Strictly-Lytic, Commensal-Sparing Phage Cocktail Editing of Fusobacterium nucleatum to De-Repress Antitumor Immunity in Checkpoint-Inhibitor Non-Responders

Funder / mechanism: NCI · U01 (Research Project Cooperative Agreement) Indication: Microbiome conditioning to convert immune-checkpoint-inhibitor (ICI) non-responders in colorectal cancer (CRC) Modality: Oral, GMP-grade, strictly lytic anti–Fusobacterium nucleatum bacteriophage cocktail, given as companion conditioning alongside anti–PD-1 Status: Concept for discussion — illustrative figures flagged [ILLUSTRATIVE]; not a submitted application

Project Summary / Abstract

Response to immune checkpoint inhibitors (ICIs) is shaped by the gut microbiome, and Fusobacterium nucleatum (Fn) is a mechanistically central "response-blunting" taxon: it recruits myeloid-derived suppressor cells (MDSCs), induces PD-L1, and dampens CD8⁺ T-cell infiltration, sustaining an immunosuppressive tumor microenvironment (TME). The tools available to "condition" the microbiome before immunotherapy cut the wrong way: broad-spectrum antibiotics indiscriminately deplete the very commensals ICI efficacy depends on, while fecal microbiota transplant (FMT) and diet are nonspecific and, for FMT, donor-dependent and transfer-risk-bearing. Bacteriophages are uniquely suited to this problem: they are receptor-specific predators that can subtract a single deleterious lineage while sparing the immunostimulatory community.

This U01 will develop and rigorously test an orally delivered, defined, strictly lytic anti-Fn phage cocktail as a precision companion conditioning agent given alongside checkpoint blockade. The mechanistic premise rests on two complementary preclinical proofs-of-concept from the same platform: (i) Fn-targeting phages that inhibit Fn growth and, conjugated to drug-loaded nanoparticles, augment first-line CRC chemotherapy in orthotopic and spontaneous mouse tumors, with a clean piglet safety profile [Zheng 2019]; and (ii) a non-lytic Fn-binding M13 silver-nanoparticle hybrid (M13@Ag) that scavenged Fn, reduced intratumoral MDSCs, and — combined with anti–PD-1 — significantly slowed tumor growth and prolonged survival in orthotopic CT26 CRC [Dong 2020]. M13@Ag establishes the immune mechanism (subtracting Fn de-represses anti-PD-1 efficacy); our innovation is to deliver that subtraction with a strictly lytic, self-amplifying, commensal-sparing oral cocktail suitable for GMP manufacture and outpatient use.

We will (1) assemble and characterize a clinical-grade lytic cocktail with quantified Fn coverage and measured commensal sparing; (2) define how Fn depletion remodels the TME and potentiates anti–PD-1 in orthotopic CRC; and (3) complete IND-enabling safety, biodistribution, and manufacturing (CMC) to position a biomarker-selected first-in-human conditioning trial. As an NCI U01, endpoints, immune-monitoring assays, and microbiome-sequencing standards will be harmonized with NCI immuno-oncology programs under shared go/no-go milestones. The outcome is a translational, GMP-trackable phage-conditioning candidate that turns the gut from an unpredictable variable into a tunable lever on cancer-immunity outcomes.

Specific Aims

Gut-microbiome composition is a determinant of anti–PD-1 response, yet the only tools to condition it before immunotherapy — antibiotics, FMT, diet — cannot subtract a single harmful taxon without collateral damage to the commensals ICIs require. Phages offer receptor-mediated strain specificity: a defined, strictly lytic cocktail can deplete Fn while sparing beneficial Ruminococcaceae/Faecalibacterium and short-chain-fatty-acid (SCFA) producers, "de-repressing" antitumor immunity so checkpoint blockade can work. We propose:

Aim 1. Assemble and characterize a defined, strictly lytic, commensal-sparing anti-Fn phage cocktail. We will compile strictly lytic phages against a panel of clinical Fn strains, quantify host range and lytic efficiency, and demonstrate selective depletion of Fn without reducing beneficial SCFA-producing commensals in defined consortia and complex stool-derived communities. We will pre-empt resistance via complementary-receptor cocktail design and depolymerase-bearing, biofilm-penetrating components. Go/no-go: a cocktail meeting pre-specified coverage and commensal-sparing thresholds.

