NF27 Eradicated Some Mouse Tumors—Why That Is Not Yet a Cancer Treatment
NF27 eliminated measurable tumors in some mice in a peer-reviewed study, but it is not a cancer treatment and has not been tested in people. The strongest result came from small, short murine experiments in which researchers injected the peptide directly into established tumors. That result supports additional pharmacology and toxicology work. It does not establish human efficacy, durable remission, systemic safety, or clinical usefulness.
Furukawa and colleagues published the peer-reviewed NF27 paper online in the Journal of Medicinal Chemistry on October 23, 2026. The work moved through a sensible preclinical sequence: optimize a peptide, test membrane disruption and cell viability, compare cancer and noncancer cells, examine immune signals, and then test several mouse tumor models. A lower cell-assay EC50 does not prove tumor control; a mouse tumor response does not prove human benefit. None can substitute for a clinical trial.
Quick answer: The NF27 oncolytic peptide cancer study supports NF27 as a preclinical lead. It does not show that NF27 treats cancer in humans. No oncolytic peptide is clinically approved for cancer treatment.
This article is educational only. It does not provide medical advice, treatment instructions, dosing, synthesis, sourcing, purchasing, injection, or self-experimentation guidance.
A result can be convincing within a mouse model and still sit several evidence gates away from a human treatment.
What is an oncolytic peptide?
An oncolytic peptide is a short peptide studied for its ability to damage cancer-cell membranes and trigger cell death. Many are positively charged and amphipathic, meaning different regions interact with water and lipid membranes. Cancer-cell surfaces can differ from healthy-cell surfaces in charge, lipid composition, membrane area, and junctional organization. Those differences may create selectivity, but they do not guarantee it.
NF27 was derived from CKS1, a chemokine-derived peptide previously reported to have oncolytic activity. For study identification, NF27 is a 23-residue peptide with the sequence NGRKAKLNPASIVKKIIKKILKS; the research material was C-terminally amidated unless otherwise stated. That sequence is reported here only to identify the experimental molecule—not as synthesis or use guidance.
The final paper compares NF27 with LTX-315, a separate oncolytic peptide that has entered clinical trials. LTX-315 is a research comparator, not an approved standard treatment, and its presence in the study does not make NF27 clinically validated.
What did the study actually test?
The study assembled multiple evidence layers, from structure–activity optimization to local tumor experiments in mice. The important reading habit is to keep those layers separate.
Six connected experiments do not become one clinical result. Each station has its own endpoint and limitations.
| Evidence layer | What was measured | What it cannot establish |
|---|---|---|
| CKS1-derived optimization | How sequence and predicted structure related to cell-killing activity | Human efficacy or safety |
| Membrane and viability assays | Rapid membrane leakage, LDH release, metabolic viability, mitochondrial potential, caspase-3 | Tumor control in a living organism |
| Cancer-versus-healthy-cell comparisons | Relative cytotoxicity in selected cell lines and NCI-60 screening | Selectivity across normal human tissues |
| Hemolysis | Red-blood-cell rupture ex vivo | Whole-body toxicity |
| Immune readouts | ATP release, dendritic-cell cytokines, immune-cell infiltration | Durable antitumor immunity in people |
| Mouse tumor experiments | Tumor-volume trajectories after direct intratumoral exposure | Human response, recurrence prevention, metastatic control, or systemic safety |
What did structure–activity optimization add?
The CKS1-derived library identified useful design relationships, but the process was lead optimization—not treatment validation. Researchers changed charge, hydrophobicity, proline placement, and helical organization, then compared predicted structures with cytotoxicity.
Several findings were more nuanced than a simple “more charge is better” rule:
- replacing a negatively charged residue with a positively charged residue improved activity in one comparison;
- removing positive charge reduced activity, but greater overall charge density did not directly track with greater potency;
- hydrophobicity correlated with activity across the tested variants;
- proline-induced bends had context-dependent effects;
- NF42 and NF43 lost activity despite small sequence changes and favorable predictions from several machine-learning tools;
- non-amidated NF27 was less potent than the amidated study material.
The initial peptide-library screen used material reported above 80% purity and, because of resource limits, was not repeated like most later assays. Individually synthesized follow-up peptides were reported above 90% purity. That progression is reasonable for discovery work, but early screening uncertainty should remain visible.
Did NF27 rupture membranes and kill cancer cells?
Yes, NF27 rapidly disrupted membranes and reduced viability in the tested cell systems. In standard LDH assays, 10,000 cells per well were exposed to serial peptide concentrations for two hours. Dose–response curves were fitted with a four-parameter model to estimate EC50 values. A separate alamarBlue assay followed two-hour exposure, washing, and three hours of viability readout.
