Systemic Treatments and Molecular Biomarkers for Perivascular Epithelioid Cell Tumors: A Single-institution Retrospective Analysis

Perivascular epithelioid cell tumors (PEComa) are a large family of mesenchymal neoplasms, with variable clinical course. Evidence regarding treatment of advanced PEComas is scarce, with only one FDA-approved treatment available. The goals of this study were to provide data regarding systemic treatments for advanced PEComas and to identify biomarkers of prognostic relevance. This is a single-institution retrospective study of patients with advanced PEComas requiring systemic treatment, including malignant PEComa, angiomyolipoma (including the epithelioid variant), and lymphangioleiomyomatosis. Outcomes measured were overall survival (OS), first-line and combined progression-free survival (PFS), and tumor response. Kaplan–Meier, univariable, and multivariable Cox proportional hazards analysis were performed. A total of 29 patients were included, most with malignant PEComa (n = 17). Median OS was 204.9 months, while median PFS was 92.4 months from first-line, and 15.8 months for all lines combined. TFE3 overexpression correlated with higher risk of death (HR: 11.8, P = 0.04), and shorter median OS (P = 0.001). Chemotherapy and mTOR inhibitors showed similar OS (P = 0.84), and first-line PFS (P = 0.67). Combined PFS was similar between individual mTOR inhibitors, chemotherapy, immune checkpoint inhibitors and other treatments (P = 0.19). Different mTOR inhibitors demonstrated similar efficacy, making cost and availability important considerations when choosing a specific agent. mTOR inhibitors showed similar outcomes as chemotherapy, suggesting that these should be preferred whenever possible for patients with PEComas given the morbidity associated with chemotherapy. TFE3 overexpression highlighted a subgroup of PEComas with worse prognosis and more aggressive behavior. Significance: This study examines systemic treatments for advanced PEComas, a rare group of sarcomas, and identifies molecular biomarkers of prognosis. Our results show that mTOR inhibitors have similar efficacy as chemotherapy, and that TFE3 overexpression, on IHC or FISH, correlates with a more aggressive disease course.


Introduction
Perivascular epithelioid cell tumors (PEComa) are a heterogeneous family of soft-tissue sarcomas (STS) characterized by the presence of epithelioid cells with perivascular distribution and a peculiar IHC profile with strong expression of melanocytic and smooth muscle markers (1)(2)(3).
albumin-bound version of sirolimus, which has been approved on the basis of the results of the AMPECT trial (16). Other treatment modalities, including cytotoxic chemotherapy, tyrosine kinase inhibitors, and antiangiogenic agents, have also been used for patients with PEComas, but are overall less studied and not as effective as mTOR inhibitors (refs. 17-19; Supplementary Table S1).
Here we report on the clinical outcomes of patients with PEComas treated at our institution over the last 22 years, focusing on those with advanced disease that required systemic treatment. Specifically, we compared outcomes of patients with PEComas that received different systemic treatment modalities and explored the prognostic and predictive role of several patient characteristics and tumor biomarkers by correlating these with outcomes. In addition, we reviewed the current literature on systemic treatment of PEComas and compared it with our findings, hoping to provide valuable insights into the effectiveness of different systemic therapies for this rare and poorly understood tumor family that can help guide clinical management.

Patient Selection
Patients were selected from a database of pathology cases of PEComa, AML (including the epithelioid variant), and LAM reviewed at the Stanford Pathology Department ( Supplementary Fig. S1) between 2000 and 2022. A pathologic diagnosis of malignant PEComa was made if more than two of the following features were present: size > 5 cm, infiltrative growth pattern, high-grade nuclear atypia, high cellularity, mitoses > 1 per 50 high-power field, necrosis or lymphovascular invasion (20). Patients that required systemic treatment, either because of presence of metastases, unresectable disease, or for symptomatic control were included in the analysis. Patients treated with surgery only, those that received their treatment entirely at an outside institution, and those diagnosed incidentally during another surgery were excluded from the analysis. Patients with LAM that underwent lung transplant before receiving mTOR inhibitors or other systemic therapies were excluded as well. There were no inclusion limitations in terms of age or primary tumor site.

