Duvelisib with Docetaxel for Patients with Anti–PD-1 Refractory, Recurrent, or Metastatic Head and Neck Squamous Cell Carcinoma

Despite aggressive multimodality treatment for locoregionally advanced head and neck squamous cell carcinoma (HNSCC), many patients experience persistent, recurrent, or metastatic (R/M) disease (1). Given that several studies have demonstrated antitumor activity with immunotherapy, treatment for R/M disease often includes the use of immune checkpoint blockade targeting PD-1 (refs. 2–4). KEYNOTE-048 established the combination of platinum and 5-fluorouracil chemotherapy and the PD-1 inhibitor pembrolizumab as a first-line standard regimen in R/M HNSCC. Furthermore, pembrolizumab monotherapy is approved for patients with tumor ligand of PD-1 (PD-L1) combined positive scores (CPS) of 1 or greater (5). In 4-year follow-up, these regimens continue to demonstrate a survival benefit versus cetuximab-based chemotherapy (6). However, an open question remains as to the preferred second-line (2L) treatment option in the R/M setting after patients experience progression on PD-1 inhibitor therapy.

Outside of a clinical trial, regimens incorporating EGFR-targeted therapy with cetuximab, platinum, and/or taxane-based chemotherapy are all options for 2L treatment and beyond. In fact, our own data and others suggest that these agents may result in improved response rates and clinical benefit after immunotherapy exposure (7–9). Taxane therapy remains a long-term standard therapy in both locoregionally advanced and R/M HNSCC (10–14), with trials establishing their favorable safety profile as compared with regimens integrating 5-fluorouracil in the advanced disease setting (15). Notably, the development of taxane resistance has been an important area of study in HNSCC and other cancers.

The PI3K signaling cascade is often activated in HNSCC and has been nominated as a pathway that facilitates the development of taxane resistance in solid tumors. Preclinical models suggest that PI3K pathway inhibition may enhance taxane antitumor activity (16, 17). BERIL-1 was a randomized, multicenter phase 2 trial comparing the combination of the oral pan-PI3K inhibitor buparlisib with weekly paclitaxel versus the taxane alone regimen in R/M HNSCC that demonstrated a manageable safety profile and an improvement in median progression-free survival (PFS; 4.6 vs. 3.5 months, P = 0.01; ref. 18). The BURAN study is a phase 3 trial investigating the overall survival (OS) benefit of the same combination compared with taxane alone (NCT04338399) and recently completed accrual. Importantly, the BERIL-1 trial completed accrual in 2015 before the approval of immunotherapy for HNSCC, with all patients receiving prior platinum-based chemotherapy. The benefit of adding a PI3K inhibitor to taxane therapy in the contemporary post-immunotherapy era has not yet been explored.

Building on these observations, we performed a phase 2 clinical trial exploring the safety and antitumor activity of the selective oral PI3K δ/γ dual isoform specific inhibitor duvelisib with docetaxel in patients with anti–PD-1 refractory R/M HNSCC. Duvelisib targets the two (δ/γ) isoforms expressed in leukocytes aimed at modulating innate and adaptive immune cell functions.

The study protocol (see Supplementary Appendix 1) was approved by the Institutional Review Board at Dana-Farber Cancer Institute (DF/HCC# 21-393), and the trial is registered with ClinicalTrials.gov (NCT05057247). The trial was conducted in accordance with the principles of the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines. All patients provided written informed consent. Study funding and duvelisib were provided by Secura Bio, Inc.

