High Programmed Death-Ligand 1 Expression is Associated With a Shorter Progression Free Survival in ALK-Rearranged Lung Cancer Patients Treated With First Line Tyrosine Kinase Inhibitors Authors: Yunan Nie, Alyse Staley, Trista K. Hinz, Rocio Perez-Johnston, Daniel T. Merrick, Michael C. Yang, Kurtis D. Davies, Paul A. Bunn, Dara L. Aisner, D Ross Camidge, Erin L. Schenk, Lynn E. Heasley, Tejas Patil Published: 2026 ABSTRACT In 130 patients with ALK-rearranged NSCLC treated with ALK TKIs, high PD-L1 expression was independently associated with shorter progression-free survival to first line ALK TKIs. PD-L1 overexpression did not alter ALK TKI sensitivity in preclinical ALK models, suggesting that tumor microenvironment–driven immune suppression, rather than tumor-intrinsic PD-L1, underlies inferior clinical outcomes.Background:Although ALK-positive non–small cell lung cancer (NSCLC) derives limited benefit from immune checkpoint inhibitors, PD-L1 expression is frequently elevated at diagnosis. The clinical relevance of PD-L1 expression for the efficacy and durability of response to ALK tyrosine kinase inhibitors (TKIs) remains unclear. We investigated the association between tumor PD-L1 expression and outcomes with first-line ALK TKIs and used complementary preclinical models to explore potential biological mechanisms.Methods:Clinical, pathologic, molecular, and treatment outcomes were evaluated among patients with ALK-rearranged NSCLC treated at the University of Colorado. PD-L1 expression was assessed by tumor proportion score (TPS). Murine EML4–ALK lung cancer cell lines were engineered to overexpress PD-L1, and effects on ALK TKI response were evaluated in vitro and in immune-competent mouse models.Results:Among 130 TKI-naïve patients treated with first-line ALK TKIs, 57 (44%) had high PD-L1 expression (TPS ≥ 50%). After multivariable adjustment, high PD-L1 expression was associated with significantly shorter progression-free survival (PFS) compared with low PD-L1 expression (11 vs. 21 months; HR 2.29, 95% CI, 1.51–3.49; P ≤ .001), with no difference in overall survival (91 vs. 107 months; HR 1.16, 95% CI, 0.62–2.17). Objective response rates were similar between PD-L1–high and –low tumors (81.8% vs. 83.3%; Pearson r = −0.20, 95% CI, −0.40–0.07). In preclinical EML4–ALK models, tumor cell PD-L1 overexpression did not alter alectinib sensitivity, depth, or durability of response in vitro or in vivoConclusions:High PD-L1 expression is associated with shorter PFS to first-line ALK TKIs without affecting the initial radiographic response. Preclinical findings suggest that PD-L1–intrinsic tumor cell signaling does not directly impair ALK TKI efficacy, implicating tumor microenvironment–mediated mechanisms in reduced response durability. Introduction Biomarker selection in non-small cell lung cancer (NSCLC) plays an important role in guiding the selection of anticancer therapies. The treatment landscape for NSCLC has been reshaped by 2 key developments: the identification of oncogene drivers susceptible to small molecule tyrosine kinase inhibitors (TKIs) and the integration of immune checkpoint inhibitors (ICIs) into standard clinical practice. First, the recognition of specific oncogene drivers—such as ALK gene rearrangements—has enabled the use of TKIs resulting in both improved progression free survival (PFS) and overall survival (OS).1–10 Second, the use of ICIs targeting the PD-1/PD-L1 pathway is now standard of care for patients without a targetable oncogene driver, with several trials showing a replicable OS benefit.11–13 High PD-L1 expression can be found in a subset of ALK-rearranged NSCLC, possibly as a mechanism of tumor immune evasion, although the implications and underlying biology are not fully understood.14,15 Translocations within chromosome 2 lead to the formation of an ALK gene fusion, translation of a chimeric oncoprotein composed of the C-terminal kinase domain of ALK joined with various nonkinase fusion partners (typically EML4 in NSCLC), and subsequent constitutive activation of downstream signaling pathways which includes pathways associated with endogenous PD-L1 expression.10,14 However, multiple studies have shown almost no clinical efficacy with single agent ICIs in ALK-rearranged NSCLC, irrespective of PD-L1 expression. For example, the Phase 2 ATLANTIC trial (NCT02087423) evaluating durvalumab in refractory metastatic NSCLC with progression on at least 2 lines of therapy demonstrated no confirmed disease responses in the 11 evaluable patients with ALK-rearranged NSCLC.16 Furthermore, a