Module 5- Interactions/ Analysis/ Modifications Flashcards
Protein abundance in a yeast cell
~42 million proteins per yeast cell
Abundance varies from 3/5 copies to 7.5E5 copies per cell
Median abundance of 2622
Protein abundance in a human cell roughly
Cell is ~1/2 protein
~300mg/mL of protein in a cell
Human has ~20,000 protein encoding genes
Generation of protein diversity in humans
Comes from isoforms and PTMs
1 gene can lead to a range of different proteins which are different due to genetic (splicing creating isoforms) and PTM
Different protein interactions leading to biological functions
Genetic pathways- signalling pathways/ sequential interactions
Pathway scaffolding
Enzymatic reactions
Molecular machines- form stable complexes, not much changes
Things important in protein-protein interactions
Domains are cornerstone
Intrinsically disordered regions important
Features of intrinsically disordered regions
Not well defined structures
Highly modified and therefore have a lot of diversity- more tolerant to PTMs, insertions and deletions
Domains bind to disordered regions
What is a yeast-two-hybrid assay
Where a DNA-binding domain and activation domain which normally bind together are separated and have tags put on them (bait and prey)
Can use two methods to validate this: either causes transcription of essential gene and get growth or leads to transcription of reporter gene such as causing fluorescence
Pair-wise interactions
Affinity purification mass spectrometry
Extract proteins and complexes from cells, use antibodies to do mass spec on these complexes and get data on them. Doesn’t just give pair-wise interactions
First example of experiment done- what does it mean that they had the largest data set compared to literature curated data but there was still heaps missing?
Luck et al. 2020
They compared their dataset obtained from their experiment with data curated from literature and had the biggest and best dataset
Isnt complete because it was a yeast two-hybrid test so therefore, interactions in mammalian cells wont all be in yeast cells such as ones needing PTM, if a complex is required for interaction to occur, or membrane proteins being mis-localised and wont be with their partner
First example of experiment done- what two methods were used to validate interactions found in yeast two-hybrid, making hypothesis to be tested
Protein x onto jak protein, protein y onto stat
If x and y interact then jak P stat and then the P stat causes transcription of the reporter gene
X and y both attached to proteins that when they react cause a fluorescent result
Different types of protein-protein interaction interfaces
Pre-formed interface
Conformational change leading to an adaptation of shape to cause interaction
Folded/ ordered domains binding disordered structure
Disordered structure folding upon binding with domain
Two disordered structures coming together and folding
General properties of interfaces/ how are they characterised
Overall amount of surface area buried
Chemical composition of the buried surfaces (enriched for aromatic resides)
Shape and charge complementarity of occluded surfaces (close packing)
Specific interactions such as hydrogen bonds and decreased flexibility
Features of size of the interface/ buried surface area (BSA)
Difference in surface area of two proteins alone compared to when they are in a complex
Can cause underestimation as it assumes that proteins come together without conformational change which isnt always the case
Average ~18,000 which is overestimate as mainly only large stable proteins have been sized
Less weaker transient complexes have been solved which tend to be smaller
Foglizzo and how interface was found
Size exclusion of protein showed that it was a dimer and each subunit had three functional units. Interface area was small and there was three possible ways it could have come together with this small interface
Did mutations in each possible interface and size exclusion on each, showed that one interface mutation led to a monomer which must be the interface as mutation meant they could not come together as dimer
More experiments could then be done to fond out how it interacts
Composition of interface residues
Ratio of <1= residue less likely to be in interface eg acidic or polar
>1= residue more likely to be in interface eg aromatic and arg
