Patients were enrolled to one of the three open-label, single arm, phase 2 trials. All patients were treated with T cell products manufactured in the same production facility at the Sheba Medical Center, as previously described.2 9 30 Patients signed an informed consent approved by the Israeli Ministry of Health.

Patients with stage IV melanoma were treated with TIL ACT as salvage therapy after failure of multiple treatments. The ‘120Cy/125Flu’ cohort (NCT00287131) received NMA conditioning regimen with cyclophosphamide (60 mg/kg for 2 days; days −7 and −6) and fludarabine (25 mg/m² for 5 days; days −5 to −1) before cell infusion (day 0) and the ‘60Cy/125Flu’ cohort cyclophosphamide (30 mg/kg for 2 days; days −5 and −4) and fludarabine (25 mg/m² for 5 days; days −5 to −1) (NCT03166397). The ‘TBI/75Flu’ cohort received TBI (200 cGy, on day −4) and fludarabine (25 mg/m² for 3 days; days −3 to −1) (NCT03166397). TILs were administered intravenously, followed by high-dose bolus IL-2 with 720,000 IU/kg/dose, maximum of three times a day, for 5 days or up to tolerance. Patients received filgrastim for 7–10 days starting 1 day after cell infusion. Response was assessed using the Response Evaluation Criteria in Solid Tumors (RECIST V.1.1) guidelines 4 weeks following TIL administration and every 3 months thereafter or as clinically needed. Objective responders (ORs) were defined as complete and partial responders and non-responders (NRs) as patients with stable disease or progressive disease by determining the best overall response using RECIST.

Relapsed or refractory (r/r) pediatric and adult patients with ALL were treated with anti-CD19 CAR-T cells (NCT02772198). Lymphodepleting conditioning was induced by fludarabine 25 mg/m² for 3 days (days −2 to −4) and cyclophosphamide 900 mg/m² (which equals approximately 30 mg/kg Cy) for 1 day (day −2) (cohort ‘30Cy/75Flu’), followed by infusion of 1×10⁶ transduced CAR-T cells per kilogram weight.9 31 Patients did not receive IL-2. For clinical evaluation of response, patients with ALL underwent a bone marrow aspiration 28 and 60 days following CAR-T cell administration, and disease assessment was performed by morphology, flow cytometry and PCR minimal residual disease based on immunoglobulin or T-cell receptor rearrangement, with a lower level of detection of 5–10 leukemia cells in 10,000–100,000 healthy cells.

Absolute neutrophil, platelet and lymphocyte counts were obtained before NMA initiation and up to 3 months after cell infusion. The following inflammatory indexes were calculated: NLR, defined as neutrophil-to-lymphocyte ratio; PLR, platelet-to-lymphocyte ratio; and systemic immune-inflammatory index (SII), defined as absolute neutrophil count×absolute platelet count/absolute lymphocyte count. All cell counts were performed in a centralized, automated laboratory at the Sheba Medical Center.

Patients’ serum was collected on day 0, before cell infusion. IL-7, IL-15 and IL-21 serum levels were determined by ELISA (IL-7, Human IL-7 Quantikine HS ELISA kit, R&D Systems, Minneapolis, Minnesota, USA; IL-15 and IL-21, ELISA MAX Deluxe Set, BioLegend, San Diego, California, USA) according to the manufacturer’s instructions. Measurements were performed in duplicates.

Significance of variation between groups was evaluated using a parametric two-tailed Student’s t-test. The differences between proportions were tested using two-sided Fisher’s exact test. Analysis of covariance was performed to examine differences in the mean values of patients’ characteristics on variable production parameters. Cell counts and inflammatory indexes were transformed to a logarithmic scale. Linear mixed models with Tukey’s adjustment were used to compare absolute cell counts and inflammatory indexes between responder groups or the different regimens. Eastern Cooperative Oncology Group (ECOG) performance status was modeled as fixed effect. Effect sizes were calculated as Cohen’s d. Linear mixed models were fit using the R statistical software (www.r-project.org). The receiver-operating characteristics (ROC) curve analysis was used to define sensitivity and specificity through the Youden’s index, and the area under the ROC curve (AUC) to find the optimal cut-off values for NLR, PLR and SII using the MedCalc V.18.11.6 (MedCalc Software). Survival curves were plotted by the Kaplan-Meier method, and differences between groups were assessed by the log-rank test using the Prism GraphPad V.8.0.1 (GraphPad Software).

