran plot in Coot

2.1 Implementing The Bacterial Glycerate Pathway

Many bacteria possess a simple and efficient pathway for converting glyoxylate to glycerate. This pathway involves two steps: 1) condensation of two glyoxylate to tartronate semialdehyde (by glyoxylate carboxyligase (GCL)) while releasing CO₂; 2) reduction of tartronate semialdehyde by tartronate semialdehyde reductase (requiring one NADH) to glycerate. Compared with the native photorespiratory pathway (which converts glyoxylate to glycine in peroxisome and decarboxylates glycine in the mitochondrion), NH₃ release is prevented and, if this pathway can be implemented in the chloroplast, CO₂ release is localised to the chloroplast and is more likely to be re-assimilated by Rubisco instead of escaping to the atmosphere.

To implement the pathway in the chloroplast, a targeting sequence (e.g. Arabidopsis Rubisco small subunit (RbcS) or phosphoglucomutase transit peptide sequence) needs to be added to the N terminus of the gene constructs of the desired enzymes. In addition, the enzyme(s) that converts glycolate to glyoxylate also need(s) to be imported to the chloroplast. In normal plants, this conversion is catalysed by glycolate oxidase in peroxisome, and the same enzyme can be targeted to the chloroplast. However, glycolate oxidase is inefficient, as it uses O₂ as the hydrogen acceptor, which wastes reducing power. In addition, its reaction generates H₂O₂, which needs to be broken down by catalase. A more efficient bacterial enzyme, glycolate dehydrogenase (GDH), can be used to replace glycolate oxidase’s role. GDH uses NAD⁺ instead of O₂ as the hydrogen acceptor, which prevents production of toxic H₂O₂ and preserves reducing power.

According to the scheme described above, Kebeish et al. (2007) generated plants transformed with five chloroplast-targeted bacterial genes encoding glycolate dehydrogenase, glyoxylate carboligase and tartronic semialdehyde reductase (Figure 2.1, 1). In these plants, some glycolate is successfully converted directly to glycerate in chloroplasts, and reduces, but not eliminate, flux of photorespiratory metabolites through peroxisomes and mitochondria. The transgenic plants grew faster, produced more shoot and root biomass, and the content of soluble sugars (fructose, glucose and sucrose) increases significantly.

Figure 2.1: Three photorespiratory bypasses. The loop represents the native pathway.

Carvalho et al. (n.d.) designed a similar photorespiratory bypass (Figure 2.1, 3). It also uses tartronate semialdehyde (TSA) as the intermediate and thus prevents NH₃ production. However, the CO₂-releasing TSA synthesis reaction occurs in the peroxisome rather than in the chloroplast and therefore this mechanism does not have the additional benefit of reducing CO₂ loss. In the peroxisome, TSA is converted to hydroxypyruvate by hydroxypyruvate isomerase (HYI), and the rest is identical to the native pathway. Unfortunately, all plants fail to express HYI, and for those with GCL expressed, chlorosis and stunted growth are exhibited, similar to photorespiration mutants. (I think one statement regarding this piece of research at the end of page 808 of the review written by Weber and Bar-Even (2019) is not appropriate, as HYI was not even expressed)

Maier et al. (2012) used a completely different approach. Instead of taking shortcuts within the native photorespiratory pathway with 3-PGA being the final product, they introduced glycolate oxidase, catalase and malate synthase into the chloroplast, which, together with pyruvate dehydrogenase and malic enzyme which are natively present in the chloroplast, completely oxidise glycolate into two CO₂ (Figure 2.1, 2), while producing NADPH and NADH. Although this pathway requires more energy than the native pathway to restore the ‘status quo ante’ according to the calculation (see Peterhansel, Blume, and Offermann (n.d.), Table 2), and is shown in computational modelling to result in a 31% decrease in photosynthetic efficiency compared to WT (Xin et al. 2015), experimental studies on Arabidopsiis thaliana surprisingly revealed an increase in biomass in these transgenic plants. A modified version of this approach carried out later on tobacco plants by South et al. (2019) were shown to be even more successful than Kebeish’s approach which was calculated to be the most energy-efficient. The success of Maier’s method may be explained by thhe fact that, in hot climates (where sunlight is intense), ATP and NADPH supply from light reactions is often not limiting, so an improved energy balance in photorespiration does not necessarily improve the rate of carbon fixation. Instead, the benefit brought about by the rapid CO₂ generation in this bypass is more important. In addition, it is argued that the ‘wasteful’ dissipation of reducing power by glycolate oxidase may actually be beneficial in avoiding production of reactive oxygen species (ROS) in high light intensity, when the output of light reactions is in large excess. This also explains why higher plants prefer glycolate oxidase rather than glycolate dehydrogenase.

2.2 Inhibition of the Glycolate Transporter

In the two successful photorespiratory bypasses described above, glycolate metabolism inside chloroplast competes with its export via the glycolate exporter PLGG. Therefore, repression of this transporter is expected to increase the flux into the bypasses and thus further increase the efficiency. South et al. (2019) studied the effect of RNAi inhibition of PLGG1 in these two methods as well as a ‘combined’ scheme of the two. The constructs and mechanisms of action are shown in Figure 2.2. Both AP1 and AP3 (but not AP2) showed significant increase in dry weight biomass, by 13% and 18%, respectively. When the PLGG1 RNAi module is added, the benefit of AP1 is lost, while that of AP3 increases to 24% (Weber and Bar-Even (2019) implied, on page 808 when citing this work, that inhibition of PLGG always increases efficiency, which is not correct).

Figure 2.2: South’s (2019) three photorespiratory bypass constructs. AP1: five bacterial genes encoding glycolate dehydrogenase (GDH), glyoxylate carboligase and tartronic semialdehyde reductase (same as Kebeish (2007)); AP2: glycolate oxidase (GOX), malate synthase and catalase (CAT) (same as Maier (2012)); AP3: same as AP2, except that GOX+CAT is replaced by GDH.

Elimination of the AP1 enhancements by the PLGG1 RNAi module implies that this introduced pathway ‘may not have had sufficient kinetic capacity to handle the full glycolate flux under high rates of Rubisco oxygenation’, according to South et al. (2019). This is in accordance with the aforementioned statement that the efficiency of photorespiratory bypasses in accelerating carbon fixation does not depend on energy economy (conservation of ATP and NADPH), but instead on the rate of conversion of 2-PG into TCA cycle intermediates (CO₂ and/or 3-PGA). That is to say, the real culprit for photorespiration-induced reduction in carbon fixation activity is not loss of energy in the form of ATP and NADPH, but is likely to be the inefficiency of removing 2-PG, as well as the low CO₂:O₂ ratio. Thus, in South’s AP3 construct where 2-PG is efficiently cleared up and CO₂ is produced in a large amount, the most prominent increase in carbon fixation and biomass is observed.

1.1 Rubisco Does not Exert Much Control on the Flux of the Calvin Cycle Unless in Special Conditions

Rubisco catalyses the carbon fixation step in the Calvin cycle and was initially widely thought to act as a control point on the rate of photosynthesis. Indeed, the enzymatic activity of Rubisco is modulated by a number of ways, such as activation by resersible carbamylation by CO₂ and by Rubisco activase, and inhibition by a number of sugars resulting from Rubisco’s side reactions.

1 produced a series of tobacco plants (Nicotiana tabacum) that exhibit a range of reduced amounts of Rubisco by Agrobacterium-mediated transformation of plants with antisense mRNA to the gene for the small subunit of Rubisco (rbcS). Such ‘antisense’ plants were used by² to examine the control exerted by Rubisco on the rate of photosynthesis. They found that in optimum enviornmental conditions the amount of Rubisco in a leaf could be reduced by more than one-third before any significant effect on the rate of photosynthesis, with \(C^J\) = 0.05-0.15 (very small). It was shown that reduction of enzyme amount is compensated for by an increase in Rubisco activation (from about 55% to almost 100%) due to 1) an increase of substrates (ribulose 1,5-bisphosphate and CO₂) and decrease of products (3-phosphoglycerate), and 2) an increase of ATP/ADP ratio in the chloroplast stroma. Also, Rubisco is produced in large excess in WT plants, also explaining the lack of impact of its knock-down.

However, when plants were grown in low light and are then suddenly exposed to high light intensity, there was a near-proportional relation between the amount of Rubisco and the rate of photosynthesis³. A similar result can be obtained with low CO₂ concentration. These experiments show that the contribution of Rubisco to the control of photosynthesis depends on both current and past conditions (CO₂ concentration and light intensity).

3.1 The Role of the Plastoquinone Pool

The oxidative status of the plastoquinone (PQ) pool reflects the relative activity of PS-II and PS-I, and is involved in the signalling pathway of both short-term and long-term coordination (described in Section 3.2 and 3.3, respectively).

Upon illumination by PS-I light (in state 1), PS-I activity is greater than PS-II, leading to oxidation of the plastoquinone pool (most plastoquinone molecules are in the oxidised form, PQ). Conversely, the plastoquinone pool becomes reduced upon preferential excitation of PS-II (Figure 3.1).

Figure 3.1: Preferential activation of either PS-I or PS-II causes a shift in the oxidation state of the plastoquinone (PQ) pool.

3.2 Reversible Redistribution of LHC-II

LHC-II plays an important role in the coordination between PS-I and PS-II. The structure of one LHC-II protein has been determined by a combination of electron microscopy and X-ray crystallography (reviewed by Barros and Kühlbrandt (2009)). It contains three membrane-spaning \(\alpha\)-helices, and binds about 15 chlorophyll a and b molecules plus a few carotenoids. LHC-II complexes of different subtypes (of which Lhcb1, 2 and 3 are the most common) spontaneously assemble into trimers (can be heterogeneous), which then associate with PS-II and PS-I to increase the efficiency of their corresponding photochemical reaction.

Figure 3.2: Phosphorylation by STN7 causes conformational changes in LHC-II. Top: stroma; bottom: thylakoid lumen.

The LHC-II subtypes Lhcb1 and Lhcb2 (but not Lhcb3) can be phosphorylated by a specific protein kinase called STN7 (a.k.a. Stt7) at a threonine residue near the N-terminus (Mullet 1983). This happens during transition to state 2, when the STN7 kinase, which is associated with a specific region of the Rieske protein of the cytochrome b₆f complex (Cytb₆f), is activated upon binding of PQH₂ to the Q_o site of Cytb₆f. Phosphorylation on LHC-II causes a conformational change (Figure 3.2), reduces its affinity to PS-II and increases its affinity to PS-I. The increase in affinity to PS-I is due to not only the conformational change, but also the tight association between the phosphate group and a specific arginine residue in PsaL, a protein that facilitates docking of LHC-II to PS-I core complex. As a result, LHC-II migrates from PS-II towards PS-I, thus enhancing efficiency of PS-I relative to PS-II (Figure 3.3). This is supported by the findings that plants without either STN7 (Bellafiore et al. 2005) or Lhcb1+Lhcb2 (Andersson et al. 2003) do not undergo state transitions. It was later shown by Longoni et al. (2015) that only Lhcb2 phosphorylation is relevant to state transition (Lhcb1 phosphorylation causes its exclusion from the complex). The N-terminal region of phosphorylated Lhcb2 interacts with PsaL, PsaH and PsaO proteins, and PsaO bridges PsaA (PSI reaction centre) and Lhcb2 through contacts within the membrane and on the stromal and luminal surface. PsaO is associated with two chlorophyll molecules which facilitates resonance energy transfer from LHC-II to PsaA.

The phosphorylation can be reversed by a specific kinase called TAP38 (a.k.a PPH1). When PS-I is over-excited relative to PS-II, STN7 is inactive, and TAP38 dephosphorylates LHC-II and causes them to return to PS-II (Figure 3.3).

Figure 3.3: Oxidative status of plastoquinone determines the phosphorylation state of LHC-II, which in turn regulates relative activity of PS-I and PS-II.

3.3 Transcriptional Regulation

Regulation of relative activity of PS-I and PS-II can also be achieved by controlling the amount of transcription of PS-I genes (e.g. psaA) and PS-II genes (psbA).

Puthiyaveetil et al. (2008) described a sensor kinase called chloroplast sensor kinase (CSK) that couples oxidation state of PQ to transcriptional control of psbA. As shown in Figure 3.4, PS-I light causes the PQ pool to be oxidised, and then CSK, which is activated by PQ (the PQ analogue, DBMIB (2,5-dibromo-3-methyl-5-isopropyl-p-benzoquinone) binds to CSK similarly), autophosphorylates itself, and then phosphorylates sigma factor 1 (SIG1) subunit of PEP (RNA polymerase). This results in specific repression of PS-I genes (psaA), thus decreasing the stoichiometry of PS-I to PS-I. Conversely, CSK is inactive with a reduced PQ pool under PS-II light, and SIG1 phosphatase removes the repression of PS-I gene transcription, thus increasing the stoichiometry of PS-I to PS-II.

Figure 3.4: Transcriptional control of PS-I genes mediated by PQ, CSK, SIG1 and PEP.

2.1 Glycogen and Glucose

2.2 Fatty Acids

During exercise, the usage of fatty acids has an important role of delaying the onset of glycogen depletion and hypoglycemia. Its contribution to total energy expenditure increases as excercise proceeds. However, its usage is limited during strenuous exercise.

2.1 Adipocyte Lipolysis

The major source molecules for ketogenesis are fatty acids obtained via lipolysis in white adipose tissues. Lipolysis involves the sequential hydrolysis of triacylglycerol (TAG) into 3 non-esterified fatty acids (NEFA) and glycerol by adipocyte triglyceride lipase (ATGL), hormone-sensitive lipase (HSL) and monoglyceride lipase (MGL). Activation of ATGL and HSL are dependent on PKA, which is activated by cAMP that are produced when catecholamines bind to cell-surface \(\beta\) adrenergic receptors.

Insulin is a potent inhibitor of lipolysis. It activates PKB via IRS and PI3K, and PKB in turn phosphorylates and activates phosphodiesterase 3B (PDE-3B). PDB-3B decyclises cAMP, thus diminishing PKA, and thus lipolytic, activity (Figure 2.2). During starvation, insulin levels are low, and adipocyte lipolysis is released from inhibition. Adipocytes are then able to release NEFA into bloodstream.

Figure 2.2: Left, regulation of lipolysis by insulin in adipocytes; right, regulation of uptake of long-chain fatty acids into mitochondria by insulin.

2.2 Regulation of CPT1

Once NEFA molecules arrive at hepatocytes, they are first broken down into AcCoA via \(\beta\)-oxidation before conversion into ketone bodies. Long chain fatty acids needs the carnitine shuttle to enter mitochondria where \(\beta\)-oxidation takes place, and thus the first enzyme of this shuttling system, carnitine palmitoyl transferase 1 (CPT1) is the rate limiting step and is under control. CPT1 is inhibited by malonyl CoA, which is the lipogenetic precursor produced by acetyl-CoA carboxylase (ACC). Insulin activates protein phosphotase 2A (PP2A, which is also activated by xylulose 5-phosphate during fed state), which dephosphorylates and activates ACC, thus promoting lipogenesis and suppressing \(\beta\)-oxidation and ketogenesis (Figure 2.2).