literature review of 73 patients with ALK-rearranged NSCLC who received ICIs reported an objective response rate (ORR) of 8%.15 Given the uniformly poor responses to immune checkpoint inhibitors in ALK-rearranged NSCLC, it remains unclear why PD-L1 expression is frequently elevated, despite no evidence that it reflects effective suppression of cytotoxic T cell activity.15,16 Increased PD-L1 expression may be due to multiple factors, including a tumor microenvironment (TME) that is characterized by lack of infiltrating CD8 + T cells and complex interplay of constituents such as PD-L1 expressing tumor-associated macrophages (TAMs) and induced FOXP3 + T regulatory cells.17 In addition, alternate checkpoints may contribute to tumor evasion; for example, ALK-rearranged NSCLC appear to have increased expression of TIM3 relative to EGFR mutated and KRAS mutated tumors.18 Multiple studies using patient derived samples and murine models suggest that T cell TME infiltration occurs early after TKI initiation and positively contributes to the TKI response.19–26 EML4-ALK lung adenocarcinoma cell lines induced PD-L1 expression through HIF-1a and STAT3 activities under hypoxia, but STAT3 alone under normoxia, suggesting that EML4-ALK might differentially regulate PD-L1 expression depending on a normoxic versus hypoxic tumor microenvironment.27 Additional studies have suggested that the ALK gene fusions induce PD-L1 upregulation through activation of the JAK-STAT3-PD-L1 signaling pathway, which may create an immunosuppressive TME and facilitate immune evasion.27,28 These data highlight the complex biology and interplay between oncogenic ALK fusions and PD-L1 expression. To our knowledge, a systematic clinical analysis with a mechanistic description of the relationship between PD-L1 expression and first line ALK TKI efficacy has not been described previously. Here, we provide clinical data and results from in vitro and in vivo studies exploring the relationship between PD-L1 expression and the outcomes of first line ALK TKIs. Materials and Methods Patients and Clinical Data Collection The study was conducted in accordance with the Declaration of Helsinki and good clinical practice guidelines following approved institutional review board (IRB) policies. All patients with metastatic TKI-naïve ALK-rearranged NSCLC (AJCC 8th edition) seen at the University of Colorado between January 2015 and January 2022 who received a first line ALK TKI (alectinib, brigatinib, or crizotinib) were eligible (Supplemental Figure S1). Demographics, pathology, molecular testing, imaging reports, treatment history, and clinical outcomes were extracted from the electronic health record. All patients had a baseline brain MRI prior to receiving an ALK TKI. Blinded independent radiology review to capture radiographic responses via RECIST 1.1 was performed using the patient’s first on-treatment scan obtained within 8–12 weeks after starting an ALK TKI. Clinical outcome measures such as progression free survival (PFS), time to next treatment (TTNT), and overall survival (OS) were collected (Supplemental Figure S2). Exclusion criteria included: (1) nonmetastatic disease (2) absence of PD-L1 expression score prior to starting an ALK TKI, (3) incomplete follow up. Molecular Profiling Methods Molecular assessment for an ALK rearrangement was done through a combination of fluorescence in situ hybridization (FISH) and next generation sequencing (NGS) based approaches. For FISH analyses, paraffin sections were hybridized with fluorescent break apart probes (Abbott Laboratories) with ≥ 15% of cells with split signals out of 50 cells scored constituting positivity. Tissue samples were analyzed using NGS-based mutational analysis, including both Archer VariantPlex Solid Tumor and Archer FusionPlex Solid Tumor (ArcherDx, Inc.). Bioinformatics analysis was carried out using Archer Analysis software package. PD-L1 tumor proportion scores (TPS) were performed on all patient samples at diagnosis using a combination of the PD-L1 22C3 PharmDx assay (Agilent Technologies) or Ventana SP263 assays (Roche Diagnostics) using established industry thresholds for quantification.29 Statistical Analyses Demographic, clinical, and treatment characteristics are summarized and compared using paired t-tests (continuous variables) and Fisher exact test (for categorical variables). Unadjusted survival curves and corresponding medians were obtained with Kaplan-Meier methods. Cox Proportional Hazard models were used to assess the univariate associations between each survival outcome and the following