Between 2005 and 2016, 103 patients with metastatic melanoma were enrolled to a phase 2 ACT trial with TIL following preconditioning with an accumulating dose of 120 mg/kg cyclophosphamide and 125 mg/m² fludarabine2 (120Cy/125Flu cohort; table 1). To analyze the impact of NMA on bone marrow suppression and recovery, absolute cell counts were collected, starting a few days before NMA initiation (baseline) and up to 3 months post therapy. Due to cyclophosphamide-induced toxicity (detailed later), a follow-up study was initiated. The cyclophosphamide dose was halved (60Cy/125Flu) or cyclophosphamide was replaced with 200 cGy TBI (table 1). In 2017 and 2018, nine patients with metastatic melanoma were enrolled to the TBI/75Flu group, and in 2019 and 2020, eight patients received 60Cy/125Flu-induced NMA. Patients in the 60Cy/125Flu cohort and those in the TBI/75Flu cohort did not respond to TIL therapy. The TIL manufacturing procedure was identical in all three cohorts. TILs were manufactured at the same Good Manufacturing Practice (GMP) facility and by the same staff.

The three melanoma cohorts were well matched for most baseline characteristics (eg, ECOG performance status) but were significantly different with respect to prior treatments (table 2). Most patients in the 120Cy/125Flu cohort were treated in the pre-immune checkpoint inhibitor (pre-ICI) era, whereas patients in the 60Cy/125Flu and TBI/75Flu cohorts were all treated after treatment failure with ICIs.

Simultaneously, CD19 CAR-T therapy was conducted at our clinical center and both TIL and CAR-T were produced at the same GMP cell production facility. Between 2016 and 2019, patients with r/r ALL were treated with CD19 CAR-T cells following NMA with fludarabine and cyclophosphamide 900 mg/m² (which equals approximately 30 mg/kg Cy) (‘30Cy/75Flu’ cohort). Absolute cell counts were acquired from 19 patients, including 13 complete responders and 6 NRs. The patients received, on average, 4.0±1.2 prior treatments and 2 of 19 (11%) patients had brain metastasis. The 30Cy/125Flu ALL cohort included pediatric patients, thus the average age was low in this cohort. Since patients in the 30Cy/125Flu cohort had a different underlying disease and were treated with CAR-T cells instead of TIL, the results of this cohort will be demonstrated separately.

To evaluate the impact of the different preconditioning regimens on bone marrow suppression, absolute cell counts were monitored in the three melanoma groups (figure 1) and the ALL cohort (discussed later). Cell counts were performed in a centralized, automated laboratory. The number of patients with available cell counts per day is shown in online supplemental table 1. The day of cell infusion was defined as ‘day 0’.

One hundred and one of 103 (98%) patients in the 120Cy/125Flu and 100% (8 of 8) patients in the 60Cy/125Flu cohort achieved severe neutropenia, defined as absolute neutrophil count (ANC)≤0.5 K/µL. The lowest values of ANC were found 3 days after cell infusion in the 120Cy/125Flu group (0.065±0.230 K/µL; median 0.025 K/µL) and 4 days after cell infusion in the 60Cy/125Flu cohort (0.050±0.044 K/µL; median 0.030 K/µL) (p=0.871) (figure 1A, table 3 and online supplemental figure 1). TBI/75Flu preconditioning did not result in severe neutropenia and the lowest value was 1.088±0.541 K/µL (median 1.100 K/µL) 1 month after cell infusion. In contrast, five of nine (56%) patients in the TBI/75Flu cohort demonstrated even a sharp increase of ANC, with values between 13.03 and 16.13 K/µL 1 day after cell infusion when the normal range for neutrophils is between 1.8 and 7.7 K/µL (figure 1A).