Core and rim of interface
Core of interface is buried and has no contact with solvent, likely to have aromatic and hydrophobic residues, look like protein interior
Rim of interface has parts buried and exposed to solvent, likely to have more polar residues here, looks like rest of non-binding surface
Water molecules in interface interactions
Often make key connections, often sitting around edge of rim near polar residues
May be found in unexpected areas, not likely to be there though
Obligate and non-obligate
Obligate= always in a complex
Non-obligate= regulated interactions, sometimes monomers, sometimes dimers, sometimes in complexes
Shape and charge complementarity- scoring
Different scores used to measure packing of proteins in complex
How close residues are, how many holes there are, how they come together- complementarity
When interfaces are small this is hard to determine
Close packed= 0.7, likely real interface
Crystal packing, not close= 0.3/0.4
Conformational change and complementarity and size
Smaller interface has less conformational change- less entropic cost= more pre-organised and fit together
Larger interface has more conformational change- more entropic cost
Conformation and anchor residues
Small interfaces tend to have anchor residues which come together and help binding interactions
Other residues then move around these anchors to optimise contacts
Three types of ways proteins may come together and interact and which is more likely
Compact and interact together- conformational change
No conformational change, just come together
Extended interaction- residues reach and contact each other- optimise contact= most likely
What is alphafold
Take knowledge of sequence and structure and builds a protein model
Early days, can possibly be applied to protein complexes
Alphafold in yeast-two-hybrid being used to predict complexes example
Humphreys et al. 2021
Many structures were predicted
Not good for transient interactions
Better prediction for large stable complexes
Some binary complexes that were found actually normally fit into a complex but needs more info
Again, reiterates early days
Things predicted to be at the interfaces in human protein interaction network
Predict disease causing mutations are at interface and phosphorylation sites which lead to regulation of protein interactions
What is a pDockQ based on
Proximity- number of residues in close proximity, beta carbons within 10 angstroms, greater number in close proximity, more likely to be a contact
PLDDT- alpha fold prediction score, how good the prediction is
Orthogonal data eg cross-links in protein interaction determination
Cross linking reagents can be used, bifunctional with cross-linker and functional ends which bind to different proteins
Mass spec can be used to identify cross-linked residues
Can sort cross-link data based on pDockQ, scores with higher confidence observed more significant cross-links (spacer less than a certain number of angstroms)
Location of proteins and affinity of interactions
Obligate oligomers- high affinity
Non-obligate permanent PPIs dont need to interact but when they do have high affinity
Non-obligate triggered transient PPIs- high affinity, regulatory
Non-obligate co-localised PPIs- moderate affinity, regulatory
Non-obligate weak transient PPIs- dependent entirely on concentration, low affinity
Ways and approaches to validate data out of high throughput
Qualitative and quantitative
GST pulldown (qualitative)
Isothermal calorimetry (quantitative)
Surface plasmon resonance SPR (quantitative)
Features of protein-protein interaction domains
Independently folded, 35-150 aas, can still bind target if expressed independently
Binding properties of isolated domain reflect those of intact proteins
N- and C-termini close in space with ligand binding site on opposite face
Folding allows domains to be connected without disrupting function
What are SH2 (~115) and SH3 (~300) domains in human genome
Src homology
First discovered cancer causing protein, showed cancer caused by mutations
Regulate kinase activity (SH1), has Tyr527 on end of gene commonly mutated in cancer
Features of proline binding motifs making them play a key role as a docking site for signalling proteins
Unusual shape of pyrrolidine ring
Constrained dihedral angles
Substituted amide nitrogen
Relative stability of cis isomer