predefined variables: PD-L1 TPS score (continuous variable), TKI (alectinib, brigatinib, crizotinib), brain metastases (yes, no), tumor burden (defined as total number of metastatic organ systems involved), and prior chemotherapy exposure (yes, no). Martingale residuals were used to select the best functional form. Schoenfeld and deviance residuals were further used to assess model fit. Hazard ratios, confidence intervals, corresponding P-values were obtained using robust standard errors. Statistical analyses were performed using GraphPad Prism software (version 10.2.3 for Windows, GraphPad Software, La Jolla, CA) and R Studio (version 4.4.0 for Windows, R Project, Vienna, Austria). Analysis of Pretreatment Patient Samples Clinicopathologic correlation was performed using an IRB-approved protocol (COMIRB # 09–0143). H&E-stained sections of 19 pretreatment FFPE specimens from TKI-naïve patients with ALK-rearranged NSCLC were obtained. Two pathologists independently quantified lymphocytes and PMNs; discordant cases were resolved by joint review to reach consensus, and consensus values were used for analysis. Quantified lymphocytes and PMNs were correlated with PD-L1 expression, as determined through clinical IHC assays described above. Lentiviral-Mediated Transduction of Mouse CD274 (PD-L1) Into Murine EML4-ALK Cell Lines The EA1 cell line was developed at the University of Colorado Anschutz Medical Center [26] and the EA2 cell line was generously provided by Dr. Andrea Ventura at Memorial Sloan Kettering Cancer Center. Both cell lines were cultured in Roswell Park Memorial Institute-1640 (RPMI, Corning) growth media supplemented with 5% fetal bovine serum (FBS, GIBCO) and 1% penicillin–streptomycin (Corning) in a humidified incubator with 5% CO2 at 37°C. To stably express CD274/PD-L1, prepackaged pLTC lentiviral particles encoding full-length murine PD-L1 were purchased from G&P Biosciences (Santa Clara, CA; catalog number LTP-1868) and infection of EA1 and EA2 cells was achieved according to the manufacturer’s instructions. Stably transduced cells were selected for puromycin resistance, and the resulting cultures were submitted to flow cytometry after staining with an antimouse CD274/PD-L1 antibody (Biolegend, San Diego, CA; cat. #124,331). Flow cytometry revealed low and high-expressing PD-L1 populations which were collected and cultured separately. Cell Proliferation Assays Cell Proliferation Cell lines were seeded at 25,000 cells per well of 24-well plates using complete growth medium. On subsequent days 1 through 4, the cells in triplicate wells were suspended with trypsin and counted using a Cellometer Auto T4 (Nexcelom Bioscience, Waltham, MA). Clonogenic TKI Dose Response Assays Cell lines were plated at 200 cells per well in 96-well tissue culture plates. 24 hours later, cells were treated in triplicate with concentrations of alectinib or crizotinib from 0 to 300 nM. Cell number was assessed by DNA content determined after 7–10 days of culture using the CyQUANT Direct Cell Proliferation Assay (Invitrogen, Carlsbad, CA) and a Biotek Synergy 2 fluorescence microplate reader (Agilent, Santa Clara, CA). Data are presented as percent of DMSO control-treated wells. Quantitative RT-PCR Assays Cells were collected in 600 μL RNA Lysis Buffer and total RNA was purified from cells using Quick-RNA MiniPrep kits (Zymo Research, Irvine, CA) and aliquots (5 μg) were reverse transcribed in a volume of 20 μL using Maxima First Strand cDNA Synthesis Kit (Thermo Scientific, Pittsburgh, PA). Aliquots (2 μL) of a 1:5 dilution of the reverse transcription reactions were submitted to quantitative RT-PCR in 10 μL reactions with SYBR Select Master Mix for CFX (Thermo Fisher Scientific) using a CFX Connect Real-Time PCR Detection System (BioRad, Hercules, CA). The following PCR primer pairs were used: mouse PD-L1 (Fwd: 5′-AGCAAGTGATTCAGTTTGTGGC-3′, Rev: 5′-CCTTCAAAAGCTGGTCCTTTGG-3′); mouse EML4-ALK (EML4-forward, 5′-GAGCCTTGTTGATACATCGTTC-3′, ALK-reverse, 5′-CAAGGCAGTGAGAACCTGAA-3′). GAPDH mRNA levels were measured as a housekeeper gene for normalization. Orthotopic Mouse Model of EML4-ALK Lung Cancer C57BL/6 J (#000664) mice were obtained from Jackson Laboratory (Bar Harbor, ME). All procedures and manipulations were performed under protocols approved by the Institutional Animal Care and Use Committee at the University of Colorado Anschutz Medical Campus. Control and PD-L1-expressing (high levels) EA1 and EA2 cells were injected into the left lobe of the lungs of mice. Cells were prepared in a solution of 1.35 mg/mL Matrigel (Corning #354234) diluted in Hank’s Balanced Salt Solution (Corning) for injection. Mice were anesthetized with isoflurane, the left side of the mouse was shaved, and a 1 mm incision was made to visualize the ribs and left lobe of the lung. Using a 30-gauge needle, 5 × 105 cells were injected in 40 μL of matrigel cell mix directly into the left lobe of the lung and the incision was closed with staples. Tumors were permitted to establish for 7–10 days and then the mice were submitted to micro-CT (μCT) imaging to obtain pretreatment tumor volumes. Tumor-bearing mice were randomized into treatment groups (n = 10), either 20 mg/kg alectinib or diluent control (H2O) by oral gavage 5 days/week. To monitor effects of drug treatment on tumor volume, mice were imaged weekly by μCT imaging by the University of Colorado Anschutz Medical Small-Animal IGRT Core using the Precision X-Ray X-Rad 225Cx Micro IGRT and SmART Systems (Precision X-Ray, Madison, CT). Tumor volume was quantified from μCT images using ITK-SNAP software36 (www.itksnap.org). Results Patient Characteristics Between 2015 and 2023, a total of 130 TKI-naïve patients with ALK-rearranged NSCLC treated with first line ALK TKIs were identified (Table 1). Patients were separated into a PD-L1 high cohort (TPS ≥ 50%, n = 57) and PD-L1 low cohort (TPS < 50%, n = 73). The cutoff of 50% was validated by cut point analysis. The median age at diagnosis in the PD-L1 high and PD-L1 low cohorts was 52 years (range, 41–60 years) and 48 years (range, 41–58 years) respectively. Most patients in both cohorts were never smokers ([PD-L1 high, n = 49, [86%]; PD-L1 low, n = 54, [74%]). Adenocarcinoma was the dominant histology (PD-L1 high, n = 54, [94.7%]; PD-L1 low, n = 68, [93.2%]). A third of patients in both cohorts had baseline brain metastases (PD-L1 high, n = 20, [35%]; PD-L1 low, n = 27, [37%]). The distribution of first line ALK TKIs (crizotinib, alectinib, brigatinib), incidence of TP53 alterations, and ALK 3a/b variants were balanced between cohorts (Table 1). Clinical Outcomes Median duration of follow up was 48 months for both cohorts. Univariate survival curves for PFS and OS are shown in Figure 1. After multivariable adjustment for TKI (alectinib, brigatinib, crizotinib), brain metastases, tumor burden, and prior chemotherapy exposure, the PFS for the PD-L1 high cohort was significantly worse ([PD-L1 high vs. low]; 11 months vs. 21 months, HR 2.29, 95% CI, 1.51–3.49, P ≤ .001). However, there was no difference in OS between the 2 cohorts ([PD-L1 high vs. low]; 91 vs. 107 months, HR 1.16, 95% CI, 0.62–2.17, P = .60). Given the CNS liability of crizotinib, we performed a subset analysis of patients who only received first line alectinib. In this subset analysis (n = 94), ALK-rearranged patients with high PD-L1 expression (≥50%) at diagnosis also demonstrated a shorter PFS on alectinib ([PD-L1 high vs. low]; 12 vs. 25 months, HR 0.52, 95% CI, 0.31–0.86, P = .008) with no difference in OS ([PD-L1 high vs. low]; NE vs. 58 months; HR 1.17, 95% CI, 0.52, P = .70) (Supplemental Figure S3A and B). To address the possibility that subsequent lines of therapy could have influenced OS, we explored time to next treatment (TTNT) for patients who progressed on first line ALK TKIs. Data regarding subsequent line therapies was available for 93 patients (Supplemental Table 1). A significantly higher percentage of patients in the PD-L1 low cohort received subsequent alectinib ([PD-L1 low, 10/47, [21%]; PD-L1 high, 1/46, [2%], P = .0076). Due to lack of data granularity regarding therapies in the 2nd line and beyond, we used time to next therapy (TTNT) to ascertain clinical benefit treatments immediately after first line ALK TKIs. The TTNT with 2nd line treatment was significantly worse in the PD-L1 high group ([PD-L1 low vs. high; 46 mos vs. 10 mos, HR 0.60, 95% CI, 0.35–0.99, P = .04]; Supplemental Figure S4A). When we focused on patients who exclusively received 2nd line lorlatinib, there was no significant difference between the PD-L1 low or high groups ([PD-L1 low vs. high; NE vs. 9 mos, HR 0.60, 95% CI, 0.25–1.45, P = .27]; Supplemental Figure S4B). Without adjusting for confounders, there was a 10 percent increase in the odds of progression for each incremental 10 percentage point increase in the PD-L1 TPS score (95% CI: 4–16, P < .001; Supplemental Figure S5) When PD-L1 was dichotomized, the hazard of progression was 2.07 times higher for the PD-L1 high cohort versus the PD-L1 low (95% CI, 1.40–3.06, P < .001). A Cox proportional hazards model was conducted to assess the impact of