Similar results were observed for ALC (figure 1B and table 3). 120Cy/125Flu and 60Cy/125Flu preconditioning were equally effective in achieving deep lymphodepletion, with lowest average values of 0.029±0.036 K/µL (median 0.020 K/µL) and 0.024±0.017 K/µL (median 0.025 K/µL) on the day of cell infusion in the 120Cy/125Flu and 60Cy/125Flu cohorts, respectively (p=0.389), when normal values are between 1.0 and 4.8 K/µL. A direct comparison of the ALC in the 120Cy/125Flu and 60Cy/125Flu cohorts over time is shown in figure 1D. Administration of cyclophosphamide in the 120Cy/125Flu cohort was initiated on day −7 and in 60Cy/125Flu cohort on day −5, which explains the difference of ALC on days −4 and −3. In the TBI/Flu cohort, only a minor decrease in ALC was achieved (lowest value: 0.062±0.046 K/µL; median 0.050 K/µL; on day +1) (figure 1B). Lymphodepletion was significantly less efficient compared with the 120Cy/125Flu and 60Cy/125Flu cohorts (p values≤0.01; figure 1E and F).

Platelet counts at baseline were normal in the three melanoma cohorts (120Cy/125Flu cohort, 281±96 K/µL; 60Cy/125Flu cohort, 341±121 K/µL; TBI/75Flu cohort, 299±82 K/µL). In the 120Cy/125Flu and 60Cy/125Flu cohorts, an APC decrease was observed following NMA initiation; however, in the 120Cy/125Flu cohort, values were significantly lower (120Cy/125Flu, lowest value 34±26 K/µL, median 26 K/µL; 60Cy/125Flu cohort, lowest value 115±51 K/µL, median 103 K/µL; p≤0.001) (figure 1C and online supplemental figure 1D–F).

In conclusion, 120Cy/125Flu and 60Cy/125Flu preconditioning were equally efficient in achieving deep lymphopenia and neutropenia, which was not the case for TBI/75Flu NMA.

Both in the 120Cy/125Flu and 60Cy/125Flu cohorts, the patients recovered by day 6 and day 7 from their severe neutropenia, and counts within the norm (1.8–7.7 K/µL) were measured (120Cy/125Flu cohort, 3.63±5.22 K/µL, median 1.54 K/µL; 60Cy/125Flu cohort, 3.01±2.03 K/µL, median 2.73 K/µL) (figure 1A).

Interestingly, although TBI/75Flu-induced NMA caused a sharp increase of ANC 1 day after cell infusion followed by mild neutropenia, neutrophil recovery was delayed and baseline levels were only reached after 1 month.

Recovery of ALC was observed in all the cohorts 4–7 days after cell infusion and returned to the normal range of values (figure 1B). The duration of severe lymphopenia was 7 days in the 120Cy/125Flu and 60Cy/125Flu cohorts, but only 2 days in the TBI/75Flu cohort.

Platelet recovery was found 12–14 days after cell infusion in the 120Cy/125Flu cohort. In the TBI/75Flu cohort, APC steadily decreased over time and 50% (three out of six) of the patients had prolonged thrombocytopenia around 3 weeks after cell infusion (figure 1C).

To evaluate if bone marrow suppression and recovery can serve as predictive markers of clinical response, we analyzed the cell counts in the 120Cy/125Flu melanoma patient cohort (ORs, N=29; NRs, N=74). The baseline characteristics of responders and non-responders were previously reported and multivariate logistic regression revealed that ECOG performance status was an independent factor of response in this cohort.2

Responding and non-responding patients had similar ANC, ALC and APC at baseline (p=0.953, p=0.497 and p=0.343) (figure 2A–C). Following NMA initiation, there was no significant difference between responders and non-responders regarding ANC (d=0.083, p=0.508) (figure 2A) and APC (d=0.007, p=0.972) (figure 2C). However, ALCs were significantly different (d=0.605, p=≤0.001) (figure 2B). From day 4 onwards, OR patients had significantly higher lymphocyte counts, with the highest significant difference between days 5 and 21 (p values≤0.001). Absolute cell counts were further compared between ORs (N=29) and NRs (N=29), which were matched for baseline characteristics, including ECOG and prior treatments (online supplemental figure 2A–C), as well as the entire cohort, excluding patients with ECOG 2 and prior anti-PD1 therapy (OR, N=28; NR, N=53) (online supplemental figure 2D–F) and demonstrated the same results.

It was shown that the combination of absolute cell counts, defined as inflammatory indexes, can be used as a prognostic factor for response and survival in different types of cancer and may be more significant than individual counts. ORs in the 120Cy/125Flu melanoma cohort had significantly lower PLR, NLR and SII from day 5 onwards (figure 2D).