During starvation when insulin levels are low, the high fatty acid concentration in hepatocytes activates the transcription factor PPAR\(\alpha\) (activated by direct binding of fatty acids to PPAR\(\alpha\)), which in turn promotes transcription of several enzymes involved in lipid catabolism, including CPT1.

2.3 Channeling of Acetyl-CoA into Ketogenesis

AcCoA produced from NEFA via \(\beta\)-oxidation can either enter the Krebs cycle or the ketogenesis pathway. When the rate of production of AcCoA exceeds the capacity of citrate synthesis (when there is insufficient oxaloacetate), the surplus becomes the substrate for ketogenesis. This situation occurs during starvation, when insulin levels are low and glucagon levels are high, and gluconeogenesis active. Glucagon has several mechanisms to promote gluconeogenesis (and suppress glycolysis). For example, it activates PKA, which in turn activates the transcription factor CREB, which promotes transcription of gluconeogenenic enzymes such as PEP carboxykinase and glucose 6-phosphatase. Insulin, on the contary, suppresses gluconeogenesis. For instance, it activates PKB, which in turn phosphorylates and thus inactivates FOXO1, which is a transcription factor that promotes the transcription of a similar set of gluconeogenic enzymes. Insulin inhibits gluconeogenesis and thus ketogenesis.

2.4 Mitochondrial HMG-CoA Synthase

Mitochondrial HMG-CoA catalyses the irreversible step in ketogenesis, and therefore is a site of control.

Insulin promotes glycolysis and increases flux through the Krebs cycle. One of the Krebs cycle intermediate, succinyl-CoA, inactivates mitochondrial HMG-CoA by succinylation. Glucagon, on the contrary, activates gluconeogenesis, which causes depletion of Krebs cycle intermediates including succinyl-CoA, thus activating ketogenesis.

In addition, the transcription factor PPAR\(\alpha\) described previously also activates transcription of HMG-CoA synthase.

3.1 Aetiology of Diabetes

Diabetes is characterised by insufficient insulin activity, which could be caused by either lack of insulin production or insulin resistance in target cells. Most diabetes cases fall into two categories, type 1 and type 2.

• Type 1 diabetes is caused by autoimmunity agaist pancreatic \(\beta\) cells, which leads to their death and inadequacy of insulin. The primary risk factor for type 1 diabetes is genetic, although several environmental risk factors, such as infection by viruses that show tropism to pancreatic islets.
• Type 2 diabetes begins with insulin resistance, but reduction of insulin production and number of \(\beta\) cells may also be implicated as the disease advances. Type 2 diabetes also has genetic risk factors, but environmental factors, such as sedentarity and obesity, are more important.

Whichever the type, ketogenesis is hyperactive in diabetes patients (although generally more severe in type 1) due to inadequate insulin activity.

3.2 Overproduction of Ketone Bodies

Given the fact that insulin suppresses ketogenesis and insulin activity is decreased in diabetic patients, it is easy to comprehend why ketone bodies are overproduced in diabetic individuals.

Lack of insulin activity upon adipocytes leads to uncontrolled lipolysis and subsequent release of non-esterified fatty acids. As fatty acids enter the liver, they are shuttled into mitochondria, broken down into acetyl-CoA via \(\beta\)-oxidation and converted to ketone bodies. This process is facilitated by the low insulin activity, which activates gluconeogenesis (e.g. via FOXO1-induced transcription of gluconeogenenic enzymes and phosphorylation of PFK2/FBPase2 bifunctional enzyme), which in turn releases the rate-limiting enzyme, HMG-CoA from succinylation inhibition and reduces usage of acetyl-CoA in Krebs cycle. The low insulin activity also inhibits ACC, thus activating CPT1 and hence \(\beta\)-oxidation. The high fatty acid concentration also leads to activation of PPAR\(\alpha\), which promotes transcription of ketogenic enzymes such as CPT-1 and HMG-CoA synthatase.

3.3 Consequences of Ketoacidosis

Diabetic ketoacidosis (DKA) the most common acute emergency in diabetes patients. It is a potentially lethal condition that is characterised by lowered blood pH due to accumulation of acidic ketone bodies (acetoacetate and 3-hydroxybutyrate). DKA is caused by lack of insulin and thus occurs mostly in uncontrolled type 1 diabetes (DKA is not commonly seen in type 2 diabetes, because even a tiny amount of insulin activity would inhibit ketogenesis; however, there is a small proportion of “ketosis-prone type 2 diabetes”). Often, DKA is triggered by stress or illness, which are associated with increase of ketogenesis-promoting hormones, glucagon and catecholamines.

Ketoacidosis has several consequences, as described below.

• Kussmaul breathing: initally, bicarbonate ions buffur protons dissociated from ketone bodies, but this buffering system quickly becomes overwhelmed. To compensate, blood CO₂ concentration is reduced by deep, regular breaths known as Kussmaul breathing.
• Hypovolaemia: hyperglycemia and ketonaemia results in excretion of glucose and ketone bodies (which normally should be absorbed) in urine. Since they are osmotically active, their presence in urine causes osmotic diuresis, which further leads to hypovolaemia (decrease in body fluid volume).
• Electrolyte disturvance: as the the plasma pH falls, extracellular protons are exchanged for intracellular potassium ions. In addition, owing to hypovolaemia, the aldosterone concentration increases to retain water, but this hormone also leads to excretion of K⁺ and retention of Na⁺.
• The low plasma pH impair with many biological functions, especially in the brain, where potentially life-threatening coma and cerebral oedema may occur.

2.1 Insulin Signalling

The insulin receptor belongs to the receptor tyrosine kinase (RTK) family. Upon insulin binding, the receptor dimer trans-autophosphorylates each other on tyrosine residues (pY). IRS1 (insulin receptor substrate 1) binds to pY via its SH2 (src homology 2) domain. Insulin receptor phosphorylates IRS1 on tyrosine, and then PI3K (phosphoinositide 3 kinase) binds to pY of IRS1 via SH2. PI3K converts PIP2 (PtdIns(4,5)P2) to PIP3 (PtdIns(3,4,5)P3), then PIP3 serves as the docking site for PDK1 (phosphoinositide dependent kinase 1) and PKB (Akt). PDK1 and mTOR2C activates PKB by phosphorylation, and PKB then phosphorylates other protein targets to bring about most insulin-induced metabolic changes. Among these effects, a significant porportion is mediated by PP1 (protein phosphatase 1), which is activated by PKB (Figure 2.1).

Figure 2.1: Overview of insulin signalling.

2.2 Glucagon Signalling

Glucagon receptors are G-protein coupled receptors (GPCR). Upon glucagon binding, the receptor acts as a guanine nucleotide exchange factor and replace GDP bound to the G protein with GTP. The activated heterotrimeric G protein dissociates into \(\beta\gamma\) and \(\alpha\) subunits. Some glucagon receptors are associated with G_s, and their \(\alpha\) subunit activates adenylyl cyclase, which produces cAMP, which in turn activates protein kinase A (PKA). Other glucagon receptors are associated with G_q, and their \(\alpha\) subunit activates phospholipase C, which converts PIP2 (phosphatidylinositol bisphosphate) to IP3 (inositol trisphosphate) and DAG (diacylglycerol). IP3 stimulates Ca²⁺ release from the smooth endoplasmic reticulum. PKA and Ca²⁺ then activates downstream proteins (Figure 2.2 ).

Figure 2.2: Overview of glucagon signalling.

3.1 Glycogen Metabolism in Liver

In liver, net synthesis of glycogen (in order to lower down blood glucose concentration) is stimulated by insulin and net consumption is promoted by glucagon.

3.1.1 Glucagon Activates Glycogenolysis in Liver

As described previously, glucagon signalling results in activation of PKA and release of Ca²⁺. PKA phosphorylates and thus activates phosphorylase b kinase. Ca²⁺ binds to phosphorylase b kinase, also activating it. Insulin inhibits glycogenolysis via PP1, which dephosphorylates phosphorylase kinase. Glucagon counteracts the effect of insulin in another way: PKA phosphorylates the R (regulatory) subunit of PP1, causing the dissociation of the C (catalytic) subunit, thus inactivating PP1 (Figure 3.1 ).

Figure 3.1: Effect of glucagon and insulin on hepatic glycogenolysis

3.1.2 Insulin Activates Glycogen Synthesis in Liver

Insulin activates glycogen synthesis in hepatocytes. Glycogen is synthesised by glycogen synthase (and branching enzymes), whose active (a) form is unphosphorylated. It is normally phosphorylated (inactivated) by caesin kinase 2 (CK2) and glycogen synthase kinase 3 (GSK3). Insulin-activated PKB phosphorylates (inavtivates) GSK3 and hence releases GS from inhibition (Figure 3.2).

Figure 3.2: Insulin stimulates glycogen synthesis in the liver.

3.2 Glycolysis and Gluconeogeneis in Liver

Insulin activates glycolysis and inhibits gluconeogenesis in liver. Glucagon has opposite effects.

3.2.1 Short-Term Control via PFK2/PBPase2 Bifunctional Enzyme

In liver, fructose 2,6-bisphosphate (F26BP) the key regulator of glycolysis and gluconeogenesis. It is an allosteric regulator of PFK-1 and FBPase-1 and have opposite effects on them. It significantly reduces the inhibitory effect of ATP on PFK-1—without F26BP, PFK-1 essentially has no activity in normal hepatocytes because hepatic ATP concentration is very high. In contrast, F26BP is an allosteric inhibitor of FBPase-1.

The PFK2/PBPase2 bifunctional enzyme (BFE) converts between fructose 6-phosphate (F6P) and F26BP. PFK2 is active in its unphosphorylated form while FBPase2 is active in its phosphorylated form. Insulin activates PP1, which dephosphorylates BFE (PP2A has the same effect), resulting in net production of F26BP, thus activating glycolysis and inhibiting gluconeogenesis. Glucagon activates PKA, which in turn phosphorylates BFE, reducing amount of F26BP and hence deactivating glycolysis and de-inhibiting (activating) gluconeogenesis (Figure 3.3 ).

Figure 3.3: Regulation of glycolysis and gluconeogenesis via F26BP in liver.

The isoform of pyruvate kinase (PK) present in liver is also phosphorylated and inactivated by PKA, and this also contributes to glucagon-induced inhibition of glycolysis.

3.3 Glucose Uptake into Muscle Cells

Insulin sigalling in muscle cells results in activation of PKB, which in turn causes translocation of GLUT4-containing vesicles to the cell surface. This significantly increases the rate of influx of glucose into muscle cells, thus reducing blood glucose concentration.

4.1 Lipolysis in Adipocytes

Lipolysis is the process by which triacylglycerol molecules (TAG) stored in the lipid droplet of adipocytes in white adipose tissues are hydrolysed to 3 free (non-esterified) fatty acid molecules and 1 glycerol molecule and released into the bloodstream, resulting in net increase in free fatty acid concentration in blood plasma.

As shown in Figure 4.1, PKA is central to the activation of the lipolytic pathway. Catecholamines (adrenaline and noradrenaline), being the major activators of lipolysis, activates PKA via \(\beta\) adrenergic receptor, G_s, AC and cAMP. Glucagon has also been shown to activate lipolysis in vitro (via PKA), but there is no clear evidence that it acts in vivo.

Figure 4.1: Activation of lipolysis by catecholamines (adrenaline and noradrenaline) and, with less evidence, by glucagon. PKA is activated via the GPCR-G_s-AC-cAMP pathway described previously. Activated PKA has two roles. First, it phosphorylates HSL and increases enzymatic activity thereof. Second, it phosphorylates perilipin, which is the protein residing on the membrane of the lipid droplet (LD) providing protection against ATGL and HSL. The phosphate groups induce binding of HSL to perilipin, effectively translocating HSL from the cytosol to the LD. Phosphorylation also release CGI-58 from perilipin, and the free CGI-58 activates ATGL. TAG is hydrolysed to 3 fatty acids and a glycerol molecule stepwisely by ATGL, HSL and MGL.

Insulin inhibits lipolysis via PKB and PDE-3B, resulting in net decrease in blood fatty acid concentration (Figure 4.2).

Figure 4.2: Inhibition of lipolysis by insulin. Insulin activates PKB, which phosphorylates and thus activates the phosphodiesterase PDE-3B. PDE-3B de-cyclise cAMP, to AMP, thus reducing PKA activity and the lipolytic pathway downstream of PKA.

4.2 TAG Synthesis in Liver

TAG synthesis in liver is promoted after a meal. This is mainly mediated by de-phosphorylation (activation) of ACC (acetyl-CoA carboxylase) by PP2A as a result of elevated glucose concentration and pentose phosphate pathway activity. (Some books, such as Lehninger’s, say that insulin-activated PP1 also de-phosphorylates ACC). The product of ACC, malonyl CoA, inhibits CPT1 (carnitine palmitoyl transferase) and hence oxidation of fatty acids.

Long term regulation of lipogenesis is achieved via the transcription factors ChREBP and SREBP-1c. These two TFs both up-regulates enzymes involved in lipogenesis, but their regulation differ. ChREBP is activated by xylulose 5-phosphate, which happens when there is a large flux through the pentose phosphate pathway due to high glucose concentration. This effect is reduced by cAMP, whose production is induced by glucagon. In contrast, SREBP-1c is activated by atypical PKC (aPKC), which is another protein kinase activated by insulin (Figure 4.3 ).

Figure 4.3: Long-term regulation of lipogenesis via ChREBP and SREBP-1c.

4.3 TAG (Re-)Synthesis in Adipocytes

During fed state, adipocytes, with help from endothelial cells, break down chylomicrons (from the intestine) and VLDLs (from the liver), hydrolyse the containing triacylglycerol molecules, take up the resulting fatty acids, and re-synthesise the TAG. Adipocytes do not express glycerol kinase, thus they have to obtain glycogen 3-phosphate, one of the substrate in the first step of TAG synthesis, from dihydroxyacetone phosphte (by glycerol 3-phosphate), which is a glycolytic intermediate. Insulin promotes glucose uptake and hence glycolysis in adipocytes, thus promoting TAG resynthesis.

3.1 Forward Genetics and Reverse Genetics

3.1.1 In Early Experiments, Forward Genetics Methods were Used Entensively

Studies on the molecular mechanisms of C. elegans development date back to the 1980s, when genetic manipulating techniques, especially gene cloning, were emerging, but the complete genome sequencing (of C. elegans and other model organisms) had yet to start. Therefore, those pioineering studies relied heavily on forward genetics, where worms were subjected to mutagens, and abnomal phenotypes were identified, isolated, amplified (by self-fertilisation) and the respective genotypes were determined. Differential interference contrast (DIC)/Nomarski microscopes were commonly used to examine the phenotypes because of their superior imaging quuality for live and unstained samples. There were also cellular methods for manipulating gene expression (which especially aided in eludidating cell-cell interactions). First, observing fates of the cells among varying individuals. Second, using laser beams to ablate cells suspected to interact with the cell of interest. Third, isolation of cells of interest, or rearrangement of their surrounding cells.