prior chemotherapy, crizotinib use, tumor burden (defined as number of metastatic sites at diagnosis), baseline brain metastases, TP53 mutations, and ALK variant 3a/b status on PFS and OS. The overall model was not statistically significant for either PFS or OS (P > .05). Among 46 patients that had complete NGS profiling, there was no difference in the frequency of TP53 alterations or other co-occurring somatic alterations between the PD-L1 high and low cohorts (Supplemental Figure S6). Among patients with radiographically measurable disease (n = 86), objective response rates were calculated using RECIST v1.1 criteria via blinded independent review. The radiographic responses to first line ALK TKIs are shown in the waterfall plot (Figure 2). The objective response rate (ORR) was 82.5% (n = 71) and the disease control rate (DCR) was 98.8% (n = 85). In the low PD-L1 group, the ORR was 83.3% (n = 35) and the DCR was 100% (n = 42). In the high PD-L1 group, the ORR was 81.8% (n = 36) and the DCR was 97.7% (n = 43). We found no statistically significant correlation between PD-L1 expression and radiographic response rates, although they did trend towards an inverse association (Pearson’s correlation, r = −0.20, 95% CI, − 0.40–0.07, P = .06). In a subset analysis of patients who only received alectinib or brigatinib, the objective response rate was 84.6% (n = 65) and the disease control rate was 99.0% (Supplemental Figure S3). Without adjusting for a specific TKI, there was a significantly higher rate of intracranial progression in the low PD-L1 group compared to the high PD-L1 group (34.2% vs. 8.8%, P < .001). When looking specifically at the patients who had CNS progression after receiving an ALK TKI (n = 30), 52% (10/23) of the PD-L1 low group had received first line crizotinib, which could explain the higher rate of CNS progression in this cohort. A Swimmer plot of patients with CNS progression is shown in Supplemental Figure S7. Given the CNS liability of crizotinib, we repeated this analysis with only patients who received first line alectinib or brigatinib (n = 95). When focusing only on patients who received first line alectinib or brigatinib, there was a significantly higher rate of intracranial progression within the low PD-L1 cohort (30.0% vs. 6.7%, P = .004). The intracranial PFS, however, was significantly longer in the low PD-L1 group ([PD-L1 low vs. PD-L1 high]; 17 months vs. 11 months, HR 1.545, 95% CI, 0.45–5.34, P = .045). Preclinical Assessment of PD-L1 Overexpression on Alectinib Sensitivity in Murine Eml4-Alk Lung Cancer Cell Lines Classically, PD-L1 expression on tumor cells is regulated through interferons released by T cells within an inflamed microenvironment.30 In this setting, blockade of the PD-1/PD-L1 interaction reactivates T cell proliferation and killing of tumor cells. However, other distinct signals as well as genetic events can also transcriptionally upregulate PD-L1 expression. In these cases, elevated PD-L1 is not necessarily an informative biomarker of T cell inflamed tumors or response to immune therapy. In addition to immunosuppressive signaling, the literature demonstrates reverse signaling by PD-L1 driving tumor cell-autonomous functions. Of relevance here, studies by our group using murine models of ALK and EGFR-driven lung cancer have demonstrated an adaptive immunity requirement for durable therapeutic TKI responses.26 To test the aforementioned mechanisms underlying PD-L1 action in tumors, we deployed murine Eml4-Alk lung cancer cell lines to determine if ectopic PD-L1 overexpression in ALK-rearranged tumor cells directly impacts the depth or duration of response to alectinib. Murine EML4-ALK cell line models (EA1, EA2) that yield reproducible in vivo alectinib responses when implanted orthotopically into syngeneic, immune-competent mice (25) were transduced with a lentivirus encoding murine CD274/PD-L1. As a control, cells were transduced with the empty lentivirus. The transduced cell cultures were selected for resistance to puromycin, stained with an anti-PD-L1 antibody, and submitted to flow cytometry. As shown in Supplemental Figure S8A, EA1 and EA2 cells transduced with the mouse PD-L1-encoding lentivirus expressed elevated surface PD-L1 staining relative to the little or no expression in cells transduced with an empty lentivirus (EV). Cell sorting was deployed to collect populations of transduced EA1 and EA2 cells expressing “high” and “low” levels of PD-L1 as indicated (Supplemental Figure S8A). The cell surface levels of PD-L1 measured by flow cytometry associated with