ROC curves were used to obtain the optimal inflammatory indexes cut-offs on day 7, which distinguish between ORs and NRs. The optimal cut-off for NLR was 1.79, for PLR 32.7 and for SII 368 K/µL.

We found that 61% (17 of 28) of OR patients compared with 24% (16 of 66) of NR patients had an NLR ≤1.79 (p=0.002) and 61% (17 of 28) of the responders compared with 15% of the NRs had a PLR ≤32.7 (p≤0.001). SII on day 7 could not be used to distinguish between OR and NR patients.

Kaplan-Meier analysis indicated that lower values of NLR and PLR were also significantly associated with prolonged overall survival (OS) in the 120Cy/125Flu cohort. As shown in figure 3, patients with NLR≤1.79 had a median OS of 23 months compared with 13.5 months for patients with NLR>1.79 months (p=0.011); patients with PLR≤32.7 had a median OS of 39 months compared with 12.5 months for patients with PLR>32.7 (p=≤0.001).

To determine the impact of the different lymphodepleting regimens on cytokine levels, ELISA assays were performed with serum samples collected after completion of NMA and before cell infusion (day 0). Serum was available from patients with melanoma in the 120Cy/125Flu cohort (N=24, including 12 ORs and 12 NRs), 60Cy/125Flu cohort (N=7) and TBI/75Flu cohort (N=8).

IL-7 serum levels were significantly lower following TBI/75Flu-induced lymphodepletion compared with 120Cy/125Flu NMA (TBI/75Flu cohort, 12.0±7.0 pg/mL; 120Cy/125Flu cohort, 19.0±5.7 pg/mL; p=0.008) (figure 4A). IL-15 serum levels were insignificantly different in all three cohorts (figure 4B). Patients in the 60Cy/125Flu cohort had significantly higher IL-21 serum levels (240±116 pg/mL) than patients from the TBI/75Flu cohort (122±101 pg/mL; p=0.048) and 120Cy/125Flu cohort (124±71 pg/mL; p=0.002) (figure 4C).

No significant differences between IL-7 (p=0.252), IL-15 (p=0.902) and IL-21 (p=0.140) serum levels were observed between ORs (N=12) and NRs (N=12) in the 120Cy/125Flu cohort (figure 4D).

The 30Cy/75Flu cohort differed in many aspects from the other three cohorts. Patients in the 30Cy/75Flu group suffered from ALL (instead of metastatic melanoma), were treated with CAR-T (instead of TIL), did not receive IL-2 administration following infusion (instead of high-dose bolus IL-2) and received prior lines of chemotherapies, which is uncommon for patients with metastatic melanoma. Therefore, this population is discussed here separately.

30Cy/75Flu preconditioning in the ALL cohort did not yield deep neutropenia (online supplemental figure 3A). The lowest ANC value was 0.835±0.966 K/µL (median 0.537 K/µL) 9–11 days after cell infusion. In addition, only a minor decrease in ALC was achieved (lowest values: 0.164±0.167 K/µL; median 0.367, on day +1) (online supplemental figure 3B). Platelet counts were already low at baseline (119±105 K/µL; median 95 K/µL) (online supplemental figure 3C). Baseline thrombocytopenia might be a result of leukemic infiltration of the bone marrow or extensive prior chemotherapy treatments. APCs were low, but unchanged from baseline and up to 3 months post therapy.

Comparing OR (N=13) and NR (N=6) patients of the 30Cy/75Flu cohort, there was no significant difference in ANC (OR vs NR; d=1.380, p=0.053) and ALC (d=−0.246, p=0.678) (online supplemental figure 2D–E). APCs were significantly lower in non-responding ALL patients 1 day after NMA initiation and throughout the entire time of monitoring (OR vs NR; d=3.910, p=0.001) (online supplemental figure 3F). In both groups, APC was not affected by 30Cy/75Flu lymphodepletion. Thus, the lower platelet counts in non-responding ALL patients seem disease dependent, rather than NMA related, which is in accordance with the existing literature.33

Patients stayed, on average, 23±8 days in the hospital and received 1.0±1.4 units of RC and 1.3±2.4 units of PLT. Febrile neutropenia was observed in 12 of 19 (63%) patients.