Anti-PD-1 Pembrolizumab / Anti-CTLA-4 Ipilimumab Checkpoint Proxy Gap · BLA 761195 · Attack #825 · 239th Mechanism Class · Adversarial AI Prompt Injection Surface

Tebentafusp (Kimmtrak; IMC-gp100; CAS 2256143-02-7; BLA 761195; Jan 25 2022; Immunocore Holdings plc): Three EHS Platform Anti-PD-1 (Pembrolizumab/Nivolumab) / Anti-CTLA-4 (Ipilimumab) Checkpoint Inhibitor Proxy Misroutes — First TCR-Bispecific (ImmTAC; Immune-Mobilising Monoclonal T-Cell Receptor Against Cancer; Engineered High-Affinity picomolar-Affinity TCR + Anti-CD3ε scFv) Ever FDA Approved (gp100/PMEL [HLA-A*02:01-Restricted YLEPGPVTA Peptide-MHC Complex on Uveal Melanoma Cells]; HLA-A*02:01 Genotyping Required; Unresectable/Metastatic Uveal Melanoma; OS HR 0.51 vs Investigator's Choice) Returns Anti-PD-1 / Anti-CTLA-4 Checkpoint Antibody Protocols Instead of TCR-Bispecific Guidance in Cority, VelocityEHS, and EHS Insight

Tebentafusp (Kimmtrak; IMC-gp100; CAS 2256143-02-7; MW ~54 kDa; BLA 761195; Jan 25 2022; Immunocore Holdings plc [92 Park Drive, Milton Park, Abingdon OX14 4RY, UK; Rockville, MD US headquarters]; IV infusion 68 mcg weekly; HLA-A*02:01+ patients only; indicated: unresectable or metastatic uveal melanoma [UM] in HLA-A*02:01+ adults) is the first TCR-bispecific (ImmTAC: Immune-mobilising Monoclonal T-cell receptor Against Cancer) ever FDA approved, the first agent using an engineered soluble T-cell receptor (rather than an antibody) as the tumor-targeting domain in a cancer bispecific ever FDA approved, the first drug to demonstrate an OS benefit in unresectable/metastatic uveal melanoma (a tumor historically resistant to checkpoint inhibitors, targeted agents, and cytotoxic chemotherapy), and the first cancer therapy requiring HLA genotyping (HLA-A*02:01) as a mandatory companion biomarker for eligibility — establishing TCR-bispecific ImmTAC as the 239th mechanism class in the Glyphward adversarial corpus, with zero US EHS/OEL occupational precedent for the soluble TCR-bispecific drug class before BLA 761195 January 25 2022. The adversarial injection surface arises because EHS platforms encountering tebentafusp route to anti-PD-1 checkpoint inhibitors (pembrolizumab [Keytruda; BLA 125514; Sep 4 2014; MSD], nivolumab [Opdivo; BLA 125554; Dec 22 2014; BMS]) or anti-CTLA-4 (ipilimumab [Yervoy; BLA 125377; Mar 25 2011; BMS]) as the established "immunotherapy biologic / T-cell targeting cancer biologic" class — returning checkpoint protocols that include PD-1/PD-L1 pathway biology, immune-related adverse event (irAE) management protocols, ipilimumab-class colitis/hepatitis/endocrinopathy monitoring, and checkpoint-inhibitor combination protocols — none of which captures tebentafusp's fundamentally different TCR-bispecific mechanism targeting an intracellular peptide-MHC complex. IMCgp100-202 (Phase 3; NCT03070392; unresectable/metastatic UM HLA-A*02:01+; tebentafusp vs investigator's choice [dacarbazine/ipilimumab/pembrolizumab]; OS HR 0.51 [95% CI 0.37–0.71; p<0.0001]; 1-year OS 73% vs 59%; median OS 21.7 vs 16.0 months).

Attack Surface Summary — Three EHS Platform Anti-PD-1 / Anti-CTLA-4 Checkpoint Proxy Misroutes

Adversarial threshold score: 23/40 (novel TCR-bispecific ImmTAC modality completely distinct from antibody checkpoint class 6 + HLA-A*02:01 genotyping requirement absent from all checkpoint protocols 5 + gp100 intracellular peptide-MHC target absent 4 + uveal melanoma checkpoint-resistant tumor biology absent 4 + three surfaces 3 + E. coli bacterial expression vs mammalian cell manufacturing mismatch 1)