Patients 18 years or older were eligible if they had recurrent, incurable, or metastatic HNSCC, with at least one site of measurable disease by RECIST v1.1 (19), an Eastern Cooperative Oncology Group performance status of 0 to 1, and adequate organ function. Any primary mucosal subsite was permitted, including oral cavity, oropharynx, larynx, and hypopharynx. Patients must have received at least one, but no more than two, prior line of systemic therapy for R/M disease with one line including a PD-1 inhibitor (platinum-based therapy as part of definitive or adjuvant treatment counted as one prior line if the subject progressed within 6 months). Prior paclitaxel therapy was permitted if used in first-line treatment as part of chemoimmunotherapy or in the definitive setting, but receipt of a prior PI3K pathway inhibitor was exclusionary. Human papillomavirus (HPV) testing was required for patients with oropharyngeal tumors, and pretreatment tumor PD-L1 CPS was recorded. Patients were permitted to enroll regardless of tumor PI3K mutational status.

This single-arm, single-center, nonrandomized, open-label phase 2 clinical trial enrolled patients at the Dana-Farber Cancer Institute (Boston, Massachusetts). Following informed consent, patients received a 7-day lead-in of duvelisib at 25 mg orally twice daily followed by the initiation of the same dose and schedule of duvelisib with docetaxel at 75 mg/m² intravenously every 21 days until the occurrence of disease progression, unacceptable toxicity, withdrawal of consent, or up to 2 years. Docetaxel dosing every 3 weeks was chosen for ease of administration over a weekly schedule (14, 15). Duvelisib adherence was documented by drug diary. Prespecified dose adjustments were permitted: up to one dose of reduced level of duvelisib (15 mg twice daily) and two doses of reduced levels of docetaxel (56 and 37 mg/m²). Study medications interrupted for more than 6 weeks consecutively could not be reintroduced. Patients who discontinued one of the study drugs for any reason besides disease progression were allowed to continue the other at the discretion of the treating investigator.

Patients received standard premedications prior to taxane therapy that included dexamethasone and posttreatment G-CSF support after each cycle of chemotherapy. Patients were maintained on concomitant Pneumocystis jirovecii pneumonia (PJP) prophylaxis while on duvelisib, and antiviral prophylaxis was strongly recommended for those with a history of herpesvirus infection, including Cytomegalovirus. Concurrent medications were monitored given known CYP3A4 interactions with duvelisib.

In addition to routine laboratory assessments, plasma Cytomegalovirus viral load was monitored with each treatment cycle. Patients underwent imaging assessments (contrast-enhanced CT or MRI of the neck, with CT chest, abdomen, pelvis, or PET scan) at baseline and every 9 weeks or 3 cycles while on treatment. Following treatment discontinuation, patients were followed every 2 to 3 months up to 3 years from trial registration for resolution of toxicity and to document survival.

Adverse events (AE) to assess the safety and side effect profile of treatment were recorded using Common Terminology Criteria for AEs (CTCAE) version 5.0. Any grade 4 AE (except for short-term electrolyte disturbances or cytopenia) and confirmed PJP were considered unacceptable toxicity warranting study drug discontinuation. AEs were captured up to 30 days after discontinuation of study drugs.

The primary endpoint was best overall response rate (ORR) as determined by RECIST 1.1. A two-stage Simon design was utilized. When >4 of 26 patients who were eligible and began protocol treatment had disease in response (assuming >1 patient with disease in response among the first 13 patients), there was 89.2% power to rule out a 10% ORR and detect a 30% response rate (using a one-sided exact binomial test, type I error rate of 9.8%). A response rate of 30% was targeted based on published experience with the combination of a pan-PI3K inhibitor and paclitaxel (18) and the desired threshold for response in this disease setting. A safety run-in was performed with the first three patients enrolled to ensure tolerability of the regimen. Secondary endpoints included evaluation of the safety and tolerability of the combination, PFS and OS, and duration of therapeutic response.

The primary efficacy population included all eligible patients who began protocol treatment. Response rate was summarized as a proportion with a corresponding two-stage 95% confidence interval (CI; to account for accrual continuing into the second stage). The distribution of PFS and OS was estimated using the Kaplan–Meier method. Logistic regression and Cox proportional hazard models were used to estimate ORs for best overall response and HRs for survival, respectively. Subgroups were compared using Fisher exact test (categorical variables) and Wilcoxon rank-sum test (continuous variables). Two-sided P values are reported. Data as of June 13, 2024, were analyzed using R (v4.4.0) and GraphPad Prism v10.