In live 4T1 breast-cancer cells treated with 20 micromolar NF27, calcein fluorescence fell rapidly at about 10 minutes and was nearly absent by about 20 minutes, consistent with membrane leakage. The study also reported loss of mitochondrial membrane potential and caspase-3 activation six hours after exposure. FITC-labelled NF27 at 10 micromolar localized to the plasma membrane, nuclear envelope, and nucleus, with partial overlap across several organelles rather than one preferred intracellular target.
These experiments support a membrane-disruptive mechanism followed by downstream cell-death signals. They do not show that every cancer cell behaves this way in a tumor, much less in a patient.
How did NF27 compare with LTX-315?
The final supplement directly compared LDH release from 4T1 and CT26 cells. In the plotted representative experiments:
- 4T1: NF27 EC50 was 9.3 micromolar; LTX-315 EC50 was 33 micromolar.
- CT26: NF27 EC50 was 4.3 micromolar; LTX-315 EC50 was 12 micromolar.
Each figure was representative of three experiments, with plotted points representing technical replicates. This is a head-to-head cell-assay comparison under the study conditions—not evidence that NF27 is clinically superior to LTX-315.
How broad was the cancer-cell evidence?
NF27 was active across a broad panel, but “broad” still means the cells and assay conditions actually tested. The targeted cell-line work included murine 4T1, CT26, B16-F10, MCA205, and CT2A cancer cells, and human U2OS, HT-29, HepG2, SUM149, and Panc10.05 cancer cells. Noncancer comparisons included NIH/3T3 fibroblasts, human umbilical-vein endothelial cells, and normal human lung fibroblasts.
The National Cancer Institute screen expanded the panel to 59 evaluable NCI-60 lines. Cells were exposed for 48 hours at 0.01, 0.1, 1, 10, and 100 micromolar, and growth was assessed with the sulforhodamine B assay. The reported GI50 was below 10 micromolar for every evaluable line except LOX IMVI melanoma, which was 26.3 micromolar. SF-539 was excluded for quality-control reasons.
That result demonstrates breadth across established cancer cell lines. It does not reproduce drug delivery, metabolism, immune context, tumor architecture, or the diversity of healthy human tissues.
What did cell size and cell-junction correlations mean?
Larger NCI-60 cells tended to be more sensitive, while stronger expression of several cell-junction genes tended to track with relative resistance. Cell-size data were available for 54 lines; LOX IMVI was treated as an outlier in the correlation analysis. Transcriptomic analysis identified 110 positively correlated and 20 negatively correlated genes at a Benjamini–Hochberg false-discovery rate below 0.05. Cell-junction-related genes and EPCAM protein measures appeared among the resistance-associated signals.
These are correlations. They do not prove that cell size causes sensitivity or that a particular junction protein causes resistance. Established cell lines also differ in many linked ways, so a mechanistic claim would require targeted perturbation experiments.
What did the hemolysis test show?
NF27 caused little red-blood-cell lysis in the reported ex vivo assays, including a final comparison with LTX-315. Human red blood cells were prepared as a 2% suspension and incubated with peptide for one hour at 37 degrees Celsius. The paper defined more than 10% lysis as toxic hemolysis.
In the final supplement, LTX-315 crossed that 10% line as concentration increased and reached much higher lysis at the top of the plotted range. NF27 stayed below 10% across the plotted range, extending to roughly 3,000 micromolar. The earlier preprint comparison used melittin and likewise found substantially more hemolysis with the comparator.
The result lowers concern about direct red-blood-cell lysis under this assay's conditions. It cannot establish systemic safety, organ toxicity, immune toxicity, vascular effects, or tolerability after repeated exposure.
Was NF27-induced cell death immunogenic?
The study found several markers consistent with immunogenic cell death, but it did not prove a durable, tumor-specific immune memory response. NF27-treated 4T1 and CT26 cells released ATP. Conditioned media from cells exposed to 20 micromolar NF27 for two hours increased TNF-alpha and IL-6 production by DC2.4 dendritic cells after 24 hours.
For an in vivo immune readout, established CT26 tumors received one intratumoral administration. Tumors were removed five days later for flow cytometry. The reported groups were vehicle n = 10 and NF27 n = 12. NF27 increased infiltration by T cells, CD8-positive T cells, and NK cells; macrophage infiltration did not significantly change. Two-group comparisons used unpaired two-tailed Welch tests. ATP comparisons across several conditions used Dunnett tests.
