PhoreMost Ltd. Phenotypic screening for novel druggable targets using Protein-interference

Transcription

1 PhoreMost Ltd Phenotypic screening for novel druggable targets using Protein-interference

2 Company Snap-shot PhoreMost Ltd: New Biotech with a mission to drug key underlying causes of unmet diseases, the majority of which are intractable to current drug discovery technologies Ashok Venkitaraman MD, PhD, FMedSci Chris Torrance PhD Novel PROTEINi phenotypic screening platform to identify the best disease targets & how to drug them Building a pipeline of validated targets & drug discovery programs for out-licensing key pharma bottleneck Dir. MRC Cancer Unit, Cambridge University CSO of PhoreMost Founder Horizon Discovery CEO of PhoreMost Cancer focus initially; first program = novel drug target site addressing mutant KRAS cancers Also seeking collaborations on the Protein-i platform

3 The druggable genome, so far The human proteome: >100K (?) possible proteins (including splice variants) >300K (?) predicted high-confidence binary interactions Protein-protein interactions represent a vast untapped repertoire of potential targets Potential drug targets Drugged targets belong to just a few structural classes Only a tiny number of proteome targets has been drugged Eg., 1065 unique drugs (of 1357) target 324 known efficacy targets (of which 266 are human proteins) Overington et al. (2006) 68 targeted cancer therapies (July 2014) Kinases Glycosylated phosphoproteins Nuclear receptors Monoxygenases Growth factors Haemopoietic receptors GPCRs Cytokines Nimgaonkar & Venkitaraman (unpubl.)

4 Why is target-id/validation important? (Cancer e.g.,) Advances in genomics leading to deep genetic understanding of disease biology But few key disease drivers are directly druggable Those that are have largely been done, to good effect Defining novel drug sites (direct or indirect) is now a bottleneck & you can t guess! Not druggable Druggable

5 The underlying potential of phenotypic screens 1) Good targets are hidden in complex disease networks: Targets should be screened exhaustively to link them to a disease trait 2) Good drug sites are hidden in subtle and changing protein shapes in living cells: Targets should ideally be screened in with high-diversity probes that mimic biologic surfaces New solutions are required to uncover targets AND probe cryptic drug sites as they exist in nature

6 Complementary drug discovery approaches Target-based screening Phenotypic screening YES e.g., kinases with deep pockets + natural small ligands to mimic Streamlined drug discovery Validation critical Druggability essential Small list of candidates NO e.g., Protein/Protein interactions with large & discontinuous surface areas Swinney & Anthony NRDD 2011 Direct screening for desired outcome Good source of 1 st in class targets Small molecules>>diversity & target-id? RNAi/CRISPR>>>>>druggablilty? How can systematic target identification and validation be more closely linked to druggability?

7 Overview: PROTEINi phenotypic screening platform Genome-wide target-id & validation: GE Libraries e.g., CRISPR: High Target Specificity Off-targets weakly exploitable Drug sites not identified Size ~30,000 RNAi Libraries Low Target Specificity Off-targets confounding Drug sites not identified Size ~30,000 PROTEINi Libraries (diverse 3D protein-folds) High Target Specificity Off-targets co-exploitable Drug sites identified Size ~1,000,000,000 (dwarfs small mol. diversity)

8 PROTEINi : A live-cell system to find Tx-targets & how to drug them RNAi/CRISPR work indirectly to remove proteins. This is NOT what drugs do & druggable sites are NOT identified Does it usefully affect a disease? Does it spare normal cells Which patients will benefit most? PROTEINi directly inhibits target proteins with a high diversity libraries (10 9 ) of protein-fold shapes that systematically probe for cryptic druggable sites Phenotypic Screening (USP#2)

9 Genome-based 3D protein fold libraries Smaller or random in sequence Bigger e.g., Antibodies No 3D shape = No targets Derived from genomic coding sequences Goldilocks position for size (15-80aa), affinity, 3D potential & shape diversity Large footprint & very high affinity = Lots of targets No drugs Viable drugs But non-viable drugs

10 Harnessing protein-fold biodiversity (Phylomers) Collaboration with Avacta Affimers Collaboration with Medimmune scfvs Coding sequences fragments from 35 bacterial species Thermophillic/Extremophillic High propensity for self-folding subdomains Collaboration with Phylogica Ltd

11 3D bio-active protein folds for new target ID and drug discovery? Seeking the Goldilocks position from nature for high 3D-shape diversity and functionality to probe and then inform innovative low molecular weight NCE-discovery and design across the Site Diversity Site Affinity Site Size Phenotype! ( Pharmacophore ) living proteome

12 Allo-targeting Active site-directed small-molecule drugs competitively inhibit conserved catalytic sites Making it difficult to obtain selectivity, which can confound disease vs normal targeting High potency is also required to compete with natural ligands e.g., ATP in kinases Inhibitors only; principally against primary activity Allostery relies on non-competitive alteration of global 3 o /4 o protein structure Higher specificity & lower hurdles to modulation, especially for P/P interactions Allostery can be functionally effective in um range (c.f. nm for ATP-site directed inhibitors) Inhibitors or activators; primary or secondary activities

