A Six-Protein Serum Profile Taken Before Infusion Separated Melanoma Patients Who Benefited From Unconditioned TIL Therapy

Seventeen patients, no conditioning regimen

TILT-123, generic name igrelimogene litadenorepvec, is an adenovirus armed with tumor necrosis factor and interleukin-2. The design used it in place of the lymphodepleting chemotherapy and high-dose interleukin-2 that adoptive TIL therapy conventionally requires, on the reasoning that conditioning toxicity limits who can be treated at all. Median age was 67, range 25 to 75, with nine women and eight men across cutaneous, mucosal and uveal subtypes. By RECIST 1.1 the disease control rate among evaluable patients was 35.3% at day 36 and 37.5% at day 78, with objective responses in 2 of 16 at day 78. PET-based disease control ran higher, 62.5% at day 36 and 46.6% at day 78. Median overall survival was 360 days in patients who never achieved disease control and was not reached in those who did, a separation the authors report at p=0.0154.

Progressors already differed in serum on the day of infusion

The day 36 serum draw was taken immediately before TIL infusion. Patients who went on to progress already had significantly higher IL-6, hepatocyte growth factor, VEGF-A, IL-8, CCL17 and CCL14. Two of those, CCL17 and HGF, were elevated at baseline before any treatment. Circulating monocytic myeloid-derived suppressor cells followed the same pattern at day 36 (p=0.0168), and the six-protein panel stayed elevated through days 50 and 64. Baseline HGF above the cohort median predicted shorter overall survival within TUNINTIL alone (p=0.0322), while EGF, IL-8, VEGF-A, CCL3 and CCL4 did not reach significance in a cohort of 17.

Pooling three trials put the cut-off at 229.2 pg/mL

To get past that sample size the authors pooled baseline serum data from TUNIMO, TUNINTIL and PROTA, reaching 58 patients. Receiver operating characteristic analysis put baseline EGF at an area under the curve of 0.721 with an optimal cut-off of 229.2 pg/mL, against 0.553 for HGF. Combining the two added essentially nothing, at 0.722. The EGF threshold classified 44 of 58 patients correctly, 19 of 23 with disease control and 25 of 35 without, for 75.9% accuracy, 82.6% sensitivity and 71.4% specificity. Applied across trials, elevated baseline HGF and EGF both predicted shorter overall survival (p=0.0028 and p=0.0288). Restricting TUNINTIL to the nine patients below the EGF cut-off shifted the objective response rate from 11.7% to 22.2% and the disease control rate from 35.2% to 77.8%.

CD8 and CD4 rose in uninjected metastases too

Serial tumor sampling was read by multiplex immunofluorescence quantified in Indica Labs HALO software through Oracle Bio, alongside hematoxylin and eosin and adenovirus E1A immunohistochemistry. By day 64, intratumoral CD8 and CD4 T cell proportions were significantly raised over baseline in both injected and non-injected metastases (p=0.0342 and p=0.0343). FoxP3-positive regulatory T cells stayed constrained in injected lesions and PD-L1 expression in tumor was stable across both sites. Patient 10,107, a complete responder, supplied the illustrative series. Viral DNA, quantified by PCR against the IRES-IL2 region, was detectable in tumor in only a subset of patients.

Limits the authors put on their own numbers

Their stated constraints are the small cohort, a mixed population spanning three melanoma subtypes with varying prior treatment, reliance on blood sampling that cannot capture spatial structure in the tumor, and no correction for multiple testing across exploratory endpoints. Two observations carry the authors' own speculative label: a complete response in a patient with von Willebrand disease, and better survival in patients whose baseline activated partial thromboplastin time sat at or above the median, which they read as a possible low-platelet-activation phenotype. Heavily pretreated patients carried more terminally differentiated CD8 T cells at baseline and lived less long (p=0.0344).

Why This Matters to the APO|APE Reader

The 229.2 pg/mL threshold was derived and evaluated on the same 58 patients with no held-out cohort, so 75.9% accuracy is an in-sample figure that will move on an independent cohort. The authors also note that serum EGF is largely platelet-derived during activation and clotting, which turns tube type, clotting interval and centrifugation into part of the measurement rather than background handling detail. Reproducing this cut-off means pinning down those preanalytics and porting the readout from a research NPX panel to a validated immunoassay before a single number in pg/mL can gate who starts an intratumoral dosing course.