Those signals are compatible with local immune activation. The study did not report a tumor rechallenge, depletion experiment proving that a specific immune population caused tumor control, or long-term immune-memory endpoint.
What did intratumoral distribution show?
FITC-labelled NF27 spread locally through CT26 tumor tissue soon after injection, then the fluorescence diminished over time. The final supplement shows tumor sections collected at 1, 4, and 24 hours after direct intratumoral administration. Signal was strongest around the local distribution area at one hour, weaker by four hours, and faint by 24 hours.
This is best described as local tumor distribution, not comprehensive whole-body biodistribution. The experiment did not map intact NF27 across blood and major organs, and FITC signal does not necessarily prove that the peptide remained chemically intact. An ex vivo murine-serum experiment estimated an approximately 30-minute half-life at 37 degrees Celsius, but that is not an in vivo pharmacokinetic study.
What happened in the mouse tumor experiments?
Direct intratumoral NF27 slowed growth in four models and produced complete responses in some CT26 and HT-29 tumors during a short observation period. Female mice were four to eight weeks old; the study used BALB/c, C57BL/6, and BALB/c nude animals across the models. Tumors were established from CT26 murine colorectal, HT-29 human colorectal, B16-F10 murine melanoma, or 4T1 murine breast-cancer cells.
One week after tumor implantation, mice with sufficient tumors were randomized. The efficacy experiments delivered NF27 directly into tumors every two days; controls received the same volume of water. The exact reported group sizes were:
| Model | Control | NF27 | Reported result |
|---|---|---|---|
| CT26 murine colorectal | 10 | 10 | Growth suppressed; paper reported 80% complete eradication |
| HT-29 human colorectal xenograft | 12 | 11 | Growth suppressed; paper reported 60% complete eradication |
| B16-F10 murine melanoma | 10 | 9 | Growth suppressed; no complete-eradication percentage reported |
| 4T1 murine breast cancer | 8 | 8 | More modest growth suppression; no complete-eradication percentage reported |
The study administered 50 microlitres of a 16 mg/mL NF27 solution, reported as 40 mg/kg, directly into each tumor. Those details identify the experiment; they are not treatment instructions. Tumor volume was measured by calipers. Curves were reported as mean plus or minus SEM and analyzed with two-way ANOVA, with the figure marking p below 0.005.
Individual curves in the final supplement run to approximately day 16 after treatment began. For CT26, the reported 80% maps to 8 of 10 treated mice. For HT-29, the paper reports 60% while the treated group is n = 11; that rounded percentage does not disclose a clean exact denominator. It should not be silently converted into a precise fraction.
Most importantly, the paper did not report a long-term recurrence study, survival endpoint, metastatic-disease test, or tumor rechallenge. “Eradication” here means no measurable tumor on the plotted curve during the reported observation—not proof of cure.
Were the experiments randomized or blinded?
The animal allocation was randomized, but comprehensive blinding was not reported. The study-design section says sample sizes were determined by power analysis in R and that mice with sufficient tumors were randomized into treatment or control groups while concealed in a cloth-covered cage. It says no data or samples were excluded and that animal experiments were repeated at least twice.
The paper does not report that the people administering treatment, measuring tumors, analyzing images, or assessing outcomes were blinded to group assignment. In vitro assays were generally run in triplicate and repeated at least three times, except the resource-limited initial CKS1-derived screen. Reporting allocation randomization is useful; it is not the same as blinded outcome assessment.
Did the study establish systemic safety?
No. The paper reported no observable systemic toxicity, but the available experiments cannot support a broad systemic-safety conclusion. The safety-related evidence consisted chiefly of selected noncancer-cell assays, ex vivo hemolysis, local intratumoral exposure, an ex vivo serum-degradation experiment, and short mouse observations.
The final supplement does not present a comprehensive repeat-dose toxicology package with hematology, clinical chemistry, organ histopathology, dose-ranging, systemic exposure, delayed toxicity, or large safety groups. Direct tumor injection also concentrates exposure locally and does not model intravenous or otherwise systemic administration. Small groups and short follow-up are poorly suited to detecting uncommon or delayed harm.
The correct conclusion is narrow: no clear toxicity signal was reported under these particular preclinical conditions. That is not evidence that NF27 is systemically safe in animals or humans.
Why mouse tumor eradication is not human cancer efficacy
Four translation gaps remain especially large:
- Route: NF27 was placed directly into small established tumors. Human tumors are larger, more heterogeneous, and often inaccessible or metastatic.
- Model: CT26 and 4T1 are rapidly growing murine systems; HT-29 is a human-cell xenograft in immunodeficient mice. Each captures only part of human cancer biology.