Data Collection
Patient information was retrospectively collected through review of patients' charts and physician documentation. Data collected included demographics, treatment regimens, clinical outcomes, radiographic response, and nextgeneration sequencing (NGS) data. NGS was available for 12 of 29 patients and obtained through either the Stanford Actionable Mutation Panel for Solid Tumors (STAMP) panel (n = 2; refs. 21,22), the Foundation One panel (Foundation Medicine, n = 4), the Altera Tumor Genomics Profiling panel (Natera, n = 2), the UCSF 500 Cancer Gene panel (n = 2), or the GPS cancer test (NantHealth, n = 1). One patient had a germline TSC mutation detected with the NGS panel from PreventionGenetics.
Of the 59 treatment episodes recorded, 40 had available radiographic scans (CT scans or MRI) that allowed direct measurements of the change in maximum tumor diameters pretreatment and posttreatment for assessment of response based on RECIST 1.1 (23). For nine more treatment episodes, the RE-CIST 1.1 response was not assessed through direct tumor measurement but abstracted from the treating physician notes, due to unavailability of outside scans to review. In accordance with RECIST 1.1, the disappearance of all lesions was considered as a complete response (CR), a reduction in the maximum tumor diameters after treatment ≥30% was considered a partial response (PR), an increase in the maximum tumor diameters ≥20% or the appearance of new lesions was considered disease progression (PD), while a change in maximum tumor diameters between −30% and +20% was considered stable disease (SD). Objective response rate (ORR) was defined as the proportion of patients achieving CR and PR out of all the patients treated, while the disease control rate (DCR) was defined as the proportion of patients that achieved PR, CR, or SD out of all the treated patients.
Outcomes measured included overall survival (OS), clinical progression-free survival from first-line treatment (first-line cPFS), and clinical progressionfree survival combined for all treatment episodes regardless of line of therapy (combined cPFS). OS was measured from the time of pathologically confirmed diagnosis to death from any cause or was censored at the time of last followup. cPFS was measured from the start date of a new treatment until the date of the next clinically documented progression, as reported in the treating physician's notes. Disease progression was defined as either worsening radiographic tumor burden, not necessarily based on RECIST 1.1, or as worsening symptoms requiring the start of another line of therapy. When a treatment was stopped or changed because of patient's or physician's preference, because of toxicity, or for unknown reasons, patients were censored at the last follow-up or at the time of the start of the next treatment line.
As far as treatment doses, everolimus was given orally at a dose of 5 mg daily, temsirolimus was given intravenously at a dose of 25 mg once a week, sirolimus was given orally at a dose between 1 and 5 mg daily adjusted to target a therapeutic drug plasma level of 5-15 ng/mL, nab-sirolimus was given intravenously at a dose of 100 mg/m 2 on days 1 and 8 of a 21-day cycle, olaparib was given orally at a dose of 300 mg twice daily, and pazopanib was given orally at a dose of 400 mg daily. Cytotoxic chemotherapy and immune checkpoint inhibitors (ICI) were administered according to standard dosing regimens.

Statistical Analysis
Patient characteristics were summarized through frequency tables. Categorical variables were reported as proportions, while continuous variables were reported as median and range. Differences between categorical variables, including ORR and DCR, were measured through the Fisher exact test.
For survival analysis, the Kaplan-Meier method was used to create the survival curves and calculate the survival proportions as well as the median survival time with associated 95% confidence intervals (CI) where applicable. The logrank test was used to calculate differences between the survival curves. We performed both a univariable and a multivariable Cox proportional hazards (PH) analysis for both OS and cPFS to calculate the HRs with associated 95% CIs. For OS, the variables included in the univariable Cox PH analysis were first-line treatment, sex, primary tumor site (uterine or extrauterine), age at diagnosis, TSC/TSC and TP mutational status, TFE3 positivity (detected through either FISH or IHC), number of lines of therapy, histology (malignant PEComa vs. other), and history of tuberous sclerosis complex. The same covariates were included in the Cox PH model of first-line and combined cPFS, with the addition of the adjuvant treatment variable, to control for the longer PFS associated with treatments administered in the adjuvant setting. Also, for combined cPFS, given that multiple treatment episodes associated with the same patients are not independent events, a frailty covariate was included in the Cox PH model to control for random patient effect. For OS, only the variables that showed a significant P value in the univariable Cox PH analysis were included in the multivariable analysis together with the treatment variable. The same

Ethics Statement
The study was conducted according to the guidelines of the Declaration of

Data Availability Statement
The data generated in this study are available upon request from the corresponding author.