These extensive genetic screens led to identification of a number of genes that cause abormal cell lineage (the lin# genes) phenotypes, notably vulvaless (Vul)/Egg-laying defective (Egl) or multivulva (Muv) (Figure 3.1). Epistasis analyses of these phenotypes grouped these genes into distinct pathways (notably EGFR-Ras-MAPK and Notch) and helped elucidating where each component functions in the respetive pathway. The general idea is that, if the phenotype of gene A masks the phenotype of gene B, then gene A is likely to act downstream of gene B.

Figure 3.1: Egg-laying defective (Egl) and Multivulva (Muv) phenotypes caused by LET-23 EGFR-Ras-MAPK signalling mutants. (a): Wild type. Eggs are neatly arranged in the uterus. (b) and (c): Defective LIN-10 (which helps to localise LET-23) can lead to disorganisation of eggs (b) and the bag-of-worms phenotype (c), in which self-fertilised eggs hatch and larvae fills up their mother’s carcass. (d): let-60 gain-of-function results in Ras signalling in all VPCs, resulting in a Muv phenotype. Reproduced from 10.

Figure 3.2 shows the earliest expetiments that demonstrated the role of the AC and the LET-60 protein in vulval induction.

Figure 3.2: (a) in wild-type cells, signals from the AC results in the invariant VPC pattern 3\(^\circ\)-3\(^\circ\)-2\(^\circ\)-1\(^\circ\)-2\(^\circ\)-3\(^\circ\); (b) ablation of the AC renders a Vul phenotype, where all 6 VPCs adopts the 3\(^\circ\) fate, this shows that the AC positively regulates vulval induction; (c) inactive LET-60 (Ras) also renders a Vul phenotype even in the presence of the AC, therefore LET-60 should be downstream of the signal from the AC; (d) LET-60 hyperactivity leads to a Muv phenotype, further demonstrating LET-60 as a positive regulator of vulval induction.

3.2 Observing Cell Fates

3.3 Cross-Species Genomic Analysis Reveals Homology and Helps to Elucidate Roles of Proteins with Greater Accuracy

As sequencing techniques mature, complete genome sequences of more organisms are becoming available. This is accompanied by advance in computing power and bioinformatic algorithms, and they allow identification of a significant amount of homology among species.

The studies on C. elegans Ras signalling also benefitted from these developments. For example, MPK-1 was initially identified as a suppressor of Ras and was named SUR-1 before being identified as an Erk1/2 homologue. The scaffold protein KSR-1 was originally named SUR-3 due to the same reason. With homology available, one can easily guess the function of a novel gene/protein identified in one organism, if its sequence is found homologous with another gene with known functions.

3.4 Studying EGFR Localisation and Trafficking

The LET-23 EGFR upstream of LET-60 Ras has a polarised localisation in VPCs: vuval induction requires that EGFR to be expressed on the basolateral membrane. Immunostaining and live imaging can be used to monitor LET-23 trafficking and identify regulators involved.

3.4.1 Immunostaining

Immunostaining can be used to visualise subcellular localisation of components of regulators of the EDGR-Ras-MAPK pathway. This technique was used by 24 to establish the role of LIN-2/7/10 complex binding to LET-23 C-terminuc in regulating LET-23 basolateral localisation (Figure 3.3).

Figure 3.3: Top: LET-23 are localised to the basolateral membrane of VPCs in wild-type C. elegans; Middle: lin-7 mutant shows mislocalisation of LET-23 to the apical membrane; Bottom: let-23(sy1) mutant, in which a premature stop codon removes 6 C-terminal amino acids of LET-23, similarly resulted in mislocalisation of LET-23. LET-23 staining is shownin green, and staining for the cell junctions using the monoclonal antibody MH27 is shown in red. Reproduced from 24.

Teratogen and congenital anomaly

According to the US’ National Research Council, ~3% of live births suffer from major developmental aberrations, ~70% of neonatal deaths are due to developmental defects.

• exogenous agents that cause birth defects are teratogens, they can be chemicals, radiations, or viral infections, and so on
• thalidomide
  • used to treat anxiety
  • malformed arms and legs
  • thee first major evidence that drugs could induce congenital anomalies
  • led to the development of greater drug regulation and monitoring in many countries
• alcohol interferes with fetal neural development and is shown to cause mental retardation in the newborn
  • fetal alcohol syndrome
  • abnormal facial development
  • most prevalent type of congenital mental retardation syndrome in the US
• in pregnant women, Zika virus directly infects the neural progenitor cells of the fetal cortex (vertically transmitted infection), resulting in the death of these cells
• Model organisms such as Xenopus and Denio, are often used to screen compound that have a high probability of being hazardous, because the early development of these organisms rely on the same basic paracrine factors and TFs as humans do.
  • studies on model organisms have revealed three probable pathways for alcohol teratogenesis. first, generation of superoxide redicals; second, downregulation of Sonic hedgehog; third, intereference of the cell adhesion molecule L1

Nutrients

to sustain normal development of the fetus, healthy diet is needed. This is quite obvious, however, there is one vitamin, that cannot be obtained with enough amount just from food, it is folic acid

• Folic acid
  • it is found to significantly reduce the chance of neural tube defects in infants, but only at high doses
  • it’s recommended, by NHS, that all pregnant women should take a daily supplement of 400 migrograms of folic acid before they’re pregnant and during the first 12 weeeks of pregnancy, when the baby’s spine is developing

ES Cells

• ES cells are the natural source of pluripotent cells
• ICM (embryoblast) and trophoblast
• the study of the normal sequence of embryonic inductions yielded the methods for induced differentiation of these cells
• derived from the inner cell mass of blastocysts from IVF clinic
• harvesting human ES cells means destroying human embryos (5 days after fertilisation)
• raised ethical controversy

iPS cells

• Reverting to undifferentated state
  • functionallty equivalent to embryonic stem cells
  • capable of differentiating into almost all cell types
• Shinya Yamanaka factors: Oct4, Klf4, Sox2, c-Myc
  • screened 22 factors
• Regenerative Medicine
  • SMA clinical trial in Japan
• Organoids
  • An organoid is a miniaturized and simplified version of an organ produced in vitro in three dimensions that shows realistic micro-anatomy

1.1 Physical Principles

Electromagnetic waves (EM waves) has an electrical (E) component and a magnetic (M) component perpendicular to each other and both are perpendicular to the axis of propagation. Usually, only the E component is depicted for simplicity.

Most light sources have E components in all orientations, and when they pass through a slit, they become plane polarised, and can be represented by a simple sinusoidal wave. Adding up two plane polarised light perpendicular to each other produces interesting resultant waves, which can be visualised with plotly in R (Sievert 2018) (which is an uncommon use ‘invented’ by me¹). Assuming we are superposing two sinusoidal waves of the same amplitude and frequency with various phase differences, we can then compute the pattern of the resultant wave as a function of the phase difference:

library(plotly)
library(tibble)

calc_waves <- function(phase_diff = 0){
  # compute y and z values for given x values of the waves 
  # first arbitrarily choose evenly-spaced points on the x_axis
  x_axis = x=seq(1,20,0.1) 
  # compute the y, z values of the first sin wave on the xy plane
  sin1 <- tibble(x=x_axis,y=sin(x),z=0)
  # compute the y, z values of the second sin wave on the xz plane, 
  # taking the phase difference into account
  sin2 <- tibble(x=x_axis,y=0,z=sin(x + phase_diff))
  # compute the axes of the resultant wave
  # each y and z value is the sum of corresponding values in the 2 component waves
  resultant <- tibble(x=x_axis, y = sin1$y+sin2$y, z = sin1$z+sin2$z)
  
  return(list(sin1, sin2, resultant))
} 

plot_waves <- function(wave_list, components = c('sin1', 'sin2'), resultant = 'resultant'){
  # plotting
  plot_ly() %>% 
    add_trace(x=~x, y=~y, z=~z, name=components[1], data = wave_list[[1]], type='scatter3d', mode='lines') %>% 
    add_trace(x=~x, y=~y, z=~z, name=components[2], data = wave_list[[2]], type='scatter3d', mode='lines') %>% 
    add_trace(x=~x, y=~y, z=~z, name=resultant, data = wave_list[[3]], type='scatter3d', mode='lines')
}

When phase difference is zero, the resultant wave is also sinusoidal, and the plane in which it resides is oriented at 45\(^\circ\) to each component wave, as shown in Fig. 1.1 (this is an interactive HTML widget so you can rotate, drag and zoom! Unfortunately it cannot be rendered in LaTeX, so please visit ../tutorial/three-biophysical-methods.html if you are reeading the LaTeX PDF output).

calc_waves(phase_diff=0) %>% plot_waves()

As shown in 1.2, When the phase difference is \(\pi/2\), the resultant wave is helical and is said to be circularly polarised (for other phase differences, the pattern is helical but ellipised.

L <- calc_waves(phase_diff=pi/2)
plot_waves(L, resultant = 'resultant: left circularly polarised')

L <- L[[3]] # the third tibble represents the resultant, which will be used later

Note the handedness of the resultant wave: when the phase difference is \(\pi/2\), it is left-handed and it becomes right-handed when the phase difference is \(-\pi/2\). The latter is drawn in Fig. 1.3.

R <- calc_waves(phase_diff=-pi/2)
plot_waves(R, resultant = 'resultant: right circularly polarised')

R <- R[[3]]

As shown in Fig. 1.4 If we add up the left and right circularly polarised light, the resultant is a plane polarised light. Thus it can be said that a plane polarised light can be viewed as being made up of two circularly polarised components of equal magnitude and frequency.

plot_waves(list(L, R, 
  resultant=tibble(x=L$x, y = L$y+R$y, z = L$z+R$z)),
  components = c('left circularly polarised', 'right circularly polarised'),
  resultant = 'plane polarised')

CD measures differential absorption of these components. If the two components are absorbed to the same extent, clearly the resultant will still be planar as shown above. If there is differential absorption, as I simulate in Fig. 1.5, the resultant wave will be elliptical polarised. Here the R component is absorbed more than the L component, and the resultant ellipsed wave is left-handed. The CD signal (formally \(\Delta A=\Delta A_\text{L}-\Delta A_\text{R}\)) is generally reported in terms of the ellipticity, \(\theta\), of this resultant wave. \(\theta=\arctan(b/a)\) where b and a are the minor and major axes of the ellipse. \(\theta\) can be easily converted to \(\Delta A\) by the simple relationship: \(\theta=32.98\Delta A\). The CD spectrum is obtained when the CD signal \(\theta\) or \(\Delta A\) is measured as a function of wavelength.

L1 <- L %>% mutate(y = 0.9 * y, z = 0.9 * z) # 10% absorption of L
R1 <- R %>% mutate(y = 0.7 * y, z = 0.7 * z) # 30% absorption of R
plot_waves(list(L, R, 
  resultant=tibble(x=L1$x, y = L1$y+R1$y, z = L1$z+R1$z)),
  components = c('left circularly polarised', 'right circularly polarised'),
  resultant = 'elliptical polarised')

A CD signal will be observed when a chromophore is chiral (either intrinsically chiral, bonded to a chiral atom, or due to asymmetric enviornment). In proteins, such chromophores include the peptide bond (absorption below 240 nm), aromatic amino acid side chains (absorption in the range 260~320 nm) and disulphide bonds (weak absorptoin around 260 nm).

1.2 Experimental Setup

The EM wave used in CD is UV, usually with \(\lambda\) in the range 170~320 nm. Traditionally Xe arc lamps have been used as the light source of UV, but they can hardly achieve a wavelength below 180 nm. Now, high frequency UV can be generated by modern synchrotrons, and extending of CD data into the far UV region improves reliability of secondary structure prediction (see Section 1.3).

There are various methods by which the CD effect can be measured in a spectropolarimeter:

• modulation, in which the incident readiation is continuously switched between the L and R ocmponents
• direct subtraction, in whiich the absorbances of the 2 components are measured separately and subtracted from each other
• ellipsometric, in which the ellipticity of the transmitted radiation is measured

The modulation method is most commonly used. The experimental setup is described as follows:

• plane polarised light is split into the L and R components by passage through a modulator subjected to an alternating electric field
• the modulator normally consists of a piezoelectric quartz crystal and a thin plate of isotropic material (e.g. fused silica) tightly coupled to the crystal.
• the alternating electric field induces structural changes in the quartz crystal; these make the plate transmit circularly polarised light at the extremes of the field
• as the transmitted radiation of switched between L and R components, these are detecred in turn by the photomultiplier.

1.3 Application of CD

Far UV and near UV regions of CD give different information about a protein.

Absorption in the far UV region (170-250 nm) is mainly due to the \(\pi\rightarrow\pi^*\) and \(n\rightarrow\pi^*\) transitions in the peptide bonds, which is dependent on \(\Psi\) and \(\Phi\) torsional angles. Thus different types of secondary structures, such as \(\alpha\)-helices, \(\beta\)-sheets, \(\beta\)-turns, each have their characteristic CD spectrum (Fig. 1.6). These standard curves can be linearly combined to estimate the proportion of each secondary structure in a protein of interest.

Figure 1.6: Experimental setup of CD; characteristic curves of some secondary structures

The spectra in the near-UV region (260-320 nm) arise from the aromatic amino acids, each with a characteristic CD profile. The actual shape of the near UV CD spectrum will depend on the number of each type of aromic amino acid present, their mobility, their residing environment. Thus it can serve as a fingerprint of the of the tertiary structure of a protein.

Figure 1.7: The near UV CD spectrum of a dehydroquinase, labelled with contribution from each aromatic amino acid side chain.

By combining the far- and near-UV CD spectra, we can obtain a summary of the overall structural features of the protein of interest. Although it gives little insight into the precise 3D structure of the protein, it serves at a rapid way to detect conformational differences between two similar proteins in solution. Specifically, it can be used to:

• monitor the progress of protein folding (especially the detectiion of molten globule-like structures)
• compare the wild type and mutant forms of a protein
• confirm a modification (tagging) will not affect the protein’s native conformation and normal function
• assess thermal stability (unfolding at high temperature)
• show the formation of amyloid \(\beta\) protein in Alzheimer disease (Barrow et al. 1992)

CD can not only be applied to proteins but also to other chiral molecules. Such applications include:

• determination of nucleic acid conformations (A-RNA, A-DNA, B-DNA, Z-DNA)
• determination of nucleic acid-ligand interactions, e.g. between cationic porphyrins and DNA (Pasternack 2003)
• conformational study of biomolecular interaction with nanoparticles, where the degree of protein or nucleic acid denaturation is estimated (Liu and Webster 2007)

1.4 Other CD-based Techniques

The experiment described above is the conventional electronic circular dichorism (ECD). During the past decades, many other CD-based techniques have been developed to solve more specific questions. These include magnetic CDs (MCD, magnetic vibrational circular dichroism (MVCD), XMCD), fluorescence detected CD (FDCD), near-infrared CD (NIR-CD), vibrational CDs (VCD, FTIR-VCD), HPLC-CD, stopped-flow CD, and synchrotron radiation CD (SRCD).

