Identification and validation of tissue or ctDNA PTPRD phosphatase domain deleterious mutations as prognostic and predictive biomarkers for immune checkpoint inhibitors in non-squamous NSCLC

Although immune checkpoint inhibitors (ICIs) have recently shown promising survival advantages in treating patients with non-small cell lung cancer (NSCLC) [1, 2], only a few patients respond to current immunotherapy [3]. Therefore, identifying biomarkers that determine responsiveness to ICIs is an imperative goal. To date, programmed death ligand 1 (PD-L1) expression and the tumor mutational burden (TMB) are two predictive biomarkers for ICIs that have been validated prospectively in randomized controlled trials (RCTs) concerning NSCLC. However, both of them have several limitations. Notably, patients with high PD-L1 expression have better prognosis [4], but those with PD-L1 expression <1% could still benefit from ICIs [5]. Additionally, although PD-L1 expression is analogous, some patients still show a non-response to ICIs [6‐8]. TMB is a continuous variable, whose optimum cutoff value remains controversial. Various platforms and standards to determine the cutoff value using gene panels are non-uniform [5, 9‐11]. When it comes to liquid biopsy, PD-L1 and TMB are more unsatisfactory. Patients with pre-treatment PD-L1+ circulating tumor cells (CTCs) were associated with an inferior prognosis after treatment with PD-(L)1 inhibitors [12‐14], in contrast to PD-L1 expression correlating with favorable outcomes in tissue. In the OAK and POPLAR trials [11], bTMB ≥16 was revealed to be a predictive but not prognostic biomarker for ICIs in NSCLC. In a recent study, bTMB ≥6 was a prognostic biomarker for progression-free survival (PFS) in patients treated with ICIs [9], but whether it worked in predicting overall survivals (OS) remains uncertain.

Apart from PD-L1 expression and TMB, recent studies have reported that specific gene alterations, such as KRAS/TP53 [15], STK11/LKB1 [16, 17], EGFR [17], POLD1/POLE [18], TET1 [19], EPHA [20], KEAP1 [21], and NOTCH [22], are linked to responses to ICIs by regulating the tumor immune microenvironment (TIME) and served as biomarkers to predict the outcomes of ICIs. However, they were only prognostic but not predictive, which meant uncapable of helping selecting patients benefit from ICIs. Moreover, it would be more convenient and accessible to more patients if their mutations in ctDNA could also predict, but had not been validated yet.

Receptor-type protein tyrosine phosphatases (PTPRs) are a transmembrane immunoglobulin family comprising more than 20 members, sharing homologous phosphatase domain(s) and distributing along cell membranes [23]. By dephosphorylating target proteins, PTPRs antagonize the activities of protein tyrosine kinases (PTKs) and are involved in signal transduction, regulating tumorigenesis, and the TIME. Taking PTPRD as an example, PTPRD inactivation induced CXCL8 to promote angiogenesis and metastasis in gastric cancer [24]. Loss of PTPRD led to aberrant STAT3 activation and promoted glioma progression [25]. PTPRD deleterious mutations and deletion predicted bevacizumab resistance in metastatic colorectal cancers [26]. The mutational inactivation of PTPRD in glioblastoma was linked to melanoma malignancy [27]. Other PTPRs, including PTPRA [28], PTPRC [29], and PTPRJ [30], are also involved in regulating tumorigenesis and the TIME. More importantly, it was reported that PTPRD interacting with its ligand LRFN4 was related to responses to ICIs in bladder cancers [31] but unclear in NSCLC yet. Hence, we assumed that PTPRs also had assignable functions in response to ICIs in NSCLC.

Here, we report that the tissue and ctDNA PTPRD phosphatase domain mutations functioned as a prognostic biomarker predicting higher objective response rate (ORR) and favorable PFS of ICIs. PTPRD phosphatase domain mutations in tumor tissue predicted OS and that in ctDNA served as a predictive biomarker helping select patients benefiting from ICIs in non-squamous NSCLC (ns-NSCLC). PTPRD prediction effectiveness was independent of TMB/PD-L1 expression and may be attributed to regulating the TIME.

To the best of our knowledge, an ideal biomarker for ICIs should be both prognostic and predictive, which means being capable of both predicting the outcomes (including ORR, PFS, and OS) of ICIs and helping select patients benefiting more from ICIs than other therapies [42]. PD-L1 expression and TMB were not satisfactory enough at present. Patients with PD-L1 expression <1% could also benefit from ICIs [5], bTMB≥16 was predictive but not prognostic [11]. Additionally, both PD-L1 expression and TMB were continuous variables without clearly defined cutoff points [43] and varied largely among different detection platforms and method s[44‐46]. Hence, in addition to PD-L1 and TMB, other biomarkers are urgently needed.

