HER2 Bispecific T-Cell Engager Design for Solid Tumor Immunotherapy

Oncology

HER2 Bispecific T-Cell Engager Design for Solid Tumor Immunotherapy

Clinical-stage oncology biotech needed a developable HER2×CD3 bispecific with controlled cytokine release profile before IND-enabling studies.

Company

Clinical-stage oncology biotech (Series C, 55-person team)

Timeline

August to September 2025

Engagement

Bispecific Antibody Engineering Pipeline

Bispecific Antibody Design

5 wks: Computational delivery
82%: Developability pass rate (top 30)
6: Functional engagers confirmed
18×: Faster vs. prior format screen

The Challenge

A Series C oncology company was advancing a HER2-targeted T-cell engager for gastric and breast cancer indications. Their prior bispecific format screen (scFv-Fc, DART, and knob-into-hole IgG) had consumed 7 months and $890K in CRO expression costs with only 1 lead showing acceptable cytotoxicity but unacceptable aggregation at clinical concentrations. They needed a ranked set of HER2×CD3 bispecific designs with developability and cytokine release risk scored before the next IND amendment window.

Business Constraints

Budget: $520K (remaining biologics discovery budget for H2 2025)
Timeline: Ranked bispecific sequences in 5 weeks
Must maintain HER2 selectivity over HER3/HER4 and control CD3 affinity to limit CRS risk

ProteinForge Approach

Week 1: Format Selection and Arm Optimization

Input

HER2 ECD structure (PDB: 1N8Z) and trastuzumab-HER2 co-crystal
CD3εδ complex structure (PDB: 1XIW)
Prior failed bispecific sequences and aggregation data from CRO campaign

Methods

Format enumeration across 6 bispecific architectures (2+1 IgG, CrossMab, knob-into-hole, scFv-Fc)
HER2 arm affinity tuning (target KD 1–5 nM) with HER3/HER4 cross-reactivity screen
CD3 arm affinity windowing (target KD 50–200 nM) for CRS risk mitigation

Output

Top 4 formats ranked by predicted developability and geometry
240 HER2 arm variants and 180 CD3 arm variants per format

Week 2 to 3: Interface Modeling and Developability Filtering

AlphaFold-Multimer predicted 960 HER2×CD3 bispecific assemblies across format and arm combinations. Interface stability scoring (ΔΔG, H-bond network, buried surface area) filtered to 142 assemblies. Developability scoring (aggregation, viscosity at 150 mg/mL, immunogenicity, CHO expression probability) reduced pool to 48 candidates. Cytokine release risk model (CD3 affinity + format geometry features) flagged 11 high-risk designs for deprioritization.

Week 4 to 5: Ranking and Validation Protocol

Output

Top 40 bispecific sequences ranked by cytotoxicity potential, developability, and CRS risk tier
Expression priority list for top 15 (CHO-compatible sequences with linker optimization)
Recommended validation protocol: HER2+ cell killing (SK-BR-3, NCI-N87), cytokine panel (IFN-γ, IL-2, TNF-α)
Format comparison brief for regulatory team with developability rationale

Final Ranked Bispecific Designs

Top candidates shown; full list of 40 delivered with expression-ready sequences.

Top HER2×CD3 engagers by predicted potency, developability, and CRS risk
Rank	Format	HER2 KD (nM)	CD3 KD (nM)	Developability	CRS Risk	Status
1	2+1 CrossMab	2.1	85	0.89	Low	Priority A
2	Knob-into-hole IgG	1.8	120	0.87	Low	Priority A
3	2+1 CrossMab	3.4	95	0.86	Low	Priority A
4–5	Mixed	2.5–4.0	70–140	0.82–0.85	Low–Moderate	Priority B
6–10	Mixed	1.5–5.5	60–180	0.78–0.84	Moderate	Priority B
11–40	Mixed	1–10	40–250	0.70–0.82	Mixed	Backup

Results and impact

Speed, validation, and business outcomes

Speed vs. Prior Bispecific Format Screen

Metric	Prior CRO Format Screen	ProteinForge	Improvement
Timeline	7 months	5 weeks	6× faster
Cost	$890K	$520K	42% savings
Designs Evaluated	24 expressed constructs	960 assemblies screened	40× larger search
Functional Engagers	1 (with aggregation)	6 confirmed actives	6× more viable leads

Wet-Lab Validation Outcomes (10 weeks post-delivery)

Expression, cytotoxicity, and cytokine release results on top 12 expressed constructs.