increased PD-L1 mRNA levels measured by RT-qPCR (Supplemental Figure S8B). Both EA1 and EA2 EV cultures expressed negligible endogenous PD-L1 protein and mRNA (Supplemental Figure S8B). Finally, EA1 and EA2 cells ectopically expressing murine PD-L1 expressed equivalent levels of EML4-ALK mRNA as the EV cells (Supplemental Figure S8C). EA1 and EA2 EV, mPD-L1 low and mPD-L1 high cells were analyzed for baseline rates of proliferation in vitro. The results in Figure 3A demonstrate that ectopic expression of PD-L1 modestly increased the growth rate of EA1 and EA2 cells relative to the empty vector controls. Effects on proliferation were especially evident at later time points as the cultures became confluent. Measurement of the in vitro sensitivity of the control and PD-L1-expressing EA1 and EA2 cells to alectinib and crizotinib demonstrated that neither high nor low levels of PD-L1 overexpression decreased TKI sensitivity (Figure 3B). In fact, EA1 cells expressing low or high levels of PD-L1 exhibited lower IC50’s for alectinib and crizotinib relative to EA1 EV cells (Figure 3B). The results indicate that ectopic overexpression of PD-L1 in EA1 and EA2 cells exerted modest effects on baseline growth, but did not alter TKI sensitivity measured in vitro. Based on these findings, the impact of reverse signaling by PD-L1 on the EML4-ALK cancer cells appears to be minimal. To test the effect of lentiviral-mediated overexpression of PD-L1 on growth properties and sensitivity to alectinib in immune-competent mice, the EV and “high” PD-L1-expressing EA1 and EA2 cells were inoculated into the left lungs of C57BL/6 mice (fully syngeneic with the cell lines) and orthotopic tumors were allowed to establish. Following μCT to measure pretreatment tumor volumes, the resulting tumor-bearing mice were treated with alectinib (20 mg/kg, 5 days on, 2 days off) or diluent as a control and tumor volume was monitored weekly by μCT. As shown in Figure 4, untreated EA1 and EA2 EV and PD-L1 expressing tumors grew at similar rates. Also, EA1 and EA2 tumors ectopically expressing PD-L1 exhibited shrinkage with alectinib treatment equivalent to the EV control tumors (Figure 4). Importantly, neither the depth nor durability of the responses to alectinib over the 28 day treatment were significantly different. After ~4 weeks of alectinib treatment, the residual EA1 EV and PD-L1-expressing tumors regrew at equal rates upon termination of the therapy (Figure 4). Likewise, EA2 PD-L1-expressing tumors were eliminated following 4 weeks of alectinib treatment similar to EA2 EV tumors as demonstrated by the lack of tumor regrowth following termination of therapy (Figure 4). Thus, the findings demonstrate that ectopic expression of PD-L1 via lentivirus-mediated transfection exerted no significant effect on baseline growth rates or TKI depth and durability of response in an immune-competent mouse model of ALK-rearranged lung cancer. The results from these in vivo studies do not support a mechanism where increased PD-L1 levels on cancer cells directly influences the therapeutic response to ALK-targeting TKIs. Assessment of Pretreatment Neutrophil and Lymphocyte Levels in ALK-Rearranged Patient Samples The findings in Figure 4 reveal that simple overexpression of PD-L1 on murine EML4-ALK lung cancer cells fails to alter the in vitro and in vivo sensitivity to alectinib, indicating that neither T cell suppression nor reverse-signaling by PD-L1 into the tumor cells are dominant mechanisms mediating the shortened PFS observed in patients. As an alternative, we considered the hypothesis that elevated PD-L1 expression may serve as a proxy for an immune-suppressed microenvironment identified by increased content of myeloid derived suppressor cells (MD-SCs). To explore this possibility, lymphocytes and polymorphonuclear neutrophils (PMNs) were quantified in H&E-stained pretreatment biopsies obtained from a subset of the ALK patients (n = 19). The findings in Supplemental Figure S9 reveal that the neutrophil content, but not the lymphocyte content, positively associated with the PD-L1 expression (Spearman r = 0.562, P = .012). Based on this finding, we postulate that a microenvironment associated with high tumor cell PD-L1 expression, but mediated by elevated content of immune-suppressive neutrophil/MDSCs accounts for the reduced durability of response in PD-L1 high ALK-rearranged lung cancers. Discussion This study identifies high PD-L1 expression as a negative prognostic biomarker in TKI-naïve, ALK-rearranged NSCLC. While prior reports have suggested this association, our