2.1 Physical Principles

The the X-ray used in SAXS has wavelength of about the same size as the macromolecules in the sample, which allows formation of interference pattern (Fig. 2.1).

Figure 2.1: (A) for an array (usually a crystal) of small scatterers whose size is small compared to the wavelength of the incident light, an interference pattern can be observed due to the fixed path differences (and hence phase differences) between any two scatterers; (B) in the solution state of such small particles, their random orientation and movement make scattering to occur in all directions (so that a pattern cannot be observed); (C) when the size of the scatterers is comparable to or greater than the wavelength, angle-dependent scatterings due the nuclei/electrons of pairs of atoms within individual particles can thus be observed (larger angle, larger path difference).

2.2 Experimentation and Data Processing

SAXS uses a collimated monochromatic X-ray beam to illuminate the sample, and the intensity of the scattered X-rays a is recorded. The scattering of the pure solvent is also collected and subtracted from the sample solution scattering. The resulting 2D scattering pattern is translated into a 1D I vs. q relationship (where I is intensity and \(q = 4\pi\sin(\theta/2)/\lambda\)), and the data is transformed and plotted in a variety of ways.

Several parameters can be calculated from SAXS data, including molecular weight, excluded particle volume, maximum dimension \(D_\text{max}\) and the radius of gyration \(R_g\).

\(R_g\) can be directly extracted from SAXS data using the Guinier approximation, which states that, when the incident angle is small (approaching 0), the angular dependence of scattering can be described by the equation \(I(q)=I_z\exp(-q^2R_g^2/3)\), where \(R_g\) is the radius of gyration of the particle. When \(\ln(I(q))\) is plotted against \(q^2\) (\(\ln(I(q)) = \ln{I_0}-(R_g^2/3)q^2\)), \(R_g^2/3\) is the slope of the resulting straight line.

\(R_g\) provides a measure of the overall size of the macromolecule. It is the average root-mean-square distance to the centre of density in the molecule weighted by the scattering length density.

Some plots can emphasised sample flexibility, e.g. a Kranky plot (\(IQ^2\) agianst \(Q\)) can help identify an unfolded protein, as shown in Fig. 2.2

Figure 2.2: The Kratky plot

\(I(q)\) can be considered as a reciprocal space as a Fourier transform of \(p(r)\), which is the distribution of distances between pairs of atoms in real space. These two are related to each other by the equation:

\[p(r)=\dfrac{r^2}{2\pi^2}\int_0^\infty\dfrac{q^2I(q)\sin{(qr)}}{qr}ds\]

\(p(r)\) can be obtained from experimental data by indirect Fourier transformation. One such distribution is shown in Fig. 2.3.

Figure 2.3: Distance distribution function p(r) vs. r. Globular compact particles have a more symmetric distribution while unfolded particles have an skewed distribution with an extended tail.

To produce meaningful results, SAXS requires that the samples to be monodisperse (non-aggregated), which can be verified by dynamic light scattering (DLS) or analytical centrifugation (AUC).

3.1 Choice of Optical System

Depending on the strengths of the interaction probed (and the concentration of the sample), different optical systems are used:

• for weak interactions (\(10^{-4}<K_\text{d}<10^{-1}\)), higher concentrations (>5 mg/ml) are required and the Rayleigh interference optical system is the most appropriate
• for moderate-strength interactions (\(10^{-7}<K_\text{d}<10^{-4}\), 0.1~0.5 mg/ml), either interference optics or UV absorption optics can be used
• for strong interactions (\(K_\text{d}<10^{-7}\)), dye-labelled proteins and fluorescence optics are necessary

3.2 Experiments

3.2.1 Sedimentation Velocity (SV)

In the SV experiment, measurements on the absorption (\(A\), which is proportional to local concentration) at different radial distances (\(r\)) from the rotation centre are made at fixed time intervals (\(t\)), producing a series of s-shaped curves that shift to higher \(r\) values as sedimentation proceeds (Fig. 3.1.

The spinning rotor generates a sedimentation force on a particle of \(m\omega^2r\) (\(m\) = particle mass; \(\omega\) = angular velocity; \(r\) = distance from the centre of rotation). In solution, the particle displaces solvent, so the sedimentation force acts on an effective mass, \(m_\text{eff}=m(1-\bar v\rho)\) that is less than \(m\) (\(\rho\) = solvent density; \(\bar v\) = partial-specific volume (in ml/g)). \(\bar v\) is usually calculated (for proteins, \(\bar v\) lies in the range 0.70-0.75, leading to \((1-\bar v\rho)\) of around 0.27 (in water)). At terminal velocity (acceleration = 0), the sedimentation force is balanced by the frictional force and the velocity can be described by the equation \(v=m(1-\bar{v}\rho)\omega^2r/f\), where \(f\) is the friction coefficient, which is related to the diffusion coefficient, \(D\), by \(f=RT/N_\text{A}D\). The equation can thus be arranged to \(v=DM(1-\bar{v}\rho)\omega^2r/RT\). By defining the sedimentation coefficient as \(s=v/\omega^2r\), the Svedberg equation can be written as \(s=\dfrac{DM(1-\bar{v}\rho)}{RT}\).

The sedimentation coefficient \(s\), which is experimentally determined in SV has the unit Svedberg (S) where 1 S = 10^-13 seconds. The greater the molecular weight or more compact/spherical (less friction) the macromolecule is, the larger its \(s\) value. Usually the experimentally measured \(s_\text{exp}\) is standarised to \(s_{20, \text{w}}\), which is the value that would have been observed in water at 20\(^\circ\)C, using the relationship \(s_{20, \text{w}}=\dfrac{s_\text{exp}(\eta_\text{exp}(1-\bar{v}\rho_\text{20,w})}{(\eta_\text{20,w}(1-\bar{v}\rho_\text{exp})}\), where \(\eta\) and \(\rho\) refer to the viscosity and density of the buffer.

The Lamm equation, derived from Svedberg equation and Fick’s diffusion laws, describes the time dependence of the concentration:

\[\frac{\partial c}{\partial t}=D\left[\left(\frac{\partial^2 c}{\partial r^2}\right)+\frac{1}{r}\left(\frac{\partial c}{\partial r}\right)\right]-s\omega^2\left[r\left(\frac{\partial c}{\partial r}\right)+2c\right]\]

Solving this diffential equation leads to \(g(s)\), which is the distribution of \(s\). SEDFIT is one of the softwares specialised in solving this. SEDFIT can either give the uncorrected \(g(s)\) versus \(s\) profile, or it can give a distribution, known as \(c(s)\) vs. \(s\), which has been corrected for diffusion broadening (this assumes all particles have the same frictional ratio \(f/f_o\)). The \(g(s)\) vs. \(s\) or \(c(s)\) vs. \(s\) can further be converted to a molecular weight distribution, \(c(M)\) vs. \(M\), which is analogous to a mass spectrum (Fig. 3.1). SEDFIT is particularly good at evaluating homogeneity/heterogeneity of a prepation. Where the solution is heterogeneous, it can estimate the proportion of each sedimenting species and ascertain whether there is a reversible equilibrium.

Figure 3.1: Schematic of data processing in SV experiments. The raw data collected is a series of curves recorded at fixed intervals, each showing the variation of the absorbance with the distance from the rotation centre. By solving the Lamm equation, this is transformed to a g(s) vs. s plot, or c(s) vs. s plot if it is denoised by SEDFIT. It can be further transformed into a c(M) vs. M plot (distribution of molecular weights), assuming constant frictional ratio \(f/f_o\)

SV also gives information about the shape. The friction coefficient, \(f\), can be easily calculated from the terminal velocity (\(v=m(1-\bar{v}\rho)\omega^2r/f\)), and the friction ratio, \(f/f_o\) (where \(f_o=6\pi r\eta\)), shows diviation of the molecular shape from the sphere.

SV can also be applied for interaction analysis, and the simpest case is co-sedimentation. For example, the binding of adenosylcobalamin cofactor to the methylmalonyl-CoA mutase system from Propionibacterium shermanii was demonstrated by AUC with UV-absoption system. At a wavelength selected to detect the ligand only, in the presence of the mutase, all ligands sediment at the same rate as the protein, confirming the ligand is 100% bound.

In SV analysis, the emergence of new peaks at higher concentrations, shifts in the ratios of the peak areas, and/or shifts in peak positions are indicative of protein-protein interactions.

Fig. 3.2 shows an SV experiment which studies the binding between the Bacillus stearothermophilus 11-mer protein TRAP (trp RNA-binding attentuation protein) and Anti-TRAP (AT). As the TRAP:AT ratio increases froom 1:0 to 1:6, the peak representing TRAP shifts to higher s, indicating rapid and reversible binding between TRAP and AT (slow/irreversible interaction would result in separate peaks). The plot also shows that TRAP was saturated with AT at a 1:6 stoichiometry (as increasing ratio to 1:10 did not result in further shifts) and that the TRAP-AT complex was stable to excess AT (no negative feedback loop).

Figure 3.2: SV analysis of TRAP and AT. The TRAP concentratioin was fixed at 0.5 mg/ml, and varying amounts of AT (TRAP’s ligand) were added, with molar ratios of TRAP:AT labelled.

2.1 Chromosome Replication

Chromosome replication is principally regulated by DnaA in E. coli. DnaA initiates replication by binding its active, GTP-bound form (DnaA-ATP) to OriC. Its inactivation upon initiation of replication ensures one round of replication per cell cycle¹. Such mechanisms include competition for OriC binding by SeqA, repression of dnaA expression (also by SeqA), and inactivation of DnaA-ATP by the ATPase HdaA.

2.2 Chromosome Segregation

Segregation of chromosomes and plasmids is achieved while they are being replicated by using the Par system in some bacteria such as the budding bacteria Caulobacter crescentus, as illustrated in Figure 2.1. popZ proteins anchors the chromosome at its parS sequence (near oriC) to the old pole, which is mediated by parB. As replication starts, more parB binds to parS sequence of the newly synthesised chromosome, and is pulled to the new cell pole by the ATPase activity of ParA.

Figure 2.1: The Par system segregates replicated chromosomes in Caulobacter crescentus.

The mechanism of chromosome segregation in E. coli, which lacks Par, is not well characterised.

3.1 Z-ring Placement

The site of cell division in rod-shaped bacteria is usually at the centre of the cell, which is dependent on the corrent placement of the Z-ring. As shown in Figure 3.1, this is typically achieved using two mechanisms, nucleoid occlusion (NO), and the Min system, which are well characterised in E. coli and B. subtilis. It should be noted, however, that there is increasing evidence showing Z-ring positioning may be determined by other factors, and Min and NO systems may primarily function to ensure the efficient utilisation of this site. Such evidence include the observation of precise midcell Z-ring formation in Min- and Noc-null B. subtilis by Rodrigues and Harry (2012).

3.1.2 Nuclear Occlusion

Nuclear occlusion prevents Z-ring formation atop of nucleoids. It is mediated by SlmA in E. coli and Noc in B subtilis. Neither are normally essential in their respective organisms, but both are synthetically lethal with mutatations of Min, due to chaotic FtsZ assembly.

According to Adams, Wu, and Errington (2015), Noc has a sequence-specific DNA-binding domain and a membrane-associating domain. Noc is thought to oligomerise using their two dimerisation domains, bind to DNA at specific sequences, and then insert into the inner leaflet of the cytoplasmic membrane.

Figure 3.1: Negative regulation by nuclear occulation and the Min system ensures formation of the FtsZ ring in E coli.

3.2 Other Key Proteins in the Divisome

A fully functional divisome requires a combination of proteins in addition to the FtsZ ring. Most of these proteins have close association with the Z ring. In E. coli, It was thought that the recruitment of these proteins appears to have a linear dependecy pathway which starts with FtsA/ZipA and ends with FtsN, but a recent review by Du and Lutkenhaus (2017) shows evidence for the noon-sequential assembly of the divisome.

3.2.2 Downstream Proteins Recruited by the FtsZ Ring

Upon assembly of the FtsZ ring, other downstream proteins, notably those involved in peptidoglycan biosynthesis, are recruited to form the divisome.

FtsI, also known as penicillin binding protein 3 (PBP3), is one such protein in E. coli. It is required specifically for peptidoglycan synthesis at the septum (Chen and Beckwith 2001). Its recruitment is dependent on FtsK, which appears to be a bifunctional protein, with the C-terminal domain facilitating segregation of chromosome and the N-terminal domain carrying out a necessary, but undefined, function in septum development.

A summary of the divisome components is shown in Figure 3.2

Figure 3.2: Components of the divisome.

4.1 Alternative strategies of Binary Division

Binary division means dividing a bacterial cell into daughter cells with equal volume and physiology, which requires the FtsZ ring to form at the midcell. Apart form the NO and Min systems described above, there are other strategies to achieve this.

Figure 4.1: Alternative strategies of Binary Division

4.2 Positive Regulation in Myxococcus xanthus

The NO and Min systems described above are both negative regulatory mechanisms, where regions of low activity of relevant proteins allows for foramtion of FtsZ ring. There is evidence in some nonmodel organisms of positive regulation, where FtsZ ring formation is promoted by presence of certain proteins. This is exemplified by the deltaproteobacterium M. xanthus, in which the ParA-like protein PomZ localises to midcell in an FtsZ-independent manner before FtsZ ring formation.

4.3 Cell Division without FtsZ

In some species of Mycoplasma (which are well known for their small genomes), the FtsZ homologue is absent (Lluch-Senar, Querol, and Piñol 2010). Although the cell division mechanisms in these organisms remain elusive, they lead to reassessments on the role of FtsZ in FtsZ-dependent cell division. For example, as reviewed by Xiao and Goley (2016), it is likely that the main driving force for membrane invagination and constriction is not directly provided by FtsZ, but by the peptidoglycan synthesis.

1.1 Simple plasmids

Plasmids are extrachromosomal self-replicating cytoplasmic (usually circular) DNA elements found in prokaryotes and, less commonly, in eukaryotes. They come in various forms, from simple plasmids used for direct transformation and fosmids for phage transduction. Plasmids optimised for expressing recombinant proteins share several features, as illustrated in Fig. 1.1 and 1.2:

• An origin of replication (ori), which allows it to be duplicated in bacterial cells. The type of origin of replication affects the copy number. Eukaryotic vectors generally carry two oris, one bacterial (e.g. pUC ori) and one viral (e.g. SV40 ori). The bacterial ori is for amplifying the plasmid in bacteria before transfection, and the viral ori allows for episomal amplification of plasmids in the eukaryotic host.
  