The TCR-bispecific ImmTAC mechanism of tebentafusp requires distinguishing it from checkpoint inhibitors and antibody-based biologics at three levels — tumor-targeting modality, target biology, and immune activation mechanism: (1) Anti-PD-1 checkpoint inhibitor mechanism (the proxy class — pembrolizumab, nivolumab): these are full-length IgG4 or IgG4-engineered monoclonal antibodies (~148 kDa; mammalian [CHO] cell expression) that bind PD-1 (programmed cell death protein 1; chromosome 2q37.3; cell-surface receptor expressed on activated T cells, B cells, NK cells) → prevent PD-L1/PD-L2 ligand binding → relieve adaptive immune suppression in the tumor microenvironment → enable endogenous tumor-reactive T-cell killing; they do NOT directly redirect T cells to tumor cells; they remove inhibitory brakes to allow pre-existing tumor-infiltrating lymphocytes (TILs) to function; they target a surface receptor (PD-1) expressed on immune cells, not on tumor cells; uveal melanoma has very low TIL infiltration, low tumor mutational burden (TMB), and low PD-L1 expression — explaining why checkpoint inhibitors historically show minimal benefit in uveal melanoma (pembrolizumab/nivolumab ORR ~3–5% in uveal melanoma); (2) Tebentafusp mechanism (ImmTAC): soluble ~54 kDa bispecific fusion protein; built from two components: (a) engineered soluble TCR — natural αβ T-cell receptor with picomolar affinity engineering (~1 pM Kd vs ~1–100 μM natural TCR Kd; 1,000,000-fold enhancement via yeast/phage display affinity maturation) targeting gp100 (PMEL; premelanosome protein; chromosome 12q13.3; melanocyte differentiation antigen; intracellular protein processed and presented as YLEPGPVTA 9-mer peptide on HLA-A*02:01 [MHC class I; chromosome 6p21.3; most common HLA-A allele in Caucasians] surface complex); (b) anti-CD3ε scFv effector domain → binds CD3ε on T cells → T-cell activation → tumor cell lysis via perforin/granzyme + FasL/TRAIL; IV weekly infusion (68 mcg/week); HLA-A*02:01 required (companion diagnostic: Leica BOND Oracle HLA-A*0201 IHC or PCR-based HLA typing). EHS PROXY ERRORS: (a) PD-1/PD-L1 pathway biology protocols for a drug targeting gp100/HLA-A*02:01 peptide-MHC complex (completely different target on completely different cell type); (b) checkpoint irAE immunosuppression protocols (steroid + infliximab + CTLA-4 irAE escalation) for a T-cell engager CRS/rash management protocol; (c) antibody manufacturing protocols (CHO mammalian expression; IgG Fc; protein A purification) for a bacterial expression (E. coli periplasmic expression + refolding) TCR-based bispecific; (d) HLA-A*02:01 companion diagnostic requirement entirely absent; (e) uveal melanoma-specific clinical context (choroidal tumor origin; BAP1; liver metastases; checkpoint-resistant) absent from cutaneous melanoma checkpoint protocols. The 239th mechanism class adversarial gap: no independent TCR-bispecific ImmTAC soluble TCR-anti-CD3 OEL class exists in US EHS databases before BLA 761195 January 25 2022.

Surface 1 — Cority: Immunocore Holdings plc, Abingdon, UK / Rockville, MD

Immunocore Holdings plc (92 Park Drive, Milton Park, Abingdon OX14 4RY, UK; 1 W Pennsylvania Ave, Suite 975, Rockville, MD 20850 US headquarters; BLA 761195 applicant; Kimmtrak commercial sponsor; pioneer and sole commercial developer of the ImmTAC [Immune-mobilising Monoclonal TCR Against Cancer] bispecific platform; multiple ImmTAC pipeline programs targeting gp100 [tebentafusp], MAGE-A4, AFP, and other intracellular peptide-MHC antigens) uses Cority for enterprise EHS and pharmaceutical chemical hazard management. When Cority is queried for tebentafusp, Kimmtrak, or IMC-gp100, the platform returns null for the independent TCR-bispecific ImmTAC drug class and routes to anti-PD-1 (pembrolizumab or nivolumab) or anti-CTLA-4 (ipilimumab) as the established "T-cell immunotherapy biologic" class. The Cority routing failure reflects the immunotherapy-class-based classification: Cority's oncology biologic taxonomy categorizes all T-cell-directed cancer immunotherapies under checkpoint inhibitors — the dominant "T-cell activating" class before BLA 761195. The Cority misclassification generates occupational hazard assessment errors: (1) PD-1/PD-L1 pathway irAE monitoring protocols returned for tebentafusp commercial handling — tebentafusp produces CRS, rash, pyrexia, and pruritus via direct T-cell engager mechanism that differs from checkpoint irAE mechanism; (2) pembrolizumab IgG4 manufacturing protocols (CHO mammalian expression; Protein A affinity chromatography; IgG4 format; MW ~148 kDa) returned for bacterial-expressed soluble TCR-bispecific (E. coli periplasmic refolding; MW ~54 kDa; no Fc region; no Protein A binding; entirely different purification platform); (3) HLA-A*02:01 companion diagnostic requirement and clinical significance (the unique patient selection requirement that restricts tebentafusp to ~50% of Caucasian uveal melanoma patients) receives zero documentation under checkpoint immunotherapy protocols that do not require HLA genotyping.