Aim 2. Define how phage-mediated Fn depletion remodels the TME and potentiates anti–PD-1. In orthotopic CRC mouse models colonized with Fn, we will test whether oral cocktail dosing reduces intratumoral MDSC recruitment, lowers PD-L1, restores CD8⁺ T-cell infiltration, and — combined with anti–PD-1 — inhibits tumor growth and extends survival, benchmarked against the published M13@Ag Fn-scavenging result [Dong 2020]. Go/no-go: combination superiority over either agent alone on pre-specified immune and survival endpoints.

Aim 3. Complete IND-enabling safety, biodistribution, and CMC to enable a first-in-human conditioning trial. We will conduct GLP-style repeat-dose oral safety/biodistribution studies, manufacture cocktail to quality standards for clinical phage products with endotoxin/purity release specifications, and prepare an FDA emergency/expanded-access IND (eIND) pathway and IRB-reviewed protocol for ICI-eligible CRC patients selected by stool sequencing. Go/no-go: an acceptable safety package plus a release-spec'd, GMP-trackable cocktail.

Impact: Success would deliver the first precision microbiome-editing agent designed to convert predicted checkpoint-inhibitor non-responders into responders, establishing strictly lytic phage conditioning as an oral, outpatient adjunct to immuno-oncology.

Significance

ICIs have transformed oncology, but a majority of patients still do not respond, and gut-microbiome composition is an established modulator of that response. The microbiome is causal and editable — responder-derived FMT can produce objective responses in ICI-refractory disease — but FMT is donor-dependent, variable, and carries transfer risk, and it cannot be reduced to a defined release-specified product. Among response-blunting taxa, Fn is mechanistically central: it recruits MDSCs, induces PD-L1, and dampens CD8⁺ T-cell infiltration, building an immunosuppressive TME that blunts checkpoint blockade. The standard conditioning levers cut the wrong way — broad-spectrum antibiotics deplete exactly the Ruminococcaceae/Faecalibacterium and SCFA producers ICI efficacy depends on. There is therefore a clear unmet need for a subtractive, strain-specific tool.

Engineered and naturally lytic phages meet that need, and the preclinical evidence base is directly on point. Fn-targeting phages inhibit Fn growth and, as drug-conjugated nanoparticles, augment first-line CRC chemotherapy in orthotopic and spontaneous tumors, with negligible changes in piglet haemocyte counts, immunoglobulin/histamine levels, and liver/renal function [Zheng 2019] — establishing both target tractability and a favorable safety signal. Critically for immuno-oncology, the M13@Ag Fn-binding hybrid links Fn subtraction directly to reduced intratumoral MDSCs and improved anti–PD-1 efficacy with prolonged survival in orthotopic CT26 CRC [Dong 2020] — the exact causal chain a conditioning agent must exploit. The broader engineering toolkit (cocktail design, depolymerases/biofilm penetration, CRISPR-armed specificity, GMP manufacture, and documented synergy with checkpoint blockade) is reviewed in [Xu 2026].

An NCI U01 is the right vehicle: converting non-responders to responders requires harmonized immune-monitoring assays, microbiome-sequencing standards, and shared go/no-go criteria across a cooperative program — the infrastructure NCI immuno-oncology networks provide. (NIAID, given its phage-therapy and microbiome portfolios, is a logical alternate or partner home.)

Innovation

1. Conditioning, not cytotoxic adjunctivity. Prior phage work paired Fn depletion with chemotherapy [Zheng 2019] or used a metal-armed binding capsid [Dong 2020]. We reframe a defined, strictly lytic, orally delivered cocktail as pre-/co-immunotherapy microbiome editing whose explicit endpoint is conversion to ICI responsiveness.
2. Strictly lytic and self-amplifying — distinct from the published constructs. Unlike the non-lytic M13@Ag binding/scavenging hybrid (which depends on stoichiometric silver-nanoparticle payload) [Dong 2020], a replicating lytic cocktail amplifies at the target and carries no metal payload, simplifying CMC and the regulatory profile for repeat oral dosing.
3. Commensal-sparing by design — as a release criterion. Unlike antibiotics, the cocktail is validated to subtract Fn while quantitatively preserving SCFA producers, turning strain specificity into a measurable specification rather than an aspiration.
4. Mechanism-anchored immune readouts. We tie depletion to MDSC reduction, PD-L1 lowering, and CD8⁺ restoration as the causal chain enabling checkpoint blockade — the relationship M13@Ag demonstrated [Dong 2020].
5. Engineering headroom. The platform is compatible with biofilm-penetrating depolymerases for mucosal/tumoral biofilms, with M13-display Fn-binders as an orthogonal fallback, and with a roadmap toward CRISPR-armed precision editing [Xu 2026].