How does isothermal calorimetry work
Reaction cell filled with protein solution and injected syringe filled with ligand solution
Small volumes of ligand injected into cell triggering binding reaction
Exothermic- samples becomes warmer and causes downward peak sequence
When binding saturation reached, remaining heat effects (if present) due to mechanical and dilution
Area under peak plotted versus molar ratio, gives Ka
How does surface plasmon resonance (SPR) work
Measures based on kinetics, measures kon and koff rates
One substrate immobilised on a surface, the other run over top
Reflected light tells how much is bound, can measure the association and dissociation
Compare sensorgrams for different interactions
PPI paper 1 and how they validated novel interactions found at hugh confidence
Burke et al. 2023
Cross-linking data between proteins
Disease causing mutations prevalent at interfaces
Phosphorylation sites at interfaces
Recognition of prolines for proline binding motifs
Often stretches of proline (polyproline type II helices) are favourable for binding
3 residues per turn, ring and carboxyl regions regularly positioned- backbone restricted
Carbonyls are free so make no intra-molecular H bonds and are free to make interactions which binding proteins take advantage of for interaction
SH3 domain and proline rich regions
SH3 has two antiparallel B-sheets and two variable loops (RT and n-Src)
Binds ligand in polyproline II helix
Recognition relies on N-substitutionm proline
Variability in loops confers some selectivity
More on how SH3 binds proline-rich regions
One or two residues at end of PPII helix recognised by loops eg Arg- loops required for specificity
Two xP grooves for proline binding
Three aromatic residues make areas for proline to bind (xP grooves)
How to break SH3 binding and how was it found that three aromatic residues are important and loops are important
Lim and Richards 1994
Break by determining if residue is in interface and mutating or by unfolding the protein
Can mutate residues to see if they are in interface and if no binding occurs, shown that mutating 3 aromatics the binding is broken= evidence for being in interface
Mutating regions in RT-loop can make SH3 better or worse at binding (if was bad before, mutations can make better and vice versa)
Features of the WW domain
Triple stranded B-sheet with 2 conserved Trps with conserved length of the loop, where the Trps are tells where binding interface is
Binds ligand in PPII helix and recognition relies on N-substitution of proline
Usually has one xP binding pocket
WW domain binding to proline rich region and compared to SH3
One xP groove where proline sits
Specificity loops recognise a specific residue which confers binding
Both little domains which bind proline rich regions, aromatics in the exact same place- have evolved to have residues in the same space showing that having aromatics making xP grooves is best for binding
What does an SH2 domain do
Docking site for signalling proteins as tyrosine phosphorylation allows recognition to be tightly regulated by kinases
Various evolution has occurred, used in many places and in many proteins
Features of SH2 domains
4 stranded anti-parallel B-sheets, 2 helices conserved and a positively charged pocket with Arg binds pTyr (neg charged) specifically provided by C-terminal- it always binds in the same place
SH2 conserved regions binding to pTyr
BetaB5 Arg and alphaA2 arg/lys
BetaB7 Ser and betaD His
All important for the interaction and shows that SH2 recognise pTyr in the same way, gives specificity for the pTyr
Importance of loop conformation in SH2- how it was found
Kaneko et al. 2011
Took 70 SH2 and evaluated binding at semi-high throughput, screened against peptide libraries containing pTyr residues, led to 2 models of binding
Loops block different pockets (+2 +3 or +4) allowing one pocket to be free for recognition and binding of proline rich protein
Surrounding residues in SH2
Permissive and non-permissive interactions combine to influence binding
Surrounding residues have an impact and influence binding and can cause unfavourable interactions and block binding
Additional sequences in each protein can give specific interactions
SH domains and kinase interactions
The Tyr527 on the end is phosphorylated and goes into SH2 which inactivates it. The SH3-SH2 linker is proline-rich and binds to SH3 which keeps it inactivated and overall, kinase is inactive