Exploratory objectives aimed to correlate clinicopathologic, tumor genomic, and immunologic profiling data (namely, pretreatment PI3K pathway alteration status, PD-L1 CPS, and T-cell immune parameters) with response and outcomes. A prior study showed that TP53 alterations, HPV-negative status, low tumor mutational load, or high infiltration of CD8⁺ T cells or tumor infiltrating lymphocytes derived benefit with the combination of buparlisib and paclitaxel (20). Formalin-fixed, paraffin-embedded tumor samples obtained before treatment were submitted for targeted next-generation sequencing with either FoundationOneCDx or Caris Molecular Intelligence profile. PD-L1 CPS was assessed by means of IHC staining with the use of either the 22C3 or E13LN clones. We performed multiplex immunofluorescence (MxIF) on pretreatment formalin-fixed, paraffin-embedded whole-slide tumor specimens using Leica Bond Rx Automated Stainer (Leica Biosystems). We used hematoxylin and eosin–stained sections, and board-certified pathologists identified regions of interest (ROI), including the invasive tumor front on stained sections. To delineate immune and tumor architectures, the MxIF panel included 4′,6-diamidino-2-phenylindole (DAPI; nuclear stain), pan-cytokeratin (AE1/AE3 epithelial tumor marker), CD3e, CD8, CD20, and CD68. For detection, we used Akoya’s opal polymer anti-mouse and anti-rabbit HRP with multiple opal fluorophores (NEL821001KT, Akoya Biosciences); for antibody clones and fluorophore assignments, see Supplementary Table S1. We performed image acquisition at 20× using the Vectra Polaris Automated Imaging System (PerkinElmer). Data analysis was performed using proprietary tools developed by BostonGene (21). Briefly, we used the MxIF raw full-channel image files, applied signal thresholding to eliminate background noise, and overlapped ROIs. After exclusion of artifacts, we segmented individual and classified and counted cells based on their marker status. We used raw cell counts for downstream statistical analyses.

Requests for data are limited to those data sets described in this article, and data are not publicly available due to patient privacy concerns. Deidentified participant data generated from this clinical trial are available upon reasonable request from the corresponding author up to 6 years following publication. Raw tumor genomic sequencing data were not available to the sponsor, as testing relied on commercial assay reporting, but the deidentified reports received are available from the corresponding author upon reasonable request.

Between November 2021 and September 2023, 28 patients enrolled to the study. Two patients failed screening, one because of elevated liver function tests (LFT) and one because of cytopenia, both related to the disease. All remaining 26 patients began protocol treatment and are included in safety and efficacy analyses.

Median age was 64 years (range, 34–74), with most identifying as male at birth (25, 96%), and many were former or current smokers (18, 69%). Oropharynx (12, 46%) and oral cavity (11, 42%) were the most common primary disease subsites; 14 (54%) patients had tumors causally related to HPV (Table 1; Supplementary Table S2). No patient had a history of prior taxane or EGFR antibody exposure, but most patients had >1 line of prior systemic therapy for advanced disease (18, 69%).