- Duration: roughly two weeks of plotted response cannot establish recurrence-free control or survival.
- Safety: limited local-exposure observations cannot establish systemic toxicology or a therapeutic window in people.
The model dependence is visible inside the study itself. CT26 responded far better than 4T1. The authors proposed that tumor stiffness and intratumoral distribution could contribute: 4T1 tumors were harder to inject and may distribute material less evenly. That explanation is plausible, but it was not directly proven.
For a broader guide to this distinction, see What Preclinical Actually Means. A separate mouse cancer-vaccine example, the SaPAC evidence review, shows why immune markers, tumor-volume curves, and human benefit must remain separate endpoints.
What changed from the 2025 preprint?
The project progressed from a bioRxiv preprint posted in August 2025 to peer-reviewed publication in October 2026. The preprint used the more outcome-forward title “Oncolytic peptide NF27 effectively inhibits tumor growth and eradicates tumors in multiple cancer types.” The final Journal of Medicinal Chemistry title shifted emphasis toward discovery and structure–activity relationships.
Peer review and the final analytical records make the evidence easier to audit, but they do not move the work out of the preclinical category. The final supplement added or clarified direct LTX-315 comparisons, local distribution imaging, individual tumor trajectories, and analytical certificates. The peer-reviewed abstract still describes NF27 as a promising candidate—not an approved or clinically effective therapy.
What conflicts of interest should readers know?
Several authors have direct intellectual-property and company interests in the technology. Natsuki Furukawa, Adam C. Mirando, Niranjan B. Pandey, and Aleksander S. Popel disclosed that they are inventors on a pending US patent application, “Oncolytic peptides,” submitted by Johns Hopkins University and covering the peptides described in the work. The same four authors disclosed that they are co-founders of Terebra Therapeutics.
Eric S. Christenson also disclosed consulting relationships with Seres Therapeutics, SirTex, Urogen, and Roche, plus research collaborations with Haystack, Pfizer, Affirmed, NextCure, and Regeneron. The relevant response is not to discard the study automatically, but to favor preregistered follow-up, independent replication, and transparent reporting of negative results.
Bottom line
NF27 is an interesting preclinical oncolytic-peptide lead, not a cancer treatment. The study provides a coherent chain from CKS1-derived optimization through membrane rupture, broad cell-line activity, relative selectivity, immune signals, local tumor distribution, and mouse tumor responses. It also exposes the central limits: direct intratumoral delivery, model-dependent effects, small groups, brief follow-up, incomplete systemic toxicology, and no human data.
“Some mouse tumors became unmeasurable” is the accurate exciting result. “NF27 cures cancer” would be fiction wearing a lab coat.
Frequently asked questions
What is NF27?
NF27 is a 23-residue experimental oncolytic peptide derived through structure–activity optimization of CKS1. It has been studied in cell assays and mouse tumor models, not in humans.
Did NF27 cure cancer in mice?
The paper reported complete eradication of measurable tumors in 80% of treated CT26 mice and 60% of the HT-29 group during a roughly 16-day plotted period. It did not report long-term recurrence testing, survival, or rechallenge, so “cure” is not established.
Has NF27 been tested in people?
No human trial was reported in the peer-reviewed paper.
Is NF27 approved for cancer treatment?
No. NF27 is not approved, and the paper states that no oncolytic peptide has been clinically approved for cancer treatment.
Was NF27 better than LTX-315?
NF27 had lower EC50 values than LTX-315 in representative 4T1 and CT26 LDH assays and caused less hemolysis across the plotted ex vivo range. That does not establish clinical superiority.
Did the paper prove NF27 is safe?
No. It found no clear toxicity signal under limited preclinical conditions. Formal systemic toxicology and human safety remain unestablished.
Sources
- Furukawa N, Chao AR, Yang W, et al. “Discovery of a Highly Potent and Selective Oncolytic Peptide NF27 through Structure−Activity Relationship Studies.” Journal of Medicinal Chemistry. Published online October 23, 2026. doi:10.1021/acs.jmedchem.6c02442.
- Furukawa N, et al. Final ACS Supporting Information, including Figures S1–S23 and Tables S1–S4.
- Furukawa N, et al. “Oncolytic peptide NF27 effectively inhibits tumor growth and eradicates tumors in multiple cancer types.” bioRxiv. Posted August 22, 2025. doi:10.1101/2025.08.17.668569.
- National Cancer Institute Developmental Therapeutics Program. NCI-60 Human Tumor Cell Lines Screen.