• A cloning site where the GOI is inserted. Typically it is crowded with restriction enzyme cutting sites (compatible with RE/ligase dependent cloning), but it can also be made compatible with other cloning methods, such as the LIC-compatible pBLIC-puro¹ plasmid developed by Patel et al. (2012).
• A promoter and associated operator/enhancer. The promoter sequences for bacteria and eukaryotes are different (-35/-10 consensus sequence v.s. TATA box) For a bacterial vector, the promoter is usually controlled by an inducible operator so that protein expression is induced manually only when the optimal bacterial density is reached (e.g. lac operator induced by IPTG). In contrast, for a eukaryotic vector, the promoter is usually made as processive as possible, because the natural rate of transcription is generally low, and cell growth and division are not important (they do not serve to amplify the plasmid but to express the protein). For example, the pEF-BOS vector developed by Mizushima and Nagata (1990) contains the promoter of EF-1\(\alpha\) (EF-1\(\alpha\) is one of the most abundant proteins in eukaryotic cells, so its promoter is highly processive). Recently, promoters of viral origin (e.g. CMV, SV40) are gaining popularity.
• A transcription terminator/polyA signal. The mechanisms of transcription termination in bacteria and eukaryotes differ. In bacteria, the terminator sequence consists of a symmetric DNA sequence, and its transcribed product folds into a hairpin structure, releasing RNA polymerase (RNAP) from DNA. In eukaryotes, CstF and CPSF bound on RNAP C-terminal domain recognise 3’-end processing (polyadenylation) sequence (AAUAAA and GU-rich sequence) and induce cleavage and polyadenylation.
• A selectable marker. Usually an antibiotic resistance gene (bacteria and eukaryotes are sensitive to different antibiotics). Fluorophore genes are also common in eukaryotic vectors.
• Some plasmids possess a start codon and associated Shine-Dalgarno sequence (for bacteria) or Kozak sequence (for eukaryotes).
• Some plasmids have tag sequences near the start or end of the ORF for easy purification of the recombinant protein by affinity chromatography (e.g. hex-His, GST, HSV) and associated cleavage sequence (e.g. TEV) for easy removal of the tag.

Figure 1.1: The bacterial expression vector pETBlue-2

Figure 1.2: The eukaryotic expression vector pCMV-GFP-LC3

1.2 Other bacterial vectors

There are other forms of circular DNA, each with different disirable properties (larger insertion size, low copy number, phage-compatibility, etc.). They are sometimes called ‘high capacity vectors’ because their insertion sizes are greater than simple plasmids.

Bacterial artificial chromosomes (BACs) contain some regions derived from a special plasmid called the F (fertility) factor: the region containing the origin of replication as well as genes that ensured its precise segregation during bacterial cell division. A great advantage of BAC vectors is the large insertion size (100-200 kb). But the very large insertion can be a problem, in that it cannot be manipulated by restriction enzymes. Instead, in vivo homologous recombination-based strategies are used to insert DNA fragments into BACs. (Fig. 1.3)

Figure 1.3: Making a BAC

Fosmids are hybrids of \(\lambda\) phage DNA and bacterial F plasmid DNA. Fosmids are packaged into \(\lambda\) phage particles, which delivers them into bacterial cells. Due to the presence of the F plasmid origins of replications, fosmids are maintained at a very low copy number (normally single copy).

Fig. 1.4 is a schematic of 6 high capacity vectors (Saraswathy and Ramalingam 2011).

Figure 1.4: Schematic of 6 high capacity vectors.

1.3 Other Eukaryotic Vectors

Yeast artificial chromosome (YAC) has a cloning capacity up to 3000 kbp. It is introduced into the yeast cells by electroporation, then it is maintained as a linear DNA-like chromosome. It is replicated along with other chromosomes in yeast and its copy number of one is maintained after cell divition.

Some viruses (e.g. Adenovirus, Lentivirus and Baculovirus) and bacteria (e.g. Agrobacterium tumefaciens) are reliable vectors for stable transfection of eukaryotic cells (see section 3.2.3)

2.1 Simple plasmids (for both bacteria and eukaryotes)

Traditionally, restriction endonucleases and T4 ligase are used to insert GOI into pladmids. This method has several drawbacks:

• the association between the short 2-4 nt overhangs (sticky ends) is weak
• each DNA insert has to be inspected for any internal restriction sites
• due to the above two reasons, it has very limited capability of constructing multi-fragment plasmids
• incomplete DNA digestion and poor ligation yields
• unwanted amino acids can be introduced to the expressed protein
• the GOI can only be cloned in the vector position where the selected restriction site is present

Ligation-independent cloning (LIC) overcomes many of the problems described above. In a T4 DNA polymerase-dependent approach, the sticky sequences are made by the 3’-to-5’ exonuclease activity of T4 polymerase. These sequences are long, allowing formation of stable recombinant plasmid without in vitro ligase treatment. The nicks in sugar-phosphate backbone are later fixed by the host’s ligase. As T4 polymerase (exonuclease activity) always proceeds from 3’ ends, any internal sticky sequences will not be disrupting and thus the custom stiky sequence can be used for any DNA inserts.

Fig. 2.1 compares the mechanisms of the traditional RE/ligase-based cloning and the T4 polymerase-dependent LIC approach.

Figure 2.1: Left, traditional cloning based on restriction endonuclease and ligase; right, ligase-independent cloning based on T4 polymerase

There are also other alternatives to the traditional method:

• the Gateway cloning system exploits the site-specific recombination system utilized by bacteriophage lambda to shuttle sequences between plasmids bearing flanking compatible recombination attachment (att) sites
• Gibson assembly is a molecular cloning method which allows for ‘stitching’ multiple overlapping DNA fragments together in a single, isothermal reaction, using exonuclease, polymerase, and ligase.
• TOPO cloning expoits Taq polymerase’s feature that adds non-specific A to the 3’ end, and uses topoisomerase I to hold vectors open and to promote ligation

2.2 Other vectors for bacteria

BACs and fosmids: see section 1.2

2.3 Viral vectors for eukaryotes

See section 3.2.3.1

3.1 Transformation of bacterial cells

3.2 Transfection of Eukaryotic Cells

Transfection of eukaryotic cells can be transient or stable. Transient gene expression is generally used in academic settings, when the product of the GOI (usually a protein) is used for short-term research purposes. Stable transfection is usually used in indusdries, but it is also used in academia for specific purposes.

3.2.3 Biological methods

The physical and chemical methods described above are generally used for transient transfection. Although they may produce stably transfected cells (when the rare homologous recombination events occur), the probability is very low and the GOI is inserted randomly and is thus not expressed with greatest efficiency (the chances are low that the GOI is inserted into a highly ‘transcriptionally open’ region).

Biological methods significantly improve the chance of integration of GOI into the host’s genome, and can even direct the site of insertion.

3.2.3.1 Lentiviral transduction as an example of viral transfection (Merten, Hebben, and Bovolenta 2016)

Lentiviruses (LV) carry a genome made of (+) strand RNA. Upon infection, the viral RNA genome is reverse-transcribed and the cDNA is stably integrated into the host’s genome. Unlike \(\gamma\)-retroviruses, LVs exploit active nuclear transport and thus do not rely on cell division to access the nucleus. Fig. 3.1 is an example of the third generation HIV-1 vector system.

Figure 3.1: Production of recombinant LV particles and their infection. Production of recombinant LV particles is usually accomplished by transfection of HEK293T cells with a set of three helper plasmids and additionally a LV vector plasmid that encodes the GOI as well as optional reporter (e.g resistance, fluorophore) genes. LV particles released from HEK helper cells infect target cells and integrate the GOI into their genome.

3.2.3.2 Ti plasmid and Agrobacterium tumefaciens infection

Agrobacterium tumefaciens is the bacterium that causes crown gall disease in plants. When the it infects a plant cell, a part of its Ti plasmid, called T-DNA, is transferred and inserted (at random) into the genome of the plant cell. Thus, the T-DNA can be replaced with any GOI for stable transfection of plant cells.

3.2.3.3 CRISPR-Cas9 Transfection

The CRISPR-Cas9 system comprises a short guide RNA (gRNA) and the Cas9 nuclease. The gRNA guides Cas9 to a specific gene locus that is complementary to its crRNA portion, then Cas9 makes a double strand break, which is dependent on the PAM (protospacer adjacent motif) sequence immediately downstream of the recognition sequence.

Following DNA cleavage, the break is repaired by either non-homologous end joining (NHEJ) or homology directed repair (HDR). It is the HDR mechanism that may cause integration of GOI into the cleavage site.

Figure 3.2: gRNA directed DNA cleavage by Cas9

The Ribosome

The bacterial 70S ribosome comprises two subunits, the 50S (large) and the 30S (small) subunit. The 30S subunit mediates selection of cognate aminoacyl tRNAs and the 50S subunit contains the peptidyl-transferase centre (PTC), which catalyses addition of new amino acids to the elongating polypeptide chain.

The process of protein synthesis

The 3 phases of protein synthesis are initiation, elongation and termination.

During initiation, the 3’ end of the 16S rRNA of the 30S subunit base-pairs with the Shine-Dalgarno sequence, which positions the start codon (usually AUG) in the P site for the binding of the initiator tRNA (usually fMet-tRNA) and IF1, 2, 3. Joining of the 50S subunit and dissociations of IFs primes the elongation phase.

The elongation phase, where new amino acids are appended to the growing polypeptide, involves 4 steps:

1. Decoding. An aminoacylated tRNA (aa-tRNA) is delivered to the A-site of the ribosome by elongation factor EF-Tu complexed with GTP. GTP hydrolysis facilitates discrimination between cognate and non-cognate tRNAs.
2. Peptidyl transfer. The peptidyl-transferase centre (PTC) on the 50S subunit catalyses the formation of peptide bond between the amino acids attached to the tRNAs in the A- and P-sites by transferring the polypeptide chain from the P-site tRNA to the aa-tRNA in the A-site.
3. Translocation. EF-G catalyses the movement tRNAs from the A and P sites to the P and E sites
4. the growing polypeptide chain passes through an exit tunnel

Finally, when the stop codon (UAA/UAG/UGA) is encountered, it is recognised by release factors (RF1/RF2) that forces hydrolysis of the peptidyl-tRNA bond in the P-site, thus releasing the polypeptide chain from the ribosome.

NMR

Protein folding is a multistep process that proceeds through intermediate states and studying these intermediates is difficult due to their transient nature, low populations under non-denaturing conditions, and difficulty of their isolation.

Over the past two decades, many NMR methods have been developed and some became invaluable tools in studying molecular details of protein folding. Tansverse relaxation optimized spectroscopy (TROSY) is the first method developed for probing of dynamics and interactions of large (up to 1MDa) protein assemblies. Relaxation dispersion (RD) and saturation transfer (ST) methods provide a detailed look into the pathways of biomolecular processes, allowing studies of transient intermediates during protein folding.

RD and ST methods allow studying the minor nonnative conformations during protein folding with short lifespans (often \(\mu\)s-ms). The two major types of RD experiments include (1) Carr-Purcell-Meiboom-Gill (CPMG) methods that exploit modulation of R_ex by a sequence of evenly spaced refocusing pulses, and (2) rotating frame \(\text{R}_{1\rho}\) relaxation experiments that use modulation of R_ex by an on- or off-resonance continuous wave (CW) RF field. The ST experiments exploit modulation of R_ex by a weak RF field, and are conceptually similar to off-resonance \(\text{R}_{1\rho}\) measurements. Transient nonnative protein states can be studied by other NMR experiments, such as paramagnetic relaxation enhancement.

Dynamics of protein folding, on the time scale of seconds, can be studied by other methods. Changes in NMR spectra can be monitored in real-time. Logitudinal magnetization (ZZ) exchange experiments can probe interconversions between states with comparable populations. Hydrogen exchange measurements detect evanescent populations of disordered nonnative states transiently sampled by proteins under native-like conditions.

Other Methods

There are some methods that can assess the extent of folding qualitatively, but they do not give as much quantitative information as NMR does.

CD (Circular Dichroism)

Plane-polarized light is the sum of two circularly polarized beams, L and R, which are rotating in opposite directions. CD arises from differential absorption of L and R component beams in a smaple of chiral molecules. CD is detected by a double beam instrument with separate L and R paths (which are usually produced by high frequency modulation of a photoelastic modulator (PEM) device). The high intensity light source is produced by synchrotron.

The CD spectrum shows the variation of \(\Delta\epsilon\) (difference between the extinction coefficient of the L and R beams) with wavelength (\(\lambda\)). Because of the chiral nature of amino acids (except glycine), peptides are optically active. The CD spectra in the 170-250 nm region are distinct for proteins with different structure. Standard curves for different secondary structures (\(\alpha\)-helix, \(\beta\)-sheet, \(\beta\)-turn, and random coil) are available (Figure 2) and the proportion of different secondary structures of a protein is calculated by linear combination of the standard curves.

Figure 2: CD components and standard curves

DSC (Differential Scanning Calorimetry)

DSC determines the heat capacity (the difference between the heat capacity in the sample cell and in the reference cell), C_p, of a molecule in aqueous solution, as a function of temperature. This is done by increasing cell temperatures while keeping the two cells at the same temperature and recording the power supply throughout the experiment. The variation of \(C_\text{p}\) with temperature is plotted. On such a curve, there is a peak if the sample molecule undergoes state changes within the temperature (Figure 3). For a protein, the peak corresponds to the melting temperature, \(T_\text{m}\), which occurs when the concentrations of its native state ([N]) and the denatured state ([D]) are equal. Stabilization of protein, for example by altering the pH and forming complexes, increases \(T_\text{m}\)

Figure 3: Principles of DSC

1.1 ESI-MS (Electrospray Ionization Mass Spectrometry)

In mass spectrometry, the analytes are first ionised in vacuum and these charged molecules are passed into a electric/magnetic field, and their path through the field can be used to deduce their mass/charge (m/z) ratio (using deflection, TOF, quadrupole, or other analyzers).

Traditional ionisation methods were not suitable for biological macromolecules because the need for heating (or other treatment) to achieve gas phase would cause rapid decomposition of the molecules. Later, MALDI (matrix-assisted laser desorption/ionisation) and ESI (Electrospray Ionization) techniques were deveoped and they are suitable for this purpose. In MALDI, proteins are placed in a light-absorbing matrix, then ionisation and desorption is triggered by a short pulse of laser light. In ESI, a solution of analytes is passed through a charged needle kept at a high electrical potential, dispersing the solution into a microdroplets. The solvent around macromolecules rapidly evaporates, leaving charged molecules in gas phase. ESI can directly accept inputs from many other purification methods such as SDS-PAGE and chromatography and in addition it is good for detecting native states and different conformations.