Best ORR was 19.2% (95% CI, 6.8–40.7), with five partial responses (PR) but no complete responses (Fig. 1A–C). An additional 12 (46%) patients achieved stable disease, whereas 8 (31%) experienced progression of disease (PD), including two with clinical PD before first restaging. One patient was unevaluable, having withdrawn consent after the lead-in phase of duvelisib and before first restaging. Response rates were similar in HPV-positive vs. -negative patients (4/14 vs. 1/12; P = 0.33). Median time to response was 2.3 months (range, 1.4–2.9), and median duration of response was 5.1 months (range, 0.7–15.5 months), with one responder remaining on-treatment at the last follow-up. Another patient maintained a PR >1 year after discontinuing treatment for toxicity. The primary reason for treatment discontinuation (two remained on-treatment at the last follow-up) was PD (10, 38%), with six (25%) experiencing unacceptable toxicity (Supplementary Table S3). Patients received a median number of 2.5 cycles of treatment (range, 1–16) and remained on study treatment a median of 2.3 months (range, 0.7–21.9). The primary reason for coming off study among 21 patients (five remain on study) was death (17, 81%), whereas three (14%) patients withdrew consent from study participation and one (5%) was lost to follow-up. No independent clinical or molecular features significantly predicted response to therapy (Supplementary Table S4). Furthermore, ORR was not different regardless of whether the patient received prior anti–PD-1 therapy alone versus in combination or sequentially with platinum-based therapy in the R/M setting (18% vs. 22%, respectively; P = 1.00).

Among all patients, 18 (69%) experienced a grade 3 to 5 AE regardless of treatment attribution. Table 2 describes the treatment-related AEs (TRAE) experienced by >10% of the study cohort or any treatment-related grade 3 to 5 events. Six (23%) patients had toxicity as the reason for study drug discontinuation (most attributable to acute liver injury or elevated LFTs), but no treatment-related deaths occurred. The most common TRAEs included anemia (11, 42%), elevated LFTs (10, 38% each), hyponatremia (10, 38%), maculopapular rash (9, 35%), and fatigue (9, 35%). The most frequent grade 3 TRAE was elevated LFTs (either aspartate or alanine aminotransferases; 7, 27%), while grade 4 TRAEs included elevated LFTs (4, 15%), neutropenia, hypercalcemia, and gastric bleeding (each 1, 4%). Two (8%) patients experienced grade 3 treatment-related diarrhea, with one suspected to have delayed recrudescence of immune-mediated colitis in the setting of prior PD-1 blockade.

At a median follow-up of 6.5 months (range, 0.7–26+), the median PFS was 2.8 months (95% CI, 1.9–7.0), with 17 (65%) patients experiencing a progression event, including 5 with clinical or radiologic evidence of PD, and two patients who died without evidence of progression (one from cardiopulmonary arrest attributed to anemia from gastrointestinal bleeding, and the other from aspiration pneumonia in a patient with laryngeal cancer; Fig. 2A). The 6- and 12-month estimated PFS was 31.4% (95% CI, 13–51.9) and 18.9% (95% CI, 5.1–39.4), respectively. Median OS was 10.2 months (95% CI, 6.7–15.9) for the cohort with 17 (65%) patients having died at the last follow-up (Fig. 2B). The 6- and 12-month estimated OS was 75.3% (95% CI, 52.9–88.1) and 45.2% (95% CI, 23.7–64.5), respectively. No independent clinical or molecular features significantly predicted OS (Supplementary Table S5). Of note, median PFS and OS were numerically longer for HPV-negative patients in the present study compared with those with HPV-positive disease: 7 (95% CI, 1.3–not available) versus 2.4 months (95% CI, 1.8–5.8), and 15.5 (95% CI, 2.4–not available) versus 7.5 months (95% CI, 5.8–12.9), respectively.

Median pretreatment PD-L1 CPS was 18 (range, 0–100; available for n = 24), and median scores were not significantly different between responders and nonresponders (15 vs. 20; P = 1.00; Fig. 3A and B) and regardless of prior platinum exposure in the R/M setting vs. anti–PD-1 therapy alone (15 vs. 20, P = 0.83). Furthermore, median PFS was numerically higher among those patients with lower PD-L1 CPS (<20) compared with those with higher scores of ≥20 (5.8 vs. 2.8 months). Median pretreatment tumor mutational burden (available for n = 21) was 4 mutations/Mb (range, 1–21) which did not differ significantly based on response (3 vs. 4.8 mutations/Mb; P = 0.16; Fig. 3C). Among 25 sequenced patients, 7 (27%) harbored PIK3CA mutations or amplifications (4 with E545K/E542K hotspot alterations, 1 amplified). A mutation in PI3K signaling (mTOR/Akt) was observed in a single patient (4%), and another patient (4%) demonstrated PTEN loss—both had a best ORR of stable disease.