data demonstrate that high PD-L1 remains an independent predictor of shorter progression-free survival (PFS) after adjustment for tumor burden, TP53 status, and intracranial TKI activity. Notably, this effect was observed in relation to the durability of response rather than initial sensitivity, as PD-L1 status did not correlate with the depth of radiographic response or overall survival (OS). Our study demonstrates several novel findings. We found that high PD-L1 expression was associated with shorter PFS to first line ALK TKIs. This is consistent with prior reports,31–35 but we extend these observations by demonstrating that the association persists even after adjustment for established negative prognostic factors, including tumor burden, prior chemotherapy, brain metastases, or intracranial TKI coverage (crizotinib vs. next-generation TKIs with CNS activity). Higher incidence of TP53 alterations and additional co-occurring somatic alterations were not seen in the PD-L1 high group. We found that there was 10 percent increase in the odds of radiographic progression for each incremental percentage point increase in PD-L1 score. We also found that the initial depth of radiographic response was not influenced by PDL1 expression, suggesting that the shortened PFS in low PDL1 expression group cannot be explained fully by blunted initial sensitivity to the ALK TKI. We observed a higher cumulative rate of CNS progression in the low PD-L1 group. Prior studies have shown a high concordance between PD-L1 expression in intracranial and extracranial metastatic sites and high PD-L1 expression is associated with higher rates of synchronous brain metastases.36,37 However, we suspect that our counterintuitive finding likely reflects a "time-at-risk" phenomenon rather than an inherent biological predisposition for brain metastases in low PD-L1 tumors. Because patients with low PD-L1 expression experience longer systemic PFS on first-line ALK TKIs, the cumulative duration of risk for the development of intracranial metastases is higher than patients with high PD-L1 expression. This is supported by the longer intracranial PFS observed in the low PD-L1 group despite the higher cumulative incidence of CNS events. Conversely, patients in the high PD-L1 cohort may progress systemically and transition to subsequent therapies before CNS progression is documented. We found a significant difference in PFS without a significant difference in OS. This finding warrants further validation with large-scale datasets and prospective studies. Although detailed analysis of subsequent systemic therapies was not performed in the context of this study, most patients subsequently received lorlatinib after progression. When focusing on the patients who exclusively received lorlatinib after progression on a first line ALK TKI, we found no difference in TTNT between the high PD-L1 expressors and low PD-L1 expressors. A prior retrospective series found that high PD-L1 expression (≥50%) was associated with the ALK variant 3a/b phenotype and co-occurring somatic alterations such as TP53 and CDKN2A mutations.38 Notably, variant 3a/b has been associated with development of resistance mutations such as G1202R.39 A recent analysis of patients with ALK-rearranged NSCLC treated with first line lorlatinib found no difference in median PFS when separated by underlying variant status40 An intriguing hypothesis that our findings raise is whether ALK-rearranged tumors with high PD-L1 expression have a higher propensity towards the development of on-target ALK resistance mechanisms (such as ALK G1202R), that can be durably salvaged with 3rd generation ALK TKIs, which could account for the lack of overall survival benefit seen in our series. As ALK TKIs are being evaluated with adjuvant clinical trials such as ALINA (NCT03456076),41 the opportunity to further the characterize the relationship between ALK fusion breakpoints, co-occurring somatic alterations, PD-L1 expression, and tumor immune microenvironment will hopefully expand on our findings. High PD-L1 expression may be the natural consequence of tumor-microenvironment interactions that facilitate an immunosuppressive environment conducive to cancer cell growth. For example, release of cytokines such as IL-1β, IL-17, and TNF-α can upregulate PD-L1 and induce expression in tumor cells, which in turn promotes tumor survival and inhibits the antitumor immune response.42–46 Simple overexpression of PD-L1 on murine EML4-ALK lung cancer cells fails to alter the in vivo response to alectinib. As an alternative, we