For kinase to be active, interactions need to be turned off by a phosphatase removing the P or a competing proline rich sequence binding instead of the linker
How cancer affects kinase of SH domains
Elevated phosphatases shifting the equilibrium allowing the phosphate to be removed and kinase to be turned on
Mutated Tyr so cant be phosphorylated
Suppression of the kinase that adds the phosphate onto Tyr527
Ways to help specific binding for SH2 and SH3 domains
RT loops
Conformation of loops, determines if +2 +3 pr +4 is important for binding (not much specificity as most SH2 belong to +3)
Domains have additional regions for interaction eg flanking or extended specificity surface, multiple modules, multiple recognition surfaces
Approaches to achieve high-fidelity signal transmission
Co-localisation- organise on cell surface
Compartmentalization (organelle)
Scaffold-mediated complex assembly
GRB2 structure and domains
Adapter protein with no enzymatic function
Has 2 SH3 domains and 1 SH2 domain
SH3 domains on outside binding polyproline sequences at the ends with slightly different specificity
SH2 domain in the middl binding phosphotyrosine
SOS and GRB2
SOS most commonly binds to SH3 domains in GRB2 as it has polyproline sequences
SOS is a GTP exchange factor causing MAPK signalling
Weak interactions with GRB2 for controlling signalling
Transduction of TCR in experiment results
Su et al 2016
LAT is embedded in membrane on a microscope slide due to His-tags in the N-terminus= immobilised
LAT is also phosphorylated and when GRB2 and SOS are added they mediate the formation of clusters
Adding phosphatase removes the clusters
High concentration of SH3 binding sites on GRB2 for SOS so if one interaction is knocked off, another can be made and is the bridge for cross-link formation
Clustering in TCR with valency of GRB2 binding sites
When Tyr is mutated to phenylalanine clusters struggle to form
By increasing the concentration of GRB2 and SOS, even with lower [tyrosine] still get clusters
No tyrosine= no clustering
Multivalent binding- need to bring molecules together to get interactions/ results
Clustering in TCR and phosphatases
As clusters form phosphatases are excluded- builds a wall to keep out phosphatases allowing the signal to occur
GRB2 and EGFR (example of RTK) signalling
Kinase receptor tyrosine is phosphorylated, which GRB2 SH2 domain binds to
SOS binds to GRB2 SH3 domains, and nucleotide exchange occurs on RAS which causes signalling and MAPK activation
GRB2 structure- is dimerisation important?
Forms a dimer in crystal structure
Tyr160 is phosphorylated normally, but this is buried in the dimer so cant be phosphorylated- evidence to think that dimerisation is not important or always done
Ways it is shown GRB2 is a dimer in solution- gel
Ahmed et al. 2015
Have the wildtype, one with a mimic of phospho-tyrosine, one with two asparagines mutated to glutamic acid for force interactions apart
When run on a gel, evidence shows they are dimers
Ways it is shown GRB2 is a dimer in solution- fluorescence
Ahmed et al. 2015
FRET used to measure interactions
Co-expression with mutant shows no decrease in fluorescence lifetime
Wildtypes show decrease in lifetime= have dimers
Monomer:dimer equilibrium in solution
Monomeric GRB2 in tumours
Ahmed et al. 2015
Shown with antibodies specific for phospho-tyrosine
When monomer dominates= have high proliferative cells
Phosphorylated Tyr160 more prevalent in high grade tumours= monomer
Clustering of EGFR and GRB2- experiment done
Lin et al. 2022
EGFR sumo molecule attached to his tag= embedded in membrane on microscope slide
With ATP and Kinase, when wt-GRB2 added clusters form
When GRB2 with P-Tyr160 (=monomer) no clusters
When mutated GRB2 mimicking P-Tyr160 with glutamic acid, no clusters
Shows GRB2 dimers are required for cross linking ang cluster formation
GRB2 and EGFR clusters and RAS
Linked to RAS activation: when there are clusters there is more RAS activation
SOS and RAS activation with EGFR
SOS is not needed for cluster formation and can get clusters without it
However, SOS presence causes better cross-linking, enhancing clustering and RAS activation
Clusters recruit SOS and activate RAS signalling
Characteristics of protein interfaces