Construct	Predicted HER2 KD	Observed EC50 (pM)	Cell Killing?	CRS Panel	Notes
Rank 1	2.1 nM	42	Yes	Low	No aggregation at 150 mg/mL; Tm 68°C
Rank 2	1.8 nM	58	Yes	Low	Strong SK-BR-3 killing; acceptable viscosity
Rank 3	3.4 nM	71	Yes	Low	Best CHO yield (3.1 g/L)
Rank 4	2.5 nM	95	Yes	Moderate	Acceptable for backup series
Rank 5	4.0 nM	180	Yes	Moderate	Moderate potency; excellent stability
Rank 6–8	Mixed	Active	Yes (3/3)	Low–Moderate	3 additional functional engagers
Rank 9–12	Mixed	Weak/None	1/4	High (2/4)	High CRS risk flagged correctly pre-expression

82%: Developability pass rate (top 30)
6/12: Functional T-cell engagers confirmed
75%: CRS risk tier prediction accuracy
10 wks: Timeline to first functional engager

Immediate Wins

IND amendment supported: Rank 1 bispecific advanced to GLP tox with developability data package
Avoided repeat CRO spend: $370K saved by filtering aggregation-prone formats in silico
Board update: presented 6 validated engagers with CRS risk stratification at Q3 review

Strategic Advantages

CD3 affinity windowing prevented ultra-high-affinity engagers linked to cytokine release in prior campaign
Format comparison brief gave regulatory team developability rationale for 2+1 CrossMab selection
HER3/HER4 cross-reactivity screen eliminated 19 off-target binders before expression

Follow-on engagement
Q1 2026: Fc silencing and half-life extension on Rank 1 engager via ProteinForge. Estimated cost: $285K. Target: clinical candidate nomination within 8 months.

Model validation

Lessons and recommendations

What Worked

CD3 affinity as primary CRS predictor outperformed format-only heuristics (AUC 0.82 vs. 0.61)
Viscosity-at-concentration scoring caught 4 candidates that passed standard aggregation filters
Negative training from prior failed CRO sequences improved developability model calibration

Challenges and Mitigations

Rank 7 showed strong cytotoxicity but elevated IFN-γ in cytokine panel despite moderate CD3 affinity. Root cause: avidity effects from 2+1 format geometry.

Mitigation: Added format-specific avidity correction to CRS model; re-tiered 2+1 designs with geometry features.

Two knob-into-hole constructs mispaired during expression despite in silico compatibility scores.

Mitigation: Integrated experimental mispairing data into CHO expression probability model; flagged low-confidence hole-arm pairings.

When to use ProteinForge for bispecific design

T-cell engager programs with CRS or developability constraints
Prior bispecific format screens that produced binders but failed manufacturability
Timeline pressure before IND or clinical milestone deadlines
Programs requiring simultaneous optimization of both binding arms and format geometry

ROI: approximately 5:1 (cost savings + avoided repeat CRO campaign + faster IND timeline).

Next steps: advance top engager through Fc engineering and stability optimization; pair with patient-derived organoid validation if available.

About This Engagement

Client profile: Series C oncology biotech, 55 employees, 2 approved INDs
Project duration: 5 weeks (computational delivery) + 10 weeks (validation)
Total cost: $520K
Date: August to September 2025

This case study is anonymized at client request. Bispecific sequences and institutional affiliations have been redacted. Full protocols available under NDA.

Run a similar program with ProteinForge.

Tell us about your target. We will scope a 2 to 4 week pilot.

Schedule a Consultation [email protected]