The sequence of a protein is often too long to be obtained by MS at one time, so protease can be used to break peptides into small fragments which are sequenced individually. In addition, tandem MS (MS/MS) is can be used, in which one peptide is first analysed by one mass analyzer (MS1), and then further fragmented by a ‘collision gas’ such as He and Ar, and the m/z ratios of these fragments are analysed by MS2. This produces a spectrum with many peaks (greater m/z correspond to longer fragments), and the successive peaks differ by the m/z of a particular amino acid in the original peptide. This information can be used to deduce the original sequence.

2.1 CD (Circular Dichroism)

Plane-polarized light is the sum of two circularly polarized beams, L and R, which are rotating in opposite directions. CD arises from differential absorption of L and R component beams in a smaple of chiral molecules.

CD is detected by a double beam instrument with separate L and R paths (which are usually produced by high frequency modulation of a photoelastic modulator (PEM) device). The high intensity light source is produced by synchrotron.

The CD spectrum shows the variation of \(\Delta\epsilon\) (difference between the extinction coefficient of the L and R beams) with wavelength (\(\lambda\)).

Because of the chiral nature of amino acids (except glycine), peptides are optically active. The CD spectra in the 170-250 nm region are distinct for proteins with different structure. Standard curves for different secondary structures (\(\alpha\)-helix, \(\beta\)-sheet, \(\beta\)-turn, and random coil) are available (Figure 2.1) and the proportion of different secondary structures of a protein is calculated by linear combination of the standard curves.

Figure 2.1: CD components and standard curves

2.2 DSC (Differential Scanning Calorimetry)

DSC determines the heat capacity (the difference between the heat capacity in the sample cell and in the reference cell), C_p, of a molecule in aqueous solution, as a function of temperature. This is done by increasing cell temperatures while keeping the two cells at the same temperature and recording the power supply throughout the experiment. The variation of \(C_\text{p}\) with temperature is plotted. On such a curve, there is a peak if the sample molecule undergoes state changes within the temperature (Figure 2.2). For a protein, the peak corresponds to the melting temperature, \(T_\text{m}\), which occurs when the concentrations of its native state ([N]) and the denatured state ([D]) are equal. Stabilization of protein, for example by altering the pH and forming complexes, increases \(T_\text{m}\)

Figure 2.2: Principles of DSC

2.3 DLS (Dynamic Light Scattering)

Random motion of macromolecules in a suspension causes fluctuations in local concentration and thus local variarions in refractive index and intensity of the scattered light. These time-dependent fluctuations can be analyzed by a coherent laser source. The observed fluctuations give rise to diffusion coefficients, \(D\), which is related to the size of the molecules (large particles diffuse more slowly). DSL can measure polydispersity and the presence of aggregates in protein samples.

2.4 AUC (Analytical Ultracentrifuge)

An AUC is a specialised ultracentrifuge equipped with absorbance and interference detection systems. Each cell contains a sample meniscus and a reference cavity, allowing the absorbance of the solvent to be corrected.

Two types of experiments are typically performed by AUC: sedimentation velocity and sedimentation equilibrium. Sedimentation velocity is performed at high speed, which depletes particles from the centrifuge cell and creates a pellet at the bottom of the cell. It can give information on molecular shape, mass, and interactions with themselves and with other components.

Sedimentation equilibrium is perfomed at low speed and does not create a pellet, and the sedimentation of molecules down the centrifuge cell is balanced by their diffusion back up the cell. The main application of sedimentation equilibrium is the detection of complexes and self-association and the quantification of binding between species

3.1 Fluorimetry/Fluorescence anisotropy

If fluorophores are excited with plane polarized light and the fluorescence is observed through analyzing polarizers, the fluorescence is also polarised.

The fluorescence anisotropy is defined as \(A=\dfrac{I_\parallel - I_\bot}{I_\parallel+2_{\bot}}\), where \(I_\parallel\) and \(I_{\bot}\) are the fluorescence intensities polarised parallel and perpendicular to the direction of the excitation beam. \(A\) is a direct measure of the molecular rotation in solution and can be used to study complex formation, as a macromolecule will rotate more slowly when it is in a complex thatn when it is alone.

3.2 Surface Plasmon resonance

Surface plasmon resonance (SPR) is used for measuring molecular interactions between a pair of molecules.

A surface plasmon is an electron oscillation generated at a surface interface between a metal and a dielectric. A plasmon resonance occurs when EM wave in visible light couples optimally with the oscillating electrons in the metal, and this results in a maximal reduction in the reflected light intensity. The resonance angle, \(\theta_\text{spr}\), is found by changing the angle of incidence of the light beam, giving a dip in a plot of intensity against angle. \(\Delta\theta_\text{spr}\) is sensitive to changes in the refractive index of the medium near the metal surface and this is a measure of the mass change at the sensor surface (in the evanescent region).

In an SPR experiment, one type of ligand is immobilised at the sensor surface, and a the analyte is passed through the cell. If the ligand binds to any binding partner in the analyte, \(\Delta\theta_\text{spr}\) would increase. Then, non-specific binding is washed off by buffer, and \(\Delta\theta_\text{spr}\) would decrease and \(\Delta\theta_\text{spr}\) due to specific binding can be found. Finally, regeneration solution is applied to remove all binding and reset \(\Delta\theta_\text{spr}\) to zero. (Figure 3.1)

Figure 3.1: Principles of SPR

SPR can be used to determine \(K_d\) of complex formation, which equals \(\dfrac{k_{-1}}{k_1}\) where \(k_1\) is the rate of association and \(k_{-1}\) is the rate of dissociation. \(k_1\) and \(k_{-1}\) can be deduced from the plot.

3.3 Isothermal titration calorimetry (ITC)

ITC measures heat changes when a complex is formed at constant temperature.

In ITC, an insulated reaction cell containing protein is kept at a temperature (usually 8\(^\circ\text{C}\) above the environment) which is equal to the temperature of a reference cell, and the reference cell is kept at a constant temperature by a thermostat. Then, increasing amounts of ligand is added into the chamber, and they form complexes with the protein, which can be exothermic or endothermic. The heat change is compensated by a power supply, which can be converted to \(\Delta H\) of the reaction. As more ligands are added, proteins become saturated and \(\Delta H\) approaches zero. The raw data obtained (power supplied to compensate the heat change caused by each addition of ligands) can be integrated and corrected to give a plot of \(\Delta H\) against the molar ratio of the ligand and the protein, and \(\Delta H\), K_d and stoichiometry can be inferred from the curve. Subsequently, \(\Delta G\) and \(\Delta S\) can also be calculated (Figure 3.2).

Figure 3.2: ITC data manipulation

3.4 MST (Microscale Thermophoresis)

MST is a relatively new method for analysing interactions of proteins or small molecules in complex bioliquids such as blood serum or cell lysate. The technique depends on the phenomenon that molecules move within temperature gradients (thermophoresis).

The instrument uses an IR laser to create a temperature gradient, and the movement of the molecules within this gradient is monitored by fluoresence. The fluoresence can either be intrinsic (due to tryptophan) or extrinsic (attached dye or fluorescent protein).

Compared with traditional methods for studying protein interactions, MST has several advantages:

• A minuscule amount of sample (a few \(\mu\)l) is needed,
• no limitations on size and affinity
• no limitations on buffer; tolerates impurity; can be used in in complex bioliquids such as blood serum or cell lysate
• no need for immobilization

Figure 3.3: Sample data produced by MST

Effect of AMPK on secreted A\(\beta\) levels in neuronal cultures

1. In ELISA (enzyme-linked immunosorbant assay), antibodies are used to bind an antigen of interest (usually protein) with high specificity, and the presence and concentration of the antigen is reported by the extent of reaction of the enzyme conjugated with the antibody. Using two types of antibodies binding to different regions of the antigen allows sandwich ELISA (see below), which has greater specificity. Horseradish peroxidase converts its substrate to a coloured product, and the intensity of the colour, which is related to the concentration of A\(\beta\) peptide (the antigen), can be measured by a spectrophotometer as absorbance of EM wave at a certain visible wavelength (the relationship is not always linear so a calibration is needed).

2. By plotting ‘measured value’ against the number of dilution (the power to which \(\frac{1}{2}\) is raised), it can be shown that the realtionship between them is roughly linear when measured value is around 100.

standard <- c(206, 215, 201, 199, 205, 197, 199, 202, 190, 120, 67, 41, 27, 13, 8, 8)
untreated <- c(211, 197, 200, 202, 198, 195, 197, 164, 88, 47, 27, 15, 6, 9, 7, 7)
AICAR <- c(207, 202, 205, 199, 189, 200, 183, 101, 58, 30, 18, 12, 7, 7, 8, 6)
pow <- 0:15

df <- tibble(pow, standard, untreated, AICAR)
df %>% gather(standard, untreated, AICAR, key = 'treatment', value = 'measured value') %>% 
  ggplot(aes(pow, `measured value`, color = treatment, shape = treatment, linetype=treatment))+
  geom_point()+xlab('number of dilution (i.e. number of times the concentration is halved)')+
  geom_smooth(se=FALSE, method='loess', span=0.4, size=0.5)+
  geom_hline(yintercept = 100)+
  scale_x_continuous(breaks = 0:15)

## `geom_smooth()` using formula 'y ~ x'

Reading off from the plot:

• for the standard solution, a \(\dfrac{1}{2^{9.5}}\) dilution corresponds to 100 units. Thus, 100 units correspond to \(10\times \dfrac{1}{2^{9.5}}=0.01381\text{ ng/ml}\)
• for the untreated sample, a \(\dfrac{1}{2^{7.9}}\) dilution corresponds to 100 units, or \(0.0138\text{ ng/ml}\), so the undiluted sample has a A\(\beta\) concentration of \(0.01381\times2^{7.9}=3.299\text{ ng/ml}\). The volume of 10⁶ cells is 100\(\mu\)l, so the level of secretion is \(3.299\times0.100\times 1000=329\text{ pg/10}^6\text{ cells/day}\)
• for the sample with AICAR, the undiluted sample has a A\(\beta\) concentration of \(0.01381\times2^{7.2}=2.031\text{ ng/ml}\). The level of secretion is \(2.031\times0.100\times 1000=203\text{ pg/10}^6\text{ cells/day}\)

3. Upon activation by AICAR, AMPK reduces secretion of A\(\beta\)

Effect of added A\(\beta\) on AMPK activity

4. Triton X-100 is a detegent, disrupting the cell membrane and lysing the cell; PMSF inhibits serine proteases, preventing degradation of target protein; protein A, which is immobilised on sepharose, binds specifically to the F_c region of anti-AMPK antibody

5. This ensures that all measured enzymatic activity is due to AMPK but not other kinases that would be present in the whole cell extract.

6. Substituting \(t_{\frac{1}{2}}=14\text{ days}\) and \(t = 14\) into \(A=A_0e^{-kt}=A_0e^{-\dfrac{\ln{2}t}{t_{\frac{1}{2}}}}=A_0\times(2)^{-\dfrac{t}{t_{\frac{1}{2}}}}\), \(A=A_0\times(2)^{-\frac{9}{14}}=0.6404A_0\), i.e. \(A_0=1.5614A\), where \(A_0\) is the original activity and \(A\) is the activity after 9 days. \[\begin{aligned}A_0 \text{ cpm}\div80\%\div(2.2\times10^{12})\text{ cpm/Ci}\div(0.5\times10^{-3})\text{ Ci/nmol}\div(50\times10^{-3}\text{ mg cell protein})\div20\text{ min}\div\frac{50}{50+5+5}\\=x \text{ nmol substrate phosphorylated/min/mg cell protein}\end{aligned}\] where x is the final answer. This simplifies to \[x=1.3636\times10^{-9}A_0=2.129\times10^{-9}A\]
   • Activity (+A\(\beta\)): \(2.129\times10^{-9}\times(11800-278)=2.45\times10^{-5} \text{ nmol substrate phosphorylated/min/mg cell protein}\)
   • Activity (no A\(\beta\)): \(2.129\times10^{-9}\times(2940-278)=5.67\times10^{-6} \text{ nmol substrate phosphorylated/min/mg cell protein}\)

7. AMPK activity increases as A\(\beta\) is added, and AMPK may help to degrade A\(\beta\) according to (c).

Association of \(\tau\) protein with microtubules

8. tau alone is soluble and exists solely in the supernatant. Unphosphorylated tau binds to microtubules strongly and exists entirely in the pallet (insoluble). Phosphorylated tau has less affinity to microtubules and most of them are found in the supernatant.

Phosphorylation sites on tau

1. • Induce mutations on different positions and observe their effect on phosphorylation. The mutated position(s) can be determined by DNA sequencing and the effect of phosphorylation can be determined by isoelectric focusing (phosphorylated protein has a more negative charge).
   • Fragment the protein and test each piece for phosphorylation.

10. These phosphorylation sites are all located on the MT binding domain (235-368) and thus phosphorylation can directly (at least sterically) hinder binding to MT. In addition, MT binding domain contains many postively charged lysine (K) residues and few negatively charged (D and E) residues. This suggest that the binding between tau and MT is electrostatic (+ve on tau, -ve on MT). Phosphorylation adds -ve charge on tau and thus hinders electrostatic binding.

2.1 Identifying and amplifying the gene

If the coding sequence of the protein is unknown but the protein is purified, mass spectrometry (or chemical methods such as Edman degradation) can be used to determine the amino acid sequence of the protein, which can then be used to search for the the corresponding DNA sequence.

Usually, instead of using the full DNA sequence, cDNA is used in cloning because it does not contain introns. cDNA is made by:

1. extracting total RNA from cells by TRIzol and then isolate the mature mRNA by affinity chromatography (with poly-T coated resins which binds to poly-A tails of mature mRNA)
2. reverse transcription of the template DNA strand by viral reverse transcriptase (RT) followed by RNA degradation (often by the RNAse H activity of RT)
3. synthesis of the coding strand by DNA polymerase

Often, the latter two steps are done repeatedly in RT-PCR to amplify the DNA fragment, if the sequence of the GOI is known and specific¹ primers are designed. The primers include additional restriction enzyme cleavage sites for the convinience of traditional ligation-dependent cloning, or other sequences for ligation-independent cloning (see Section 2.3).

If the protein is already known to have a match in malarial cDNA library, we can skip most of the above steps and PCR-amplify the cDNA of interest from the library directly.

When we know the sequence of GOI, we can do a BLAST to find possible homologous proteins, and study any relavant scientific literature. The structure of some homologous proteins might have been solved previously and we can refer to their expression and purification strategies, which might prevent some waste of time on trial-and-errors. Bioinformatic analyses, such as disorder prediction using the RONN algorithm and generation of 3D homology models using the Phyre2 server, might also be helpful.