In contrast to PD-L1 expression and TMB, specific gene mutations are easily detected using the next-generation sequencing assays and easily classified as mutation and wild type. These biomarkers, such as TP53/KRAS [15, 17], STK11 [16], POLE/POLD1 [18], and EHPA [20] mutations, were promising but they were not both prognostic and predictive. Additionally, some of them were not validated in ctDNA, which overcame errors caused by tumor heterogeneity during tissue biopsy and was non-invasive and easy to track dynamically. Biomarkers both prognostic and predictive and applicable to tissue and ctDNA are expected.

Here, we discovered that PTPRD phosphatase-mut could predict the outcomes in terms of ORR and PFS in discovery cohort and validate in 3 consolidated cohorts (total n=1920) that tissue or ctDNA PTPRD phosphatase-mut could predict ORR, PFS, and OS of ICIs, in non-squamous NSCLC, although it failed to predict OS in validation cohort 3. It could be explained by dual associations between ctDNA PTPRD mutations and OS. Our recent study reported that although higher bTMB correlated with a higher ORR and longer PFS, while it also correlated with a higher maximum somatic allele frequency, indicating a higher tumor burden and thus linked to shorter OS [47]. Hence, the final association between bTMB and OS was not significant. In our present study, detection of PTPRD phosphatase-mut took enough ctDNA as a prerequisite, which was based on higher tumor burden. Consequently, in this case, ctDNA PTPRD mutations were associated with poor OS. On the other hand, when treated with ICIs, patients with ctDNA PTPRD mutation could benefit from ICIs and thus had better OS. Taken together, ctDNA PTPRD mutations’ entire influence on OS was dual and of not statistical significance. In addition, PTPRD phosphatase-mut was also a predictive biomarker selecting patients who could benefit more from ICIs than chemotherapy. It is importantly that the above predictions worked in both tissue and ctDNA sequencing with the advantages of liquid biopsy, including noninvasiveness, repeatability, higher accessibility, and reduced deviations caused by tumor heterogeneity and applicable to more advanced patients. All the above PTPRD phosphatase-mut predictions were independent of the TMB, PD-L1 expression, or TP53/KRAS/EGFR mutations. Whether combination of PTPRD mutation and TMB/PD-L1 expression will bring higher efficiency is worthy of further investigation. Moreover, variant interpreters were more over single tools but not reflections of real influences, plasmid containing specific alteration was suggested to transfected into lung cancer cells to validate the predictions of specific mutations’ effects on PTPRD biological functions.

To explain why PTPRD mutations were linked to favorable ICIs outcomes, a potential underlying mechanism might be PTPRD regulating the TGF-β pathway and immune checkpoint molecules, T effectors, and IFN-γ effectors, finally influencing CD8+ T cell infiltration via its phosphatase activity and making TIME “hot.” Here, we provide novel clues about the associations between PTPRD and TIME, while molecular biological experiments were still warrant to verify the specific mechanism.

We also found that PTPRD was linked to OS in melanoma patients treated with ICIs and in bladder cancer patients, there was also a such tendency (p=0.062). Consistent with our findings, a recent study reported that the interaction of PTPRD (referred to WT) with its ligand LRFN4 was related to a poor response to Atezolizumab in bladder cancer s[31]. Thus, whether PTPRD mutations were a pan-cancer biomarker for ICIs was worthy of further investigations. Additionally, prospective trials incorporating tissue or ctDNA PTPRD phosphatase-mut as a biomarker are worth conducting both in NSCLC and other cancer types. We are planning to validate these findings prospectively in several coming soon randomized phase II/III study of a PD-(L)1 antibody in multiple cancer types.

This study had several limitations. First, it was a retrospective analysis based on our own cohort and 7 public cohorts. Prospective studies were requested for further verification. Second, a certain degree of research heterogeneity existed. Different cohorts used different strategies, such as the monotherapy of anti-PD-(L)1 or combination with anti-CTLA-4. The predictive efficiency of PTPRD in different drugs requires further investigation. Third, molecular biological experiments are still warranted to validate how specific PTPRD mutations influence immune-related pathways. Last but not least, whether ctDNA PTPRD mutations were consistent with tissue was still unclear. In our previous study, we had proved that ctDNA EGFR mutations were consistent (although not totally) with tissue EGFR mutations and could predict the outcomes of response to EGFR TKI [48]. Hence, like EGFR mutations, ctDNA PTPRD mutations were also supposed to be consistent with tissue mutations. In our present study, the mutation frequencies were similar (both about 10%) across tissue and ctDNA cohorts and PTPRD mutations were associated with PFS and ORR in both tissue and ctDNA cohorts; PTPRD mutations in ctDNA were consequently inferred to be consistent with tumor tissue. Nevertheless, paired tissue and ctDNA samples are expected to been sequenced to verify that in the future.

To conclude, our study demonstrated that both tissue and ctDNA PTPRD phosphatase-mut could serve as a prognostic biomarker predicting the ORR and PFS. Tissue PTPRD phosphatase-mut was also a prognostic biomarker for OS. Additionally, PTPRD phosphatase-mut could also select benefiting patients between ICIs and chemotherapy. Further prospective trials are warranted.