Approach

Aim 1 — Assemble and characterize a defined, strictly lytic, commensal-sparing anti-Fn phage cocktail

Rationale. The therapeutic advantage of phages over antibiotics is receptor-mediated strain specificity. Realizing it for an oral conditioning product requires broad coverage of clinical Fn strains, low resistance liability, strictly lytic lifecycle (no lysogeny/toxin-transfer risk), and demonstrable sparing of beneficial commensals [Xu 2026].

Experimental design. We will isolate and genomically vet strictly lytic anti-Fn phages (screening out temperate phages, integrases, and known virulence/AMR genes) and screen them against a panel of Fn clinical isolates [ILLUSTRATIVE: ~20–30 strains spanning subspecies]. We will quantify host range, efficiency of plating, and killing kinetics; combine complementary-receptor phages into a defined cocktail to suppress resistance; and incorporate depolymerase-bearing, biofilm-penetrating components. Selectivity will be assessed in defined consortia and in complex stool-derived communities by 16S/shotgun sequencing, with the pre-specified criterion that Fn is depleted while Ruminococcaceae/Faecalibacterium relative abundance and SCFA output are preserved. As an orthogonal binding modality (not part of the lytic release product), we will bank engineered M13-display Fn-binders [scaffold per Dong 2020] for use only if lytic coverage proves incomplete.

Expected outcomes. A characterized strictly lytic cocktail with documented coverage, defined resistance mitigation, and quantified commensal sparing — release-style specifications suitable for downstream CMC.

Pitfalls & alternatives. Phage resistance may emerge; we mitigate via multi-receptor cocktails and depolymerases and can iterate composition. If lytic coverage of some strains is incomplete, the banked M13-display Fn-binders provide an orthogonal target-binding mechanism [Dong 2020], explicitly flagged as a distinct (non-lytic) construct with its own CMC path.

Aim 2 — Define how phage-mediated Fn depletion remodels the TME and potentiates anti–PD-1

Rationale. Conditioning is valuable only if subtracting Fn causally de-represses antitumor immunity and improves checkpoint outcomes — the relationship M13@Ag demonstrated by scavenging Fn, reducing MDSCs, and enhancing anti–PD-1 [Dong 2020].

Experimental design. In orthotopic CRC mouse models colonized with Fn (e.g., CT26 orthotopic, as in [Dong 2020]), animals will receive oral cocktail, anti–PD-1, the combination, or vehicle/isotype controls [ILLUSTRATIVE: n per arm powered for survival and immune endpoints; pre-registered analysis plan]. Readouts: intratumoral MDSC frequency (CD11b⁺Gr-1⁺), PD-L1 expression, CD8⁺ T-cell infiltration and function, tumor growth, and survival, with paired gut-community sequencing to link bacterial depletion to immune change. The published M13@Ag Fn-scavenging combination serves as the qualitative mechanistic benchmark [Dong 2020]; we do not assume its effect size transfers to a lytic cocktail.

Expected outcomes. Oral cocktail dosing reduces MDSC recruitment and PD-L1, restores CD8⁺ infiltration, and — combined with anti–PD-1 — slows tumor growth and extends survival beyond either agent alone, establishing the causal conditioning chain.

Pitfalls & alternatives. Murine colonization and TME imperfectly model human disease; we will use orthotopic models and, where feasible, humanized-microbiome approaches. If lysis alone yields modest immune shifts, the orthogonal M13-display binders (with capsid-mediated APC engagement) offer a synergy-enhancing alternative [Dong 2020; Xu 2026]. Conditioning effect sizes are not yet established in this context; the program is powered as preclinical/hypothesis-testing, not confirmatory.

Aim 3 — Complete IND-enabling safety, biodistribution, and CMC to enable a first-in-human conditioning trial

Rationale. Translation requires GLP-style safety, manufacturing consistency, and a regulatory route. Prior piglet work indicates a favorable safety profile for Fn-targeting phage delivery [Zheng 2019], and the engineering literature defines GMP expectations for clinical phage products [Xu 2026].

Experimental design. We will perform repeat-dose oral safety and biodistribution studies; develop a manufacturing/QC process with endotoxin, purity, identity, and potency release specifications consistent with clinical phage-product quality expectations [Xu 2026]; and prepare an FDA eIND/expanded-access submission plus an IRB-reviewed protocol for ICI-eligible CRC patients, including stool-sequencing-based patient selection for Fn and response-blunting taxa.

Expected outcomes. A safety/biodistribution package, a GMP-trackable cocktail with release specs, and a regulatory/clinical dossier enabling a biomarker-selected first-in-human conditioning study.