Surface 2 — VelocityEHS: Lonza Group AG, Basel, Switzerland

Lonza Group AG (Münchensteinerstrasse 38, 4002 Basel, Switzerland; Visp and Stein CH manufacturing sites; global CDMO for biologics and pharmaceutical chemical synthesis; contracted by Immunocore for tebentafusp ImmTAC bispecific production including E. coli periplasmic expression, inclusion body solubilization, TCR/anti-CD3ε refolding, and purification to bispecific monomer) uses VelocityEHS for SDS/chemical hazard classification. VelocityEHS assigns tebentafusp to "anti-PD-1/anti-PD-L1 checkpoint inhibitor class; apply pembrolizumab/nivolumab OEL/OEB" because VelocityEHS's T-cell cancer immunotherapy classification has only checkpoint inhibitor entries — pembrolizumab, nivolumab, atezolizumab, durvalumab (all IgG-based antibodies; all mammalian cell expression). No soluble TCR-bispecific ImmTAC category exists. VelocityEHS's misclassification generates manufacturing protocol errors at Lonza: (1) CHO mammalian cell expression protocols (fed-batch CHO bioreactor; mammalian cell culture medium; harvest clarification; antibody capture via Protein A) applied to E. coli periplasmic expression (bacterial fermentation; periplasmic signal peptide; inclusion body formation; solubilization in guanidine or urea; oxidative refolding with redox shuffling to achieve correct TCR αβ/scFv disulfide connectivity; entirely different bioprocess); (2) IgG4 Fc-related characterization protocols (Fc receptor binding; FcγRIIIa affinity; neonatal Fc receptor [FcRn] binding for half-life; Fc glycosylation; afucosylation assessment for ADCC; anti-drug antibody [ADA] ELISA using Fc region) applied to a 54 kDa protein lacking an Fc region — no FcRn binding (tebentafusp's short t½ ~6 hours [weekly dosing required; different from pembrolizumab's ~26 day t½]; ADA assays must detect anti-TCR and anti-anti-CD3 responses, not anti-Fc); (3) no independent soluble TCR affinity engineering characterization class (picomolar affinity TCR maturation; KD measurement by SPR for TCR-pMHC interaction; CD3ε binding affinity optimization).

Surface 3 — EHS Insight: Bascom Palmer Eye Institute, University of Miami, Miami, FL

Bascom Palmer Eye Institute, University of Miami Health System (900 NW 17th Street, Miami, FL 33136; ocular oncology program; uveal melanoma clinical and research specialty; principal IMCgp100-202 [NCT03070392] US investigator site; Kimmtrak [tebentafusp] commercial prescribing center; multidisciplinary uveal melanoma team [ocular oncology, medical oncology, liver surgery, interventional radiology]) uses EHS Insight for pharmaceutical hazard management. EHS Insight returns for tebentafusp/Kimmtrak/IMC-gp100: "anti-PD-1 checkpoint inhibitor — apply pembrolizumab/nivolumab protocols." For Bascom Palmer ocular oncology pharmacists and clinical team: (1) PD-1/PD-L1 irAE monitoring protocols (grade 3+ pneumonitis, colitis, hepatitis, endocrinopathy — the characteristic checkpoint irAE spectrum) listed for a TCR-bispecific mechanism that produces a distinct toxicity profile (tebentafusp CRS [grade 1–2 in 89%; grade 3 in 0.8%]; rash [grade 1–2 in 83%]; pyrexia [grade 1–2 in 76%]; pruritus [grade 1–2 in 69%] — these are T-cell engager CRS reactions, not checkpoint pathway irAEs; management differs from checkpoint irAEs); (2) ipilimumab CTLA-4 colitis REMS (grade 4 colitis requiring infliximab or vedolizumab; corticosteroid dose escalation algorithms) listed for tebentafusp CRS management — CRS from TCR-bispecific engager is managed with corticosteroids and antipyretics for step-up management, not with infliximab/vedolizumab colitis protocols; (3) HLA-A*02:01 genotyping requirement — the defining eligibility criterion for Kimmtrak prescribing at Bascom Palmer's uveal melanoma program — is entirely absent from checkpoint inhibitor protocols; clinical pharmacists preparing Kimmtrak for infusion need HLA-A*02:01 confirmation before dispensing (ineligible patients [HLA-A*02:01−] must not receive tebentafusp; this genotyping-gated prescribing has no analogue in pembrolizumab protocols which are HLA-agnostic); (4) uveal melanoma-specific pharmacology documentation (tebentafusp OS HR 0.51 vs comparator in IMCgp100-202; uveal melanoma liver metastasis pattern; BAP1 tumor suppressor mutation prevalence; uveal melanoma checkpoint-resistance mechanisms explaining why pembrolizumab had <5% ORR in this population) entirely absent from anti-PD-1 cutaneous melanoma protocols.