considered the hypothesis that PD-L1 expression may serve as a proxy for increased myeloid derived suppressor cells (MDSCs), which includes polymorphonuclear neutrophils (PMNs). The pathological state of immune activation is a common feature associated with the emergence of MDSCs, which exert immunosuppressive effects within the tumor microenvironment.45,46 We have previously shown that higher PMNs, including both stromal and intratumoral subtypes, is negatively associated with treatment outcomes in murine models of EML4-ALK and EGFRdel19-driven lung cancer.47,48 Among 19 pretreatment patient samples assessed by blinded independent pathology review, we found no association with lymphocyte count and PD-L1 expression, but a significant correlation between PMN count and PD-L1 expression. There are multiple proposed mechanisms for how MDSCs facilitate the escape of tumor cells including immune cell suppression through matrix remodeling and angiogenesis promotion, neutrophil-specific deletion of adipose triglyceride lipase, and suppression of natural killer cells.22–25,49,50 There is also evidence that release of reactive oxygen species by neutrophils can impact tumor colonization and blunt T-cell receptor activation.51 One possibility is that high PD-L1 expression may act as a proxy for higher PMN presence within the tumor microenvironment, which may contribute to worsened PFS to first line ALK TKIs. These findings are provocative, but caution is warranted, since our analysis only focused on PMN counts, and further studies broadening the scope to other cell types such M1 vs. M2 macrophages and other myeloid cells are needed. There are several limitations to our analysis. Although most radiographic responses to ALK TKIs occur early, our approach of using the first on-treatment scan to assess objective response rate may underestimate the best radiographic response. Second, some ALK rearrangements in this study were determined through FISH based testing. While clinically standard of care, the lack of uniform NGS testing limits our ability to draw firmer conclusions regarding whether specific ALK fusion breakpoints correlate with elevated PD-L1 expression. Third, mouse models of EML4-ALK oncogenic rearrangements may not fully replicate the nuances of the human immune system and relevant microenvironmental cues. Fourth, we had limited analyses on subsequent lines of therapy, thus limiting our ability to draw firm conclusions on how subsequent therapies contributed to overall survival outcomes. Fifth, our study was a single center retrospective series and larger clinical analyses using independent validation cohorts are warranted. Finally, postprogression biopsies were uncommon in our study limiting our ability to draw conclusions regarding how PD-L1 expression at diagnosis influences subsequent mechanisms of resistance to ALK TKIs. Future studies using larger clinical datasets and clinicogenomic databases (CGP), expanded NGS panels, and broader assessments of MDSCs will direct additional mechanistic insights to our observation. Conclusion In this study, high PD-L1 expression was identified as an independent negative predictive marker for PFS in TKI-naïve, ALK-rearranged NSCLC, even after adjustment for negative prognostic variables such as tumor burden, TP53 status, and intracranial TKI activity. Notably, PD-L1 status did not correlate with the initial depth of radiographic response or overall survival (OS), suggesting that PD-L1 expression may influence the durability of response rather than primary TKI sensitivity. While we observed a higher cumulative incidence of CNS progression in the low PD-L1 cohort, this is likely to reflect a "time-at-risk" phenomenon due to prolonged systemic PFS rather than an inherent biological predisposition. Our translational data suggests an extrinsic mechanism of resistance; while PD-L1 overexpression did not alter alectinib sensitivity in murine models, a significant correlation between PD-L1 expression and pretreatment polymorphonuclear neutrophil (PMN) counts in patient samples suggests that high PD-L1 may serve as a proxy for a myeloid-rich, immunosuppressive tumor microenvironment. These findings suggest that while patients with high PD-L1 expression derive initial benefit from first-line ALK TKIs, they may experience more rapid disease evolution, a factor that may inform clinical decision-making regarding treatment sequencing and the use of third-generation ALK inhibitors. Supplementary Material Supplementary materials Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cllc.2026.04.010.