13 Historical hit rates for peptide libraries Phylomer hit rate data from Phylogica, unpublished

14 SITE SEEKER : Going rapidly from a target to valuable drug Value inflection point

15 SITE SEEKER : Mammalian cell functional screening PoC AP1 pathway downstream of KRAS Plasmid-based AP1 responsive luciferase construct Co-transfected into U2OS s with Phylomer plasmids 24hrs later PMA stimulated 4hrs later Luciferase read-out

16 SITE SEEKER : Primary AP1 screen data 3120 Phylomers screened well-by-well Pathway inducers and inhibitors obtained in the primary screen 25 pathway reducing hits; 14 repeat-tested positively High hit rate (1 in 200) for a peptide library 3D protein folds key

17 Hit validation data 14 confirmed hits reanalysed on AP1 FF-luc vs HSV TK Renilla-luc 6 hits show selective suppression of AP1 activity Other hits = general transcription inhibitors (>80% FF/Renilla ratio)

18 Hit validation data 3 hits retain activity on endogenous promoter 645nt promoter from sulfiredoxin1; with and without AP1 sites mutated No effect on Notch-ICD or FOXO-driven promoters

19 SITE SECURE : Target-ID

20 SITE SECURE : Target-ID Untransfected Empty vector Pyc36 6G8 25C3 48E6 3 hits taken into pull-downs V5-tagged Phylomers expressed in HEK293s Analysed by PAGE>Mass spec 25C3 (16aa) Phylomer hit: 4 clear bands on Coomassie 3 bands = same kinase Other band shares a common non-kinase domain

21 Target-confirmation Phenotype recapitulated by sirna to Kinase-A (orthogonal validation)

22 Allosteric binding-site confirmation NKD NKD Presumptive allosteric mechanism Tagged phylomer & Kinase-A non-kinase domain (NKD) expressed in U2OS cells IgG NKD IP d and immunoblotted NKD co-precipitates with Phylomer IgG

23 Allosteric binding-site confirmation Studied affinity of 25C3 binding to the Kinase-A non-kinase domain (NKD) Endogenous ligand Kd ~2x10-8 M by Biolayer Interferometry Endogenous ligand binds at Kd ~10-8 M Kinase NKD Phylomer Still determining whether it is a good target not progressing into chemistry yet NKD Substrate-P

24 PoC for drugging an allosteric site Targeting a known allosteric second site in a mitotic kinase Kinase-B previously drugged via ATP-site, but is non-selective w.r.t tumour vs normal cells Substrate-binding domain P/P interaction inhibitors with a narrower MoA may be better?

25 SITE BLOCK : Small molecule discovery Site-Block : Simple FP-based peptide displacement assay format Screened 120K small molecule library for hits Several hits bind an adjacent small druggable pocket

26 Novel KRAS Phenotype of allosteric kinase-b inhibitors In the LO stage; ~20nM on cells (+ve biomarkers) Allosteric-i has > KRAS synthetic lethality than ATPi

27 Novel resistance phenotype of allosteric kinase-b inhibitors ATP-site is easily mutated in cancers to abrogate drug binding Allosteric inhibitors are <<<< susceptible to resistance Allosteric ATPi Allosteric ATPi Many resistant colonies within 3-weeks for ATPi Small # of resistant colonies in 7-weeks with Allosteric-i Allosteric-i still works in ATPi resistant clones

28 SITE SEEKER : Moving to 10 9 shapes?

29 Emerging high-throughput PROTEINi formats Higher throughput pooled screens needed to access 10 9 library diversity Ideally covering multiple assay types; focussed >> holistic Good assay/selection methods will be key Nano BiT Protein:Protein Sensitive interaction reporters e.g., complementation Promega s NanoBiT Pathway: Turning Off signals into On signals In-house assay development Synthetic Lethal Viability drop-out screens NGS readout

30 Next-Gen sorting: Microfluidics 1) Single cells, each expressing a single protein-fold, added to individual picodroplets, and queued in a microfluidic biochip channel Picodroplets: i) FACS-like sorting rates ii) Multiple read-outs possible i) Luciferase or GFP ii) Mass-spec iii) Secreted protein detection iii) Enables high cell-viability 2) Luciferase reagent added sequentially to each individual picodroplet 3) Dark vs light cells sorted and collected for NGS determination of PROTEINi hits

31 Next-Gen sorting: Microfluidics 1) Single cells, each expressing a single protein-fold, added to individual picodroplets, and queued in a microfluidic biochip channel Picodroplets: i) FACS-like sorting rates ii) Multiple read-outs possible i) Luciferase or GFP ii) Mass-spec iii) Secreted protein detection iii) Enables high cell-viability 2) Luciferase reagent added sequentially to each individual picodroplet 3) Dark vs light cells sorted and collected for NGS determination of PROTEINi hits

32 Summary Core novel SiteSeeker Protein-interference platform to de-orphan undruggable targets/pathways Advantages over RNAi/CRISPR in assaying the proteome directly for cryptic druggable sites Early PoC stage: Good hit rates in small scale functional screens Commercial ops established to scale up screens & disease models seeking collaborations