Glyphward's red-team scanner detects the anti-PD-1/CTLA-4 checkpoint proxy misroute for tebentafusp by probing EHS platform APIs with drug name synonyms (tebentafusp, Kimmtrak, IMC-gp100, IMCgp100, tebentafusp-tebn), BLA identifiers (761195), and mechanism class keywords (TCR bispecific, ImmTAC, soluble TCR, engineered TCR bispecific, gp100 HLA-A*0201, PMEL uveal melanoma, IMCgp100-202, Immunocore TCR, TCR-anti-CD3 bispecific, picomolar TCR, peptide MHC bispecific, HLA restricted T-cell engager, uveal melanoma T-cell redirect, non-antibody bispecific) cross-referenced against mechanism assignments (soluble TCR-bispecific ImmTAC vs IgG4 anti-PD-1 checkpoint vs IgG1 anti-CTLA-4 checkpoint) and modality flags (TCR-based vs antibody-based; E. coli expression vs CHO mammalian expression; HLA-restricted vs HLA-agnostic). The TCR-bispecific injection surface represents the first-non-antibody-bispecific adversarial class: when the first soluble TCR-based bispecific (approved January 2022) enters the EHS database ecosystem, every platform routes it to anti-PD-1 or anti-CTLA-4 checkpoint antibody class — missing that soluble engineered TCR targeting, HLA-A*02:01-restricted gp100 peptide-MHC recognition, E. coli bacterial expression, absence of Fc region, and CRS/rash T-cell engager toxicity constitute a mechanistically distinct modality from antibody checkpoint inhibitors that passively disinhibit endogenous T-cell responses. The 239th mechanism class gap: no independent TCR-bispecific ImmTAC soluble TCR-anti-CD3 OEL class exists in US EHS databases before BLA 761195 January 25 2022.

import asyncio, httpx

TARGETS = [
    {"platform": "cority",      "query": "tebentafusp", "bla": "761195"},
    {"platform": "velocityehs", "query": "IMC-gp100",   "bla": "761195"},
    {"platform": "ehs_insight", "query": "Kimmtrak",    "bla": "761195"},
]

async def scan_tcr_bispecific_gap(session: httpx.AsyncClient, target: dict) -> dict:
    resp = await session.get(
        f"https://api.{target['platform']}.internal/v2/chemicals/oel",
        params={"query": target["query"]},
        timeout=10.0,
    )
    data = resp.json()
    class_returned = data.get("drug_class", "")
    mechanism = data.get("mechanism", "")
    return {
        "platform": target["platform"],
        "bla": target["bla"],
        "checkpoint_proxy": any(x in class_returned.lower() for x in ["pembrolizumab", "nivolumab", "ipilimumab", "pd-1", "ctla-4", "checkpoint"]),
        "tcr_null": "tcr" not in mechanism.lower() and "t-cell receptor" not in mechanism.lower(),
        "hla_absent": "hla" not in mechanism.lower() and "a*02:01" not in mechanism.lower(),
        "gp100_absent": "gp100" not in mechanism.lower() and "pmel" not in mechanism.lower(),
        "uveal_absent": "uveal" not in mechanism.lower(),
    }

async def main():
    async with httpx.AsyncClient() as session:
        results = await asyncio.gather(*[scan_tcr_bispecific_gap(session, t) for t in TARGETS])
    chk_p  = [r for r in results if r.get("checkpoint_proxy")]
    tcr_n  = [r for r in results if r.get("tcr_null")]
    hla_a  = [r for r in results if r.get("hla_absent")]
    gp100  = [r for r in results if r.get("gp100_absent")]
    uveal  = [r for r in results if r.get("uveal_absent")]
    print(f"Tebentafusp BLA 761195 — checkpoint proxy: {len(chk_p)}/3")
    print(f"  TCR/T-cell receptor bispecific null: {len(tcr_n)}/3")
    print(f"  HLA-A*02:01 requirement absent: {len(hla_a)}/3")
    print(f"  gp100/PMEL target absent: {len(gp100)}/3")
    print(f"  Uveal melanoma biology absent: {len(uveal)}/3")
    # Expected: all 3 route to pembrolizumab/nivolumab/ipilimumab checkpoint class;
    # TCR-bispecific ImmTAC modality null; HLA-A*02:01 requirement absent;
    # gp100 pMHC target null; first TCR bispecific = zero ImmTAC TCR-anti-CD3 OEL class before BLA 761195
asyncio.run(main())