Codon optimisation might also be needed. Codon usage can vary significantly between species as well as between genome types such as nuclear DNA and mitochondrial DNA. For example, the frequencies of codons AGG, AGA, and CGA (which code for arginine) in H. sapiens are 11.4, 11.5 and 6.3, respectively, while the corresponding frequencies in E. coli are 1.2, 2.1 and 3.6. Codons that occur in high frequency in Plasmodium but in low frequencies in E. coli would result in a condition where the pool of tRNA for that codon will be so low as to become depleted. When the rare tRNAs are depleted to produce the recombinant protein, proliferation of the host cells is restricted, leading to low yield. This problem may be partially solved by the use of an inducible expression vector (e.g. IPTG), but when the foreign gene contains a large number of rare codons, this is not enough. There are two further approaches to improve the yield. First, the rare codons in the foreign genes can be substituted with prevalent codons, but this is not always reliable. Second, host cells can be transformed with the genes that code for the rare tRNAs. This approach is reliable and several such cell lines are commercially available. Custom host cells can also be made using plasmids.

We might fail in crystallising the single full-length protein (often because of low solubility). In that case, we can try co-expressing another protein or add an appropriate cofactor and crystallising the resulting complex. If this still fails, we might consider truncating the GOI, e.g. removing a very hydrophobic region that results in low solubility. In high-throughput approaches, such as the pipeline adopted by Oxford Protein Production Facility (OPPF), many variants are constructed in parallel to maximise output.

2.2 Choosing the host organism and the vector

2.2.1 Bacteria and plasmid vectors

Typically and traditionally, the gene of interest (GOI) is incorporated into a plasmid vector via a ligation-dependent mechanism, then the plasmid is introduced into bacterial (E. coli) cells for expression. There are also ligation-independent methods for making vectors, which is more robust and easy to use (only need to design one ‘adaptor’ parts of the primers once), making it suitable for high-throughput methods.

Bacteria-based protein expression is time- and cost-efficient, but it has two major limitations:

1. the insertion size of a plasmid is small, typically 2-10 kb (bacterial artificial chromosomes (BAC) with greater insertion sizes can be used, but the procedures are more complex)
2. Bacteria might be unable provide the appropriate protein folding environment (including chaperones) and/or post-translational modifications that can be crucial to protein’s functions.

Given the protein of interest is cysteine-rich, it is susceptible that disulfide bridges are present if this protein is also extracellular. The normal intracellular reducing environment disfavours disulfide bridge formation, but Origami and related strains have an oxidising intracellular environment due to mutant thioredoxin reductase and glutathione reductase, and they are used to express proteins that require disulfide bonds to achieve their correctly-folded conformation.

The plasmid should contain selectable markers for easy identification of successful transformants. The combination of Amp^R and lacZ\(\alpha\) is a common choice, as explained in Figure 2.1.

In addition, the promoter should be strong and tightly regulated with minimum or complete lack of basal level transcription under uninduced conditions. The IPTG inducible T7 promoter system (as in pET plasmid) is one of the common choices. IPTG is an analog of allolactose, which can bind to and silence lac inhibitor, de-inhibiting the lac operator. IPTG is not digested by the bacteria, and thus its constant concentration continuously induces protein expression. The lac operator may be also made symmetric, strictly preventing transcription when uninduced.

Some commerially available plamids also include tags added to the N- or C-terminus of the insertion site (multiple cloning site), which aids protein purification. This can be a polyhistidine tag, a fusion protein (e.g. GST) or a short recognisable peptide, and often a cleavage sequence is also introduced for removal of the tags.

Figure 2.1: The pETBlue-2 plasmid vector

Plasmid-based bacterial transformation is usually permanent because plasmids can replicate themselves and be distributed into daughter cells after cell division.

2.2.3 High-throughput strategy

High-throughput protein production aims to use robots to automate the steps in making expression constructs and screening them.

Take the OPPF pipeline for example, when the GOI is an eukaryotic protein, they test expressions with different vector constructs (with varying amino acid start and end points, fusion partners, and plasmids) in both E. coli and insect cells in parallel (Figure 2.2). The expresison screening is automated and performed in 96-well format.

Figure 2.2: SDS-PAGE gel after NiNTA magnetic bead purification showing expression of constructs for DDR1 in (A) E. coli and (B) insect cells via the baculovirus system. The names of constructs correspond to amino acid start and end points and the vector being used (pOPINE, pOPINF or pOPINJ, which confers a C-terminal His-tag, an N-terminal His-tag, or an N-terminal His-GST tag, respectively)

they also used ligation-independent cloning methods (Section 2.3) to reduce variability and thus simplify the workflow.

2.3 Constructing the vector

The DNA fragment we obtained should be incorporated into a vector before they can be introduced into the host cell. Usually, a recombinant plasmid is made first and amplified in bacteria. Then, if necessary, it can be transferred into other vectors such as viruses and liposomes, depending on the strategies. Here I describe two ways of making recombinant plasmids: traditional restriction enzyme and ligation-based plasmid cloning, and T4 polymerase-based ligation-independent cloning (LIC).

Figure 2.3: Left, traditional cloning based on restriction enzyme and ligase; right, ligase-independent cloning based on T4 polymerase

As shown in Figure 2.3, in the ligation-dependent method, the PCR product contains flanking restriction sites (came from the designed primers) to be recognised by a restriction enzyme (e.g. EcoRI). The plasmid also contain the restriction site for EcoRI (located in the multiple cloning site, see Figure 2.1). Cleavage by EcoRI results in complementary ‘sticky ends’, which facilitate annealing of GOI to the plasmid. The sticky sequences are shart (usually 3-5 bp), thus, to overcome low stability, ligation must be done immediately.

The ligation-based cloning are not proper for high throughput protein production (HTPP) projects due to several disadvantages:

• incomplete DNA digestion and poor ligation yields
• each GOI has to be inspected for any internal restriction sites (for exclusion from the choices of the ligation sequence)
• unwanted amino acids can be introduced to the expressed protein
• the GOI can only be cloned in the vector position where the selected restriction site is present

Ligation-independent cloning (LIC) overcomes many of the problems described above. In a T4 DNA polymerase-dependent approach, the sticky sequences are made by the 3’-to-5’ exonuclease activity of T4 polymerase. These sequences are long, allowing formation of stable recombinant plasmid without in vitro ligase treatment. The nicks in sugar-phosphate backbone are later fixed by host ligase.

The obvious elegance of T4-LIC is that once the sticky sequence is designed and constructed, it can be used for any GOI. As T4 polymerase (exonuclease activity) always proceeds from 3’ ends, any internal sticky sequences will not be disrupting.

2.4 Transformation/Transfection

Transformation is the process of delivering GOI-containing foreign DNA into host cells. For eukaryotic cells, this is more often known as transfection (because ‘transformtion’ has other meanings). Many chemical/physical transformation/transfection methods are generally done by making transient holes on host cell plasma membrane (technically, ‘inducing competence’ in host cells). Here are two examples.

Heat-shock transformation is used for small vectors such as plasmids. The general procedures are:

1. Host cells are incubated in a solution containing divalent cations (typically CaCl₂) on ice.
2. CaCl₂ partially disrupts the cell membrane, which allows the recombinant DNA enter the host cell. Such cells are called competent cells.
3. Cells are exposed to a heat pulse (heat shock), and the thermal imbalance causes the entry of DNA through disrupted plasma membrane.

Electroporation can be used for larger vectors such as BAC and PAC². The general procedures are:

1. Host cells are placed into a cuvette, together with the vector. The cuvette is connect to electrodes.
2. The cuvette containing the mixture is subjected to intense electric pulses (2500 V/cm for bacteria, lower for animal and plant cells) each lasting for only a few milliseconds
3. Most cells would die under such treatment, but for those survived, their membranes are polarised by the electric field and are disrupted, and DNA enters through the pores. Finally, the membrane reseals after the treatment.

Depending on the identity of the host cell and the vector, other methods are also available. These are not directly relevant with our goal of expressing a malarial protein, so here is a simple listing:

• cosmid and fosmid vectors are introduced into bacterial cells via bacteriophages
• calcium phosphate co-precipitation for mammalian cells
• microinjection for animal cells
• microprojectile bombardment for plant cells
• adenovirus and lentivirus vectors for mammalian cells
• Agrobacterium-mediated transformation for plant cells
• genome editing techniques based on ‘designer nucleases’ such as ZFN, TALEN and CRISPR-Cas9 with homology directed repair

3.1 Extraction

3.2 Purification

Classical methods for separating proteins depend on variable protein properties, including solubility, size, charge, and binding affinity. In most cases, protein mixtures are sequentially subjected to different separation methods, each based on a different property. After each step of purification, the fractions are assayed (see Section 3.2.3).

3.3 Preparation for crystallography

For crystallisation, the protein sample must be homogenous and correctly folded. They can be checked by the methods descibed in Section 3.2.3.

If the affinity tag needs to be removed (which is often required for structual studies), a protease cleavage site is often incorporated before or after the fusion tag and the cleavage can be conducted either in solution following purification (the protease themselves are tagged) or immediately after enzyme capture in situ on the chromatography resin itself. AKTA Express (GE Healthcare) is an elegant procedure for on-column cleavage coupled to multidimensional chromatography that is highly amenable to high throughput protein purification.

There are three methods to grow protein crystalls, namely ‘hanging drop’, ‘sitting drop’, and microdialysis, but the underlying principles are similar. The protein is first dissolved in a ‘crystallisation cocktail’ droplet and concentrated to 2-50 mg/ml, then it is allowed to equilibrate with a more concentrated reservoir solution of the cocktail with a volume ratio of 1:1. As the protein droplet becomes supersaturated, it may start to crystalise. Robots are commonly used for automatic screeening and optimisation of crystallisation conditions. If crystallisation fails under all conditions, we can try co-crystalising with a ligand. If this still fails, we might consider using a truncated protein, as described in Section 2.1.

Scattering

• When light hits matter…
  • Vibration\(\rightarrow\)scattering (all directions)
  • Energy level transitions\(\rightarrow\)absorption and emission (fluorescence)
  • Photochemical reactions (e.g. photosynthesis)
• scattering can give rise to refraction and diffraction
• experiments on scattering
  • turbidity (reduction in intensity)
  • angular dependence
  • changes in \(\lambda\)

Wnt3-Fz8 Complex (Hirai et al. 2019)

• crystals of lysine-methylated and deglycosylated human Wnt3 (hWnt3)-mFz8 CRD complex were obtained by X-ray crystal structure was solved by molecular replacement and refined to a resolution of 2.8Å.

Two Proton Translocation Models: Loop and Pump

Mechanisms of coupling redox reactions to proton translocation can be classified into two models: the chemical loop, and the conformational pump.

In the loop model shown in Figure 1, the hydrogens (protons plus electrons) from the substrate (AH₂) first reduce an mediator molecule (B) on the N side, and this mediator is reoxidised, forcing protons to be released to the P side via one pathway and electrons to flow back to the N side via another pathway. The electrons then reduce another species (C) using protons from the N side. This mechanism is employed in the respiratory chain, where ubiquinone serves as the ‘mediator’ molecule and ETC components from Rieske Fe-S protein in complex III to Cu_B centre in complex IV as the electron pathway. The number of protons translocated in this way has to follow a fixed stoichiometry (i.e. 1 proton per electron). (Not shown in this diagram, the reduction of Q near the N face driven by QH₂ oxidation near the P face in the Q cycle is also an example of this mechanism.)

Figure 1: The loop model. Left, generic form; right, simplified example found in the respiratory chain.

The key idea of the pump mechanism is the conformational change driven by redox reactions. Conformational change can lead to alternating proton accessibility from the two faces and differential pKa values of proton binding sites, resulting in proton translocation (explained in Figure 2). The number of protons pumped in this way does not have to follow a fixed stoichiometry within the thermodynamic constraints (i.e. E_r/n must be greater than ∆p). This mechanism is thought to be employed in complex I and IV of the respiratory chain. However, recent studies have shown that complex IV may adopt a unique mechanism, as explained later in this essay.

Figure 2: The conformational pump mechanism. When the “pump” protein is reduced, the proton binding site is exposed to the N face and with high pKa, promoting proton binding; upon oxidation, the proton binding site is open to the P face, and pKa drops, promoting proton dissociation.

Complex I

As shown in Figure 3, bacterial complex I (from Thermus thermophilus) comprises two domains. The hydrophobic, transmembrane domain (a.k.a. membrane arm) has a proton-translocating P-module and subunit NuoH. The hydrophilic domain (a.k.a. peripheral domain) has an NADH-oxidising dehydrogenase module (N-module) which feeds electrons to the electron-transferring chain of Fe-S clusters, and the connecting Q-module, which reduces ubiquinone. Mammalian mitochondrial complex I is similar in gross organisation and function, although it possesses more subunits.

Figure 3: Annotated structure of bacterial complex I. PDB accession code: 3M9S

Hydrophobic domain (membrane arm)

The hydrophobic domain comprises subunits NuoAHJKLMN, of which NuoLMN are homologous to bacterial Mrp Na⁺/H⁺ antiporters (characterised by the conserved 14 transmembrane helices).

At the centre of each antiporter-like subunit (NuoLMN), two transmembrane helices are ‘broken’ and each have a kink known as a ‘π-bulge’, and there is a lysine residue in this region (Glu407 for NuoM) critical to proton pumping, confirmed by mutagenesis studies. There are two symmetry-related plausible proton channels lined by hydrophilic/charged residues, one connecting the central lysine on one broken TM to the P face and the other connecting lysine of the other TM to the N face.

Figure 4: Plausible proton paths

As shown in Figure 5, NuoL/M/N may work by using conformational changes (induced from the hydrophilic domain) to transfer a proton from the lysines on the helix with N half channel to the other lysine (or Glu in NuoM) on the helix with P half channel, and if the pK_as of the two lysines differ, this leads to unidirectional proton transport. This mechanism is in accordance with the general scheme shown in Figure 2.

Figure 5: Schematic diagram of proton translocation in complex I

The three antiporter-related subunits are thought to be responsible for the translocation of 3 protons per 2 \(e^-\). An additional channel is found at the interface of subunits NuoN,K,J,A (according to Efremov and Sazanov (2011) and Sazanov (2015) or NuoH,J,K according to Hirst (2013) and Nicholls and Ferguson (2013)) and they are thought to act together to pump another proton with a similar mechanism involving protonation/deprotonation of lysine.

Coupling

It is debated how the redox reactions in the hydrophilic domain triggers conformational change, and how this conformational change is transmitted along the hydrophobic domain.

Some features of the Fe-S centre N2 (the last Fe-S centre in complex I) led to hypotheses in which N2 plays a major role in pumping protons^1,2, but currently, the mainstream idea is that the initial conformational change in the hydrophilic domain is contributed mostly by the reduction of ubiquinone.

As claimed by Wirth et al. (2016), reduction of ubiquinone is a two-step process, where the Q binding site cycles between a conformation that permits electron transfer (E-state), and a conformation that permits proton transfer (P-state) onto intermediates of ubiquinone. The conformational change associated with this controlled process is assumed to initiate proton translocation.