MxIF profiling results were available for a subset of patients (n = 18, 69%) based on pretreatment tissue yield and tumor purity on biopsy (Supplementary Table S6). Median cells in the ROI was 121,364 (range, 1,496–640,839), with samples having a median hematoxylin and eosin tumor purity of 50% (range, 20–70). The median percentage of CD3⁺/CD8⁺ T cells in the tumor microenvironment (TME) across the cohort was 0.61% (range, 0–2.58; Fig. 3D–F). The median infiltration of CD68⁺ macrophages or monocytes was 1.81% (range, 0.09–9.63), and that of B cells was 0.86% (range, 0–17.86). Supplementary Table S7 correlates pretreatment immune markers in the TME with PFS and OS.

The results of this phase 2 study further demonstrate the antitumor activity of a PI3K inhibitor when combined with a taxane in R/M HNSCC and is the first trial to our knowledge specifically investigating the combination after anti–PD-1 exposure. With 5 of 26 responses observed, our null hypothesis was rejected. Whereas the ORR of 19% observed in the present study was lower than that reported for the buparlisib and paclitaxel combination (ORR 39%; ref. 18), it is important to recognize that these study populations were not identical, and drugs targeting different isoforms within the same therapeutic class may demonstrate variable efficacy. When compared with the population enrolled in BERIL-1, our study cohort was smaller and comprised of very few women, included mostly oral cavity and oropharyngeal primaries, had a larger number of HPV-positive cases (54% vs. 22%), and all participants in this heavily pretreated population received prior anti–PD-1 therapy.

The four isoforms of PI3K class I have unique tissue distributions and physiologic functions which impact efficacy and safety profiles (22–24). Duvelisib is a selective dual inhibitor of the γ/δ PI3K isoforms targeting immune cell functions, such that its efficacy was first established in refractory indolent non–Hodgkin lymphoma (ref. 25), as compared with pan-PI3K inhibitors. Despite the lack of α/β isoform targeting aimed at limiting the dysregulation of insulin metabolism, duvelisib carries its own risks of serious infection, diarrhea or colitis, and pneumonitis, which impacted its approval status in non–Hodgkin lymphoma. The PIK3CA gene encodes the p110α catalytic subunit of the membrane protein, which is often mutated in cancer, leading to enhanced kinase activity and cell growth (26). Furthermore, PIK3CA is among the most frequently altered genes in HNSCC (27) with a higher rate of helical (such as E542 and E545) and kinase (H1047) domain mutations observed in HPV-associated oropharyngeal cancers (28). Buparlisib targets α and other PI3K class I isoforms which would be expected to target tumors with PIK3CA gene mutations or amplifications. However, duvelisib was chosen in the current trial recognizing its immunomodulatory properties and knowing that all enrolled patients would have received prior PD-1 directed therapy to promote T-cell activation in the TME. Docetaxel was chosen in the current trial given its comparable efficacy to paclitaxel across advanced HNSCC studies (29–31) and favorable toxicity profile—with more frequent hypersensitivity reactions and neurotoxicity cited with paclitaxel use (32).

