Perivascular Epithelioid Cell–Family Tumors in Children, Adolescents, and Young Adults: Clinicopathologic Features in 70 Cases

Patients up to age 25 years of age were included in the study, which was approved by the institutional review boards of all participating institutions