Pitfalls & alternatives. Phage immunogenicity/neutralization could affect dosing; oral mucosal delivery and cocktail rotation mitigate this. If a full traditional IND is premature, the eIND/expanded-access route enables initial supervised human experience.

Timeline

[ILLUSTRATIVE] Year 1: strictly lytic cocktail assembly, genomic safety vetting, host-range/selectivity (Aim 1). [ILLUSTRATIVE] Years 2–3: mechanism and anti–PD-1 efficacy studies (Aim 2); initiate safety. [ILLUSTRATIVE] Years 4–5: IND-enabling GLP safety/biodistribution, CMC, and eIND/IRB package (Aim 3). Annual NCI cooperative-program milestone reviews with pre-specified go/no-go criteria at each Aim transition.

Budget Justification (modular sketch)

[ILLUSTRATIVE] Requested at [ILLUSTRATIVE: $X] direct costs/year in modular increments. Personnel [ILLUSTRATIVE]: contact PI, phage biologist, tumor immunologist, microbiome bioinformatician, GLP/regulatory-CMC lead, study coordinator. Other costs: phage isolation/genomic vetting and sequencing (Aim 1); orthotopic immuno-oncology studies with flow/IHC (Aim 2); GLP safety/biodistribution, GMP-style manufacturing/QC, and regulatory filing (Aim 3). Animal costs under Aims 2–3 [ILLUSTRATIVE]. Cooperative-agreement travel for NCI program meetings [ILLUSTRATIVE]. All figures are placeholders pending institutional budgeting.

Vertebrate Animals

Animal work is proposed. Mouse orthotopic CRC models colonized with Fn will be used for mechanism and efficacy (Aim 2), and rodent (and, if warranted, large-animal) studies for GLP safety/biodistribution (Aim 3), consistent with prior piglet safety work [Zheng 2019]. Justification: in vitro systems cannot capture MDSC/CD8⁺ TME remodeling or systemic safety. Group sizes [ILLUSTRATIVE] will be the minimum needed for rigorous, pre-registered endpoints; humane endpoints, analgesia, and IACUC oversight apply, with explicit attention to replacement, reduction, and refinement.

Human Subjects / Clinical Trial

No human dosing occurs in this award; Aim 3 culminates in a clinical-readiness package. The planned first-in-human conditioning study would enroll ICI-eligible CRC patients [ILLUSTRATIVE: enrollment to be determined] selected by stool sequencing for Fn and response-blunting taxa. Because the cocktail is investigational, initial human use is anticipated via the FDA emergency/expanded-access IND (eIND) route, under full IRB oversight, informed consent, and a data-safety monitoring plan, with manufacturing aligned to clinical phage-product quality expectations [Xu 2026].

Team & Environment

[Template — to be completed with real names/institutions.] Contact PI: [name/institution] (phage therapy / microbiome editing). MPI/Co-I: [name] tumor immunology/checkpoint blockade; [name] phage engineering (lytic-phage genomics, depolymerases, M13 display); [name] microbiome bioinformatics; [name] GLP/regulatory and CMC. Collaborating/translational sites: an academic phage-therapy center [ILLUSTRATIVE] for clinical translation, with platform expertise informed by published Fn-targeting phage-microbiota work [Zheng 2019; Dong 2020]. Environment: BSL-appropriate phage and anaerobe facilities, orthotopic immuno-oncology models, flow/IHC and sequencing cores, and GLP/GMP-capable partners. As a U01, the team will participate in NCI cooperative-program governance, shared assays, and milestone review.

References

1. Zheng DW, Dong X, Pan P, Chen KW, Fan JX, Cheng SX, Zhang XZ. Phage-guided modulation of the gut microbiota of mouse models of colorectal cancer augments their responses to chemotherapy. Nature Biomedical Engineering. 2019;3(9):717–728. https://pubmed.ncbi.nlm.nih.gov/31332342/
2. Dong X, Pan P, Zheng DW, Bao P, Zeng X, Zhang XZ. Bioinorganic hybrid bacteriophage for modulation of intestinal microbiota to remodel tumor-immune microenvironment against colorectal cancer (M13@Ag). Science Advances. 2020;6(20):eaba1590. https://pubmed.ncbi.nlm.nih.gov/32440552/
3. Xu M, Chen S, Pei H, Hu L, Zhang Y. Engineering bacteriophages for gut health: precision antimicrobials and beyond. Journal of Nanobiotechnology. 2026. PMCID: PMC12829040. https://pmc.ncbi.nlm.nih.gov/articles/PMC12829040/