Patient and Treatment Characteristics
A total of 29 patients were included in the study and their baseline characteristics are summarized in Table 1

Response Assessment
The best change in maximum tumor dimension for the 40 treatment episodes with available imaging data is reported in a waterfall plot in Fig. 1 Fig. 2D). TFE3 positivity did not correlate with higher risk of progression after first-line therapy   Supplementary Fig. S7), with no difference between patients treated with chemotherapy, mTOR inhibitors, ICIs, and other treatments in the log-rank analysis (Fig. 2E) Fig. 2F). Similarly, patients with TP mutations tended to have shorter median combined cPFS compared with those that were TP wild type (P = 0.09), and those with primary uterine tumors tended to have shorter median combined cPFS than those with extrauterine tumors (P = 0.09), even though statistical significance was not met in both cases ( Supplementary Fig. S7). Variables that correlated with shorter combined cPFS in the univariable analysis were age at diagnosis (HR: 1.   PEComa, perivascular epithelioid cell tumors. a Other: olaparib (n = 1), pazopanib (n = 1), pazopanib-everolimus (n = 1), anastrozole (n = 1), and levantinib-everolimus (n = 1).
Also, when analyzing only patients with malignant PEComa, we found no significant associations for combined cPFS (Supplementary Table S12; Supplementary Fig. S9), as well as no difference in combined cPFS between patients treated with mTOR inhibitors considered separately, ICI, chemotherapy, and other treatments (P = 0.93, log-rank; Supplementary Fig. S10).