It was originally suggested that the horizontal HL helix (see Figure 3) in NuoL could move horizontally like a piston to transmit conformational change along the hydrophobic domain³, but now, after realising the similarity between NuoH and NuoL/M/N, it is thought that conformational changes associated with π-bulges in the mid-membrane parts are transmitted laterally to NuoN/M/L (Nicholls and Ferguson 2013), and, as suggested by Wirth et al. (2016), this is achieved by electrostatic coupling in the central hydrophilic axis.

As I mentioned last week (Shi 2019), as there is no complete evidence demonstrating how complex I proton pump works, other hypotheses exists, including a radical one recently proposed by Morelli et al. (2019), which emphasises the elusiveness of the ‘proton entrance half channel’ (from the middle Lys to the N face) contrasted by the clarity of the ‘proton exit half channel’ (to the P face). Together with another piece of evidence that an obvious proton tunnelling is found at the centre of complex I, Morelli drafted an lateral mechanism of proton circuiting, as shown in Figure 6

Figure 6: A possible lateral proton circuit suggested by Morelli (2019). Protons are only pumped from the middle to the P side, then trasferred through the hydrophilic heads of phospholipids to ATP synthase. After a complete turn in ATP synthase, protons move back through channelling in the middle of the membrane to complex I, but not to the N side

Complex III (cyt bc₁ complex)

Complex III is a homodimer, with each monomer consisting of three subunits: cytochrome b, the Rieske iron-sulfur protein, and cytochrome c1. Each monomer has two binding sites for ubiquinone, called Q^N and Q^P, indicating their proximities to two opposite faces.

Uniquinone oxidation and proton translocation is coupled in the Q cycle, as illustrated in Figure 7.

Figure 7: The Q cycle. Left, stage 1; right, stage 2.

In stage 1, QH₂ binds to Q_N of one dimer and a Q binds to the Q_P of another dimer. The QH₂ at Q_N is oxidised by two electron acceptors in two steps: the first electron is accepted by the Rieske iron-sulfur protein and passed though cytochrome c₁ and finally to cytochrome c, and second other electron is accepted by cytochrome b, passed through its two hemes (b_H and b_L) and finally to the Q at Q_P on the other monomer, forming semiquinone. The two protons of QH₂ are released into the P face (the molecular details is unclear).

In stage 2, the same process is repeated—one QH₂ is oxidised to Q, its protons released and electrons passed 1) onto one cytochrome c and 2) to the other side—but this time the electron acceptor on the other side is semiquinone. Acceptance of an electron by semiquinone, and the addition of two protons from the matrix, produces QH2.

Complex IV

Complex IV (cytochrome oxidase) accepts electrons from cytochrome c to reduce oxygen to water and pumps 2 electrons to the P face.

Previous X-ray studies suggested that complex IV exists as dimers with each monomer comprising 13 subunits (Tsukihara et al. 1996). However, it was recently shown by Zong et al. (2018) that complex IV is actually a 14-subunit monomer, and dimerisation was an artifect due to the dissociation of NDUFA4 subunit during the detergent-based purification steps before crystallisation (NDUFA4 hampers dimerisation).

Reduction of oxygen

The key catalytic activities are found in subunit I and II, as shown in Fig. 8

Figure 8: Schematic diagram of complex IV with emphasis on the electron-transport and catalytic activities of subunit I and II.

Each cytochrome c first donates one electron to Cu_A located in the globular domain of subunit II. Cu_A is a binulear centre, but it undergoes one-electron redox reactions. The electron is then passed onto haem a, and then to haem a₃-Cu_B binuclear centre, where an oxygen molecule is bound. 4 electrons from cyt c and 4 protons from the N phase are required to reduce one oxygen molecule to two water molecules. Cu_B has three His ligands, and the forth empty coordination site mediates oxygen reduction by holding the intermediates (its malfunction leads to generation of ROS).

Pump mechanism

For each pair of electrons, 2 H⁺ from N phase is consumed to make water and another 2 H⁺ is pumped intot he P phase.

According to a recent review by Wikström and Sharma (2018), the favoured charge-conpensation mechanism is as follows: (see Figure 9)

1. when an electron is transferred from haem a to the haem a₃-Cu_B binulear centre (BNC), one proton from Glu242 is loaded onto the ‘proton loading site’ (PLS⁴). This is a purely electrostatic event (not acid-base, as pK_a of PLS is not high enough).
2. reprotonated Glu242 transfers another proton to BNC, annihilating the -ve charge in BNC, so H⁺ on PLS can leave. There must be a barrier preventing H⁺ flowing back to the N side, which can be achieved by raising the energy of the transition state of proton transfer between the Glu242 and PLS.

Figure 9: Charge-compensation mechanism of proton pumping in complex IV.

Components of pmf in chloroplasts and mitochondria

In mitochondria, the major component of protomotive force (\(\Delta p\)) is the membrane potential (\(\Delta \Psi\)), while in chloroplasts it is \(\Delta\)pH. This is caused by the physiology of these organelles and the electrical properties of membranes.

In chloroplasts, when protons are being pumped into the thylakoid space, the developed membrane potential is balanced out by efflux of K⁺ and influx of Cl^– through voltage-gated channels, such as KEA3 K⁺/H⁺ antiporter and voltage-gated VCCN1 Cl^– channel, so it results in a pure pH difference. TPK3 was originally thought to be one of such channels but was recently shown to be not critical (Höhner et al. 2019). There are not many enzymes in the thylakoid lumen, so a low pH is not unsafe. In addition, the high H⁺ concentration has regulatroy functions. (Höhner et al. 2019)

If mitochondria were allowed to develop such a pH difference (very basic matrix), this would denature the numerous enzymes, such as those involved in \(\beta\)-oxidation and TCA cycle, that operate inside the matrix. In mitochondria, there are not such ion channels to counteract change in \(\Delta \Psi\), so proton translocation results in charge imbalance, leading to \(\Delta \Psi\) change⁵, and a significant membrane potential can be established with only minuscule amount of charge movement.This can be explained by the low capacitance of the membrane. Capacitance is the amount of charge separation needed to develop unit voltage difference, i.e.

\[C=\dfrac{Q}{V}\]

Given a constant \(V\) (\(\Delta \Psi\)), lower capacitance means smaller charge difference is required to build up that voltage difference.

Figure 10 shows a simple parallel-plate capacitor, which is analogous to mitochondrial inner membrane.

Figure 10: Illustration of a simple parallel-plate capacitor

The electrical capacitance of such a plate can be calculated as:

\[C = \dfrac{\epsilon A}{d}\]

where \(\epsilon\) is the permittivity of the plate, \(A\) is the surface area and \(d\) is the distance separating charges. For mitochondrial inner membrane, although \(\dfrac{A}{d}\) is large, \(\epsilon\) is very small, so the overall capacitance is small.

Clues leading to the original chemiosmosis hypothesis

The mechanism of oxidative phosphorylation were initially thought to be analogous to substrate-level phosphotylation, that the redox reactions create ‘high-energy’ intermediates, which in turn adds P_i to ADP. In 1961, Peter Mitchell listed 6 facts to question this chemical mechanism hypothesis:

1. The ‘high-energy’ intermediate is not found.
2. The close association between phosphorylation and membranous structrures is not explained.
3. P/O ratio varies–it does not follow a fixed stoichiometry.
4. ATP hydrolysis outside mitochondria promotes NAD reduction inside, accentuated by succinate oxidation. (this observation is probably irrelevant or wrong?)
5. Uncoupling can be caused by reagents with distinct chemical properties–so they are unlikely to act on a single ‘intermediate’.
6. Unexplained swelling and shrinkage accompany phosphorylation.

Trying to explain these phenomena, he proposed his initial chemiosmosis mechanism, with following 3 features:

1. A membrane located, reversible and anisotropic ‘ATPase’
2. An electron and hydrogen translocation system
3. A charge impermeable membrane

According to this model, oxidation of substrates does not occur in a single compartment, but it instead allow protons and electrons to go in different directions, which in turn creates a concentration gradient of H⁺ and OH^-, as well as a membrane potential. ATP synthesis is driven by the one-way movement of OH^- down its electrochemical gradient.

Mitchell’s origin model was rather different from the current model, as depicted in figure 1, but it provided decent explanations for the phenomena listed above. (In his later review (Mitchell 1966), there were some amendments made on his original model, but I have no time to read it)

Figure 1: Left, Mitchell’s original chemiosmosis model; right, the current model

The principal idea was that, energy released from ‘asymmetric’ redox reactions can be used to establish a rather ‘physical’ form of energy (not the common chemical bond energy), the electrochemical gradient, and this is the first coupling. The second coupling is between the synthesis of ATP and the vectorised diffusion of some species promoted by the electrochemical gradient established by the first coupling.

It should be noted that the conception of first coupling was not new. Before the 1960s, it was known that redox reactions can create a electrical potential between two compartments, and it had been noted that some inhibitors of ion transport also has an effect (either stimulatory or inhibitory) on respiratory rate (Robertson 1960) (Fig. 2).

Figure 2: Correlation between the effect on oxygen uptake and on ion accumulation of some inhibitors

Proving proton pumping by the redox chain and determining H⁺/O ratio

The ‘first coupling’ of the current chemiosmosis model means the pumping of protons into the P compartment powered by stepwise redox reactions along the ETC, with the overall oxidation of NADH or FADH₂ as substrate by molecular oxygen.

In actively respiring mitochondria, the proton concentrations (i.e. pH) on either side of the membrane is constant because the rate at which protons are pumped into the P compartment and that flowing back to the N compartment are equal. This makes it difficult to detect the presence of proton flux.

Mitchell and Moyle (1967) devised the following procedure to solve the problem (Fig. 3):

1. mitochondria are incubated in anaerobic conditions, allowing H⁺ to equilibrate (i.e. making \(\Delta p\) fall back to zero)
2. a pH-sensitive glass electrode is used to monitor pH of the suspension (i.e. the environment outside the mitochondria)
3. a known quantity of O₂ is introduced into the suspension by ‘injecting pulses of air-saturated 150mM KCI solution.’
4. by measuring the pH change, the amount of protons pumped out of the mitochondria can be calculated, and with the known amount of O₂, the H⁺/O ratio can be calculated.

Figure 3: Determining H/O ratio (Mitchell 1967a)

Additional procedures are needed to minimise error, and they’re described in Bioenergetics pp59-60

Proving the ATP synthesis is driven by the electrochemical gradient

The electrochemical gradients is the sum of concentration gradient and electrical potential gradient, and each of them are shown to drive ATP synthesis.

Jagendorf and Uribe (1966) showed ATP synthesis can be induced in dark, by artificially creating a pH difference across the thylakoid membrane in chloroplasts (Fig 4). In this experiment, broken chloroplasts (extracted from spinach) are first incubated in an acidic solution, allowing the thylakoid space to achieve the enviornmental pH of 4.0 (different acids are used to verify its non-specificity). Next, chloroplasts are exposed to pH 8.0 to develop the proton gradient across the thylakoid membrane. While moving out from the thylakoid space, protons drive ATP synthase. The amount of ATP produced is then determined by firefly luciferase assay. By repeating with different pHs, the authors concluded that the rate of ATP synthesis is positively correlated with the pH difference between two phases.

Sone et al. (1977) used reconsitituted vesicles, made from purified \(\text{F}_o\text{F}_1\) complex and the membrane of a thermophilic bacterium, to achieve ATP synthesis by induced membrane potential (Fig 4). They first incubated mitochondria with valinomycin at a fixed pH, then exposed the suspension to KCl. Valinomycin-mediated entry of K⁺ shifts the membrane potential so the vesicle interior becomes more positively charged. Consequently, H⁺ moves out and drives ATP synthase.

Figure 4: Left, Jagendorf (2016); right, Sone (1977)

Monitoring the protomotive force and proton flux in different conditions

If the chemiosmotic model is corrent, it will have some properties resembling those of an electric circuit–the current (\(I\)) and electromotive force (\(V\)) are mirrored by the proton flux (\(J_{\text{H}^+}\)) and the protomotive force (\(\Delta\text{p}\)), respectively. These two terms can be monitored in different states of the proton circuit, and the results are in accordance with the chemiosmotic model.

Figure 5: Circuit Analogy. Top left, proton gradient; top right, oxygen concentration, whose gradient reflects proton flux; below, illustrations of each numbered state.

The different states shown in Fig 5 are explained below:

1. When ETC substrate is absent, there is absolutely no oxygen consumption. No active forces act protons so they are allowed to equilibrate, and therefore the \(\Delta\text{p}\) is zero. This resembles an electrical circuit without a power source.
2. When substrate (e.g. succinate) is provided, it is oxidised by ETC and oxygen is consumed. ETC starts to pump H⁺ into the P face, building up the electrochemical gradient If the membrane was perfectly impermeable to H⁺, ETC would quickly stop as its provided energy for proton efflux balances the electrochemical gradient which favours protons’ influx, and there would be no oxygen consumption. However, the membrane is ‘leaky’ (and there are other processes causing influx of H⁺), causing constant influx of H⁺, which in turn allows H⁺ to be steadily, but slowly, pumped to the P face, which is accompanied by oxygen usage. This resembles more of a closed circuit where energy is dissipated slowly through a high-resistance resistor than of an ‘open circuit’ as described in some books.
3. When ADP is added, ATP synthase can use the proton gradient to synthesise ATP. The proton flux through ATP synthase is huge, so the oxygen concentration (whose gradient parallels proton flux) decreases steeply. (I can’t give an rigorous mathematical explanation on the decrease of \(\Delta\text{p}\))
4. ATP is used up, and the condition is identical to state 2.
5. FCCP short-circuits the proton circuit by allowing H⁺ quickly flowing back to the N face without doing ‘useful work’ (driving ATP synthase). The proton influx is so large that it outstrips the maximal efflux powered by ETC, so \(\Delta\text{p}\) quickly falls back to zero.

Lateral proton current?

The widely accepted view that ETC pumps protons transversally from the N to the P face may need updating for the following reasons stated by Morelli et al. (2019):

1. H⁺ almost never exist as isolated protons. Instead, they bind to water to form H₃O⁺. Free protons in a mitochondrial periplasmic space are too few (fewer than 10) to drive ATP synthase.
2. Free protons are destructive for biological membranes
3. The pH inside mitochondria was shown to be higher by 0.5 units from what was previously believed (Żurawik et al. 2016)
4. Phospholipid membranes are intrinsically (significantly) permeable to protons

According to the classic view of proton pumping, complex I should have a continuous channel traversing the membrane. However, based on comprehensive X-ray studies, the ‘proton entrance half channel’ is not identified with certainty, while the ‘proton exit half channel’ is clearly identifiable, and an obvious proton tunnelling is found at the centre of complex I. This leads to an lateral mechanism of proton circuiting, as shown below:

Figure 6: A possible lateral proton circuit suggested by Morelli (2019).