Generally flat, large, highly variable, featureless and flexible
Involve many intermolecular contacts
Drugs and protein binding
Small (<500 kDa) and bind to pockets= hydrophobic
Current FDA approved protein interaction drugs
Cell surface receptors and enzymes represent ~80%
Adv of enzymes is can do assays and have binding pockets
Important drug targets
Nuclear receptors, ion channels and transporters
How many human proteins thought of as ‘druggable’
~10-15%
Features of GH interaction with GHR
Binds in asymmetric way
1:1 interaction
3nm= tight
Buried SA of 1300 angstroms with 33 residues
DeltaG= -12.3kCa/mol
Mutating 33 residues in GHR binding pocket- two important conclusions made from this
Clackson and Wells 1995
All residues mutated to Ala except 2 cys residues, measured binding
Importance of residue for binding doesnt correlate to BSA (extent to which residue correlates with interface doesnt correspond to binding)
A couple of residues are important for the binding and interaction
Hot spots in GH:GHR interaction
When mutated residues mapped onto binding SA, it is seen that certain areas or ‘hot spots’ are important for binding affinity- sit close together
Tend to not have water
Hot spots in both GH and GHR interact with each other containing key residues for the interaction
GH:GHR hydrophobic residues and importance
Hydrophobic residues are important but important residues dont correlate with hydrophobicity
Abundance of SOS compared to GRB2
SOS is in much lower abundance than GRB, might not act in a 1:1 stoichiometric ratio
Overall: multivalent binding interactions, triggers and recruitment of proteins
Multivalency critical to the way interactions form
Triggers shift equilibrium- causes molecules to bind and increases interactions while allowing repulsion of other interactions
Favourable and unfavourable interactions cause phase separation
Clusters can allow the recruitment of other client proteins eg SOS to further enhance interactions
Clustering causes high concentrations of molecules, meaning weak interactions become more powerful
Intrinsic cell death pathway
Regulated by BCL-2 protein abundance
Many inhibitors of inhibitors
In healthy cells, high BCL-2, inhibits BAX, BH3 only is low and free or absent, mitochondria intact
Apoptotic cell, BH3 only increases and binds to BCL-2 leaving BAX free to put pores in mitochondria
BCL-2 elevated in cancer= promoting
Structure and homology of BCL-2 (B-cell lymphoma) proteins
BCL-2 pro-survival and pro-apoptotic all share homology domains (BH domains)
Survival include Bcl-2, Bcl-xL, A1, Mcl-1 and Bcl-w
Apoptotic bax-like include Bax, Bak and Bok
Apoptotic BH3-only include Bim, Puma, Bmf, Bid, Noxa, Bik, Bad and Blk
Bcl-2 protein interactions
For survival, Bax-like bind into Bcl-2 and are inhibited
For apoptosis, BH3-only bind in the same place as Bax-like in Bcl-2
Features of BH3 motif
4 hydrophobic residues (h1-h4) all on one side
Conserved aspartic acid and leucine
Gly/Ala conservation in other position
Binding of Bax to Bcl-2 (pro-surivival complex)
Bax is disordered and folds as it binds
Four hydrophobic pockets in Bcl-2 BH3 create areas for hydrophobic residues to interact/ go to for interaction
BH3 and selective binding
Chen et al. 2005
BIM and PUMA both bind to every Bcl-2 protein
Others dont bind to all Bcl-2
All BH3-only have core elements but differ in sequence which confers selective binding
Interactions were different
BH3 selective binding and drug development
Selective binding by BH3 id important for drug development due to underlying key selectivity
Could get selective elements that only bind to key Bcl-2 elements= selective binding and selective inhibition
Bcl-2 flexibility
Bcl-2 proteins undergo slight conformational change and arrange differently to accomodate different peptides
Flexible grooves, some differences in core elements
End state differs depending on the ligand= plasticity
Killing correlation with Bcl selectivity
Different cells have different profiles of pro-survival molecules and rely on different pro-survival molecules
Selective binding correlates to selective killing
Development of ABT-737- screening for binders
Petros et al. 2006
NMR used as could use high concentrations to find weak binders and can look for on target binding- ones that bind in the grooves
Firstly screened 10,000 compounds with Bcl-xL, added 10 at a time, 210Da average mass of fragments