Though the combination of a taxane with a δ/γ PI3K inhibitor demonstrates clinical efficacy, there are important toxicity concerns to recognize. Duvelisib and docetaxel resulted in >30% rates of anemia, elevated LFTs (without bilirubinemia), and hyponatremia, along with fatigue, rash, and diarrhea attributable to treatment—noting the 21-day docetaxel dose and schedule investigated in the present study. Furthermore, PJP prophylaxis was required on study and could have contributed to drug-induced liver injury rates. Notably, rates of grade 3+ AEs were comparable across both BERIL-1 and the present study (82% vs. 69%, respectively), but buparlisib was more often associated with hyperglycemia and cytopenias while administered in a weekly dosing schedule (18). More than 20% of our cohort discontinued one or both agents for toxicity, often related to grade 3 to 4 acute liver injury, which resolved in all cases. Despite prior anti–PD-1 exposure and the use of a δ/γ PI3K isoform inhibitor, few suspected episodes of delayed onset immune-mediated toxicity were observed. Interestingly, a patient with a PR stopped the study drug combination for toxicity (acute liver injury, fatigue, and diarrhea) but maintained a prolonged and ongoing response a year out from drug discontinuation. During hospitalization, they underwent a skin biopsy showing psoriasiform dermatitis with dermal perivascular lymphocytic infiltrate, most consistent with an immune reaction due to drug. While speculative, this transient multiorgan inflammatory reaction may have impacted their durable response to therapy—which has been observed in other settings (33).

One of the more important findings from our study was the median OS of 10.2 months, with nearly half of patients alive at 1-year follow-up, favoring those with HPV-negative disease. This finding is noteworthy in this heavily pretreated, anti–PD-1 refractory population with R/M HNSCC. Whereas 50% of patients in our study had HPV-associated disease, the BERIL-1 study reported a similar median OS of 10.4 months in the 2L setting (18). These survival outcomes are favorable even in the face of recent data showing improved response rates to cytotoxic chemotherapy after progression on immunotherapy in R/M HNSCC with a median OS of around 8 months (9). However, selection bias is important to consider as patients entering the trial were fit for additional therapy. Whereas the response rate to duvelisib and docetaxel may be modest with tempered durability, the immunomodulatory properties of a selective PI3K inhibitor with taxane may in some way impact longer-term survival like what has been observed with immune checkpoint blockade in which PFS is limited (2–5). The OS signal observed in our study compares favorably when considering other emerging therapies in this disease setting—some of which are targeting epidermal and/or hepatocyte growth factor receptors (34) or antibody–drug conjugates aimed at various surface target proteins. Though these newer agents have their own toxicity concerns in a pretreated population, grade 3 or greater AEs or discontinuation rates <20% are preferred.

Key to the aims of the study was evaluating the impact of molecular and immunologic parameters on response and outcomes. Higher tumor PD-L1 CPS is predictive of a favorable response to PD-1 blockade and impacts survival, as shown in several R/M HNSCC trials incorporating immunotherapy (5, 6, 35). In the present study, PD-L1 CPS was primarily recorded from tissue obtained prior to the start of any systemic therapy for R/M HNSCC, but there was no difference in scores regardless of duvelisib and docetaxel response or prior R/M therapy. It may be that testing between lines of therapy exposure would have proven more informative, as the TME can be dynamic. Among sequenced patients on study, 25% harbored PIK3CA mutations or amplifications most often in hotspot domains, though the frequency of these alterations should not impact the mechanism of a δ/γ inhibitor such as duvelisib. The BERIL-1 study only showed a trend toward improved ORR in the PI3Kα-activated subgroup recognizing the small number of mutationally selected patients in the study, whereas both of our patients with alterations in mTOR/Akt or PTEN loss showed disease stability. In BERIL-1, patients with HPV-negative tumors and those with TP53 alterations derived a survival benefit from pan-PI3K and taxane therapy. This may be explained by disruptions in p53 that can lead to downstream PI3K pathway activation conferring sensitivity to PI3K inhibition (36). Whereas pretreatment TME immune biomarkers were exploratory in the present study and limited to a subset of patients, for every unit increase in CD3⁺/CD8⁺ T cells, the risk of death decreased by 2.2%—a predictor of benefit from immune-based therapies that has been observed across many solid tumor trials (37).