Discussion
In this study, we report on the outcomes of what, to our knowledge, is the largest single-institution cohort of patients with advanced PEComas receiving systemic therapy, including both patients with malignant PEComa, AML, epithelioid AML, and LAM. Furthermore, it is worth mentioning that the ORR for patients treated with nab-sirolimus in our study was 14.3% versus 39% in the AMPECT trial. This discrepancy could be attributed to differences in patient populations between the studies, as well as variability in the measurement of radiological response due to the absence of centralized radiological review in our study. However, it is crucial to acknowledge the limitations of comparing a retrospective study, such as ours, to a prospectively defined study like the AMPECT trial, as well as the potential pitfalls of interstudy comparisons.
As far as tumor response, we found no difference in both DCR and ORR between mTOR inhibitors and other treatments combined. This was true both when considering the cohort as a whole and when analyzing malignant PEComas and other histotypes separately. Even though a direct comparison of tumor response between each treatment was not feasible given the size of our cohort, the ORR and DCR observed for each mTOR inhibitor, and the other treatment modalities considered separately appeared to be similar.
To our knowledge, this is one of the few reports on ICIs use in patients with malignant PEComa (26,27). Interestingly, we observed a CR in a patient treated with pembrolizumab that also showed a TMB of 17 mut/Mb. This is consistent with what observed in other cancers where a high TMB, usually above 10 mut/Mb, correlates with a more favorable response to ICIs, likely due to an increase in neoantigens and consequently in tumor immunogenicity (28,29).
Of the two other patients in our cohort that were treated with ICIs and that had imaging available for direct assessment of response, one received ipilimumabnivolumab with disease progression, and the other received pembrolizumab plus radiotherapy achieving a PR. Interestingly, both patients had TMB ≤2 mut/Mb, suggesting that, in patients with low mutational burden, radiotherapy might have a synergistic effect with ICIs as reported in previous preclinical and clinical studies (30,31).
When considering the whole cohort, we found that patients with TFE3 positivity had worse OS and increased risk of death both in the univariable and multivariable Cox analyses, as well as a nonsignificant trend toward worse combined cPFS (P = 0.05). When analyzing patients with malignant PEComa only, we found a similar association between TFE3 and shorter OS, even though in this case TFE3 positivity was not significantly associated with worse OS in the Cox analysis, likely due to the small sample size (n = 17). We defined TFE3 positivity broadly as either TFE gene rearrangement or amplification on FISH, or as TFE3 protein overexpression on IHC, because the consequence of these molecular changes is hypothesized to be an increase in TFE3 activity. Previous reports have shown that PEComas associated with TFE gene rearrangements form a distinct molecular subtype which tends to have a more aggressive behavior (24,25,32). Also, TFE rearrangements have been found to be mutually exclusive with TSC mutations in PEComas which has been hypothesized to be responsible for the worse prognosis of these patients (25,33). However, a recent report of an aggressive PEComa harboring a TSC mutation coexisting with a TFE gene amplification has challenged the dichotomy of this molecular classification (34). According to this finding, in our study, TFE3 overexpression on IHC coexisted with TSC mutation in 2 patients, and TFE gene amplification was found in a patient with a TSC mutation. Overall, our results suggest a negative prognostic role in PEComas for TFE3 positivity broadly defined as either gene amplification or rearrangement, or only as protein overexpression. Together with prior reports, our results indicate that an increase in the activity of TFE3, not exclusively through TFE gene translocation, might be responsible for the worse prognosis of this subtype of PEComas. Our results also suggest that mechanisms different from gene fusion might be responsible for the increased activity of TFE3 in PEComas. However, a major limitation of our study is that confirmation of TFE3 negativity by either IHC or NGS was available only for 3 patients, with molecular data regarding TFE3 overexpression not available for review in most patients. For this reason, further studies will be needed to confirm the prognostic role of TFE3 overexpression in PEComas, and to define the molecular basis for TFE3 overactivity in PEComas and how this could be exploited therapeutically.
In addition, in contrast to what shown in the AMPECT trial, in our study TSC mutations did not correlate with better response, either when considering all treatments or mTOR inhibitors only, and patients with TSC/TSC mutations showed similar ORR and DCR to those without TSC/TSC mutations (16). It is possible that TSC mutations might confer a favorable prognosis specifically to patients treated with nab-sirolimus rather than with other mTOR inhibitors. However, our study was not powered to answer this question and further larger studies including different types of mTOR inhibitors will be needed to address this issue.
Finally, we found that extrauterine PEComas showed higher DCR than uterine PEComas, similar to prior reports (17). This could be in part explained by the fact that most malignant PEComas in our study arose in the uterus, while none of the AML or LAM cases had a uterine location. However, even when considering only malignant PEComas, there was still a tendency toward higher DCR for extrauterine tumors. It is possible that, regardless of the histotype, there might be biological differences between uterine and extrauterine PEComas that could account for the observed differences in response.

Conclusion
In summary, this study did not show a clear difference between mTOR inhibitors and chemotherapy, both in terms of OS, PFS, and tumor response. Together with evidence from prior studies, our results suggest that mTOR inhibitors should be the agents of choice in patients with PEComas, considering the better safety profile of these agents compared with cytotoxic chemotherapy. Also, we did not find a definitive superiority of either one of the mTOR inhibitors above the others, including the new agent nab-sirolimus, suggesting that other factors like availability and cost might be important considerations when choosing a specific agent from this class. Also, our results provide preliminary evidence that increased activity of TFE3, achieved through either gene amplification or rearrangement or through protein overexpression, might define a distinct class of PEComas characterized by worse prognosis and a more aggressive behavior. Finally, this study presents important limitations, including a retrospective design, the inclusion of patients treated at a single institution, the lack of molecular testing for most patients, and a small sample size. It should be emphasized that the small sample size can lead to overestimation of subtle differences observed in our cohort and have major implications for statistical significance. Larger prospective studies comparing specific agents, as well as larger retrospective multi-institutional reviews, will be needed to confirm our findings.