If got shifts in NMR then deconvoluted (one at a time) to find the binder
Second screen where they took compound 1 from first screen and screened another 3500 with 125Da average mass and see if better binding occurs
Development of ABT-737- compound optimisation
Took compounds from first and second NMR screens and added them together which gave a compound with good binding (took two weak binders and made them strong binders)
Made many libraries of this and looked for even tighter binders- got higher gain in affinity
Development of ABT-737- issue with albumin and overcoming
Oltersdorf 2005
Compound also bound to albimun so was mopped up and couldnt do its thing
Overlayed structure from optimisation with where it is known to bind in structural visualisation and overlayed albumin with its site, looked for sites to target so that they would be solvent exposed when bound to Bcl-xL but were required in albumin binding
Anti-design to overcome binding to albumin, led to ABT-737
Binding of ABT-737 in BH3 binding groove
Lessene et al. 2008
Bound to Bcl-xL
Selects for a deeper binding pocket in h2 and h4- makes Bcl-xL adopt a conformation allowing for deeper binding- takes advantage of plasticity
ABT-737 to Abt-263 and clinical trial outcomes
Navitoclax
Caused tumour lysis syndrome- destroyed tumours too quick, body could not deal with all of the debris and led to death
Showed it was effective and that the therapeutic window needed to be tweaked
Also caused rapid and constant loss of platelets
Why was Abt-263 causing a loss of platelets
The interaction being targeted was Bad-Bcl-xL
However, Bad is not completely specific and binds to Bcl-xL, Bcl-2 and Bcl-w
Cells have different pro-survival molecules that they rely on, platelets only have Bcl-xL to rely on so it led to a decrease in tumours and also a decrease in platelets as their survival molecules were also lost
Development of Abt-199 (venetoclax)
Souers et al. 2013
Goal was to make Bcl-2 selective
Altered Abt-263, removed a group leading to a molecule with increased selectivity but decreased affinity
Looked at crystal packing of new molecule, saw there was a Trp from somewhere else that fit in well, so added indol ring which went into the same place the Trp was in crystal packing
Abt-199 more Bcl-2 selective with Ki of 0.01nm (Ki of Bcl-xL 48nm and of Bcl-w 245nm)
Outcomes of venetoclax
Much less decrease in platelets
Many lymphoid tumours rely on Bcl-2 so led to decrease in lymphoid malignancies
First drug in this class/ first drug to target this new pathway
Why do cancer cells have increased sensitivity to venetoclax
Cancer cells are in an equilibrium of wanting to die due to damage but cant because cancer causes increased pro-survival molecules (Bcl-2)
Adding venetoclax causes disruption of this equilibrium so killing occurs rapidly
Not as much collateral damage to other cells
Good for any tumour over expressing Bcl-2
Venetoclax and resistance
A mutation in the p4 pocket develops in patients who have taken this for a number of years: G101V
BH3-only proteins still bind the same in this p4 pocket
The mutation causes a change in position of E152, which shifts Y18 leading to a small rearrangement of the bottom of the p4 binding pocket leading to a decrease in affinity for venetoclax
S55746 compared to venetoclax
Birkinshaw et al. 2019
Binds to Bcl-2 in a different way- binds in the p1-p3 pockets, not in p4
Therefore, the mutation in p4 of G101V would not affect binding of this molecule as it does for venetoclax
Optimising protein interactions and biochemistry for Abt-263- antibody
By adding an antibody which is specific for tumour cells to Abt-263 in conjugate, could cause specific delivery of the molecule to tumour cells causing death of only those cells as only targets Bcl-xL in tumour cells- under trials
Optimising protein interactions and biochemistry for Abt-263- PROTAC
Khan et al. 2019
Abt-263 is linked to another molecule which binds to an E3 ligase
This would cause the ubiquitination of the target protein (Bcl-xL) for degradation
Advantage is that a low dose is needed to destroy many targets, but with other approaches higher doses are needed as they are single use molecules for inhibition of Bcl-xL
Trials have shown loss of target, no off-target loss eg platelets and a reduction in tumour= promising