The current study lacked gender and racial diversity given the modest sample population, recognizing the restraints of single-arm (without a comparator), single-center trials and the epidemiology of the disease under study. The use of a pan-PI3K or PI3Kα-specific inhibitor may have offered improved on-target effects. Furthermore, our correlative analyses to assess predictors of outcomes were limited by subset sample size and testing multiple comparisons. Despite these limitations, our findings strengthen the observed antitumor activity and survival signal for combined PI3K inhibitor and taxane therapy in advanced HNSCC—even in patients with tumors refractory to PD-1 blockade—while recognizing the importance of monitoring safety with the combination. Despite the interest in molecularly selecting patients with PI3K-activated tumors, these and other data suggest that the cytotoxic and immunomodulatory effects of the PI3K and taxane combination may provide clinical benefit across a broader, unselected HNSCC population. We await the results of the completed registrational trial (BURAN study) comparing buparlisib–paclitaxel to paclitaxel with placebo in a similar HNSCC population.

G.J. Hanna reports other support from Secura Bio during the conduct of the study as well as other support from ImmunityBio, KSQ, Kura Oncology, Replimune, and Bristol Myers Squibb; personal fees and other support from Merck, Coherus, Astellas, and Bicara; grants from Gateway for Cancer Research; personal fees from PDS Biotech, Pyxis, Nextech, OncoSwitch, Surface Oncology, Tubulis AstraZeneca, Boxer Capital, Grey Wolf, and Inhibrx; and personal fees and nonfinancial support from Naveris, Inc., outside the submitted work. K. Sehgal reports grants and other support from Merck and other support from Exelexis, Inc., Scholar Rock, Inc., Guidepoint Global, Equinox Group Inc., AmeriSource Bergen, Pri-Med, BinayTara Foundation, MedScape Oncology, and Pfizer outside the submitted work. V. Kushnarev reports personal fees from BostonGene during the conduct of the study. A. Tkachuk reports personal fees from BostonGene during the conduct of the study. J. Lennerz reports other support from BostonGene outside the submitted work. R.I. Haddad reports personal fees from Merck, EMD Serono, Genmab, Boehringer Ingelheim, AstraZeneca, GSK, RPT, and PDS; other support from Hookipa and Nanobiotix; and grants from GSK, AstraZeneca, IO Biotech, AVEO, EMD Serono, and Merck outside the submitted work. No disclosures were reported by the other authors.

G.J. Hanna: Conceptualization, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing. L.B. Oakley: Resources, data curation, investigation, methodology, writing–review and editing. R. Shi: Conceptualization, data curation, software, formal analysis, investigation, methodology, writing–review and editing. A. ONeill: Conceptualization, data curation, software, formal analysis, investigation, methodology, writing–review and editing. K.Y. Shin: Data curation, software, formal analysis, validation, investigation, methodology, writing–review and editing. N. Scarfo: Investigation, methodology, writing–review and editing. K. Sehgal: Data curation, investigation, writing–review and editing. M.J. Dennis: Data curation, investigation, writing–review and editing. N. Quinn: Data curation, investigation, writing–review and editing. V.Y. Jo: Data curation, investigation, writing–review and editing. K. Wong: Data curation, investigation, writing–review and editing. A. Shvyrkova: Data curation, formal analysis, investigation, project administration, writing–review and editing. V. Kushnarev: Data curation, software, investigation, project administration, writing–review and editing. B.U. Shanthappa: Data curation, formal analysis, validation, investigation, methodology. A. Tkachuk: Data curation, validation, investigation, writing–review and editing. K. Kryukov: Software, formal analysis, validation, investigation, methodology. A. Sarachakov: Data curation, software, visualization, methodology. V. Svekolkin: Software, formal analysis, visualization, methodology. J. Lennerz: Resources, software, supervision, investigation, methodology, writing–review and editing. S. Waters: Resources, supervision, funding acquisition, project administration, writing–review and editing. R.I. Haddad: Conceptualization, resources, data curation, supervision, investigation, methodology, writing–review and editing.

This research received support and funding from Secura Bio, Inc. The authors would like to extend their gratitude to the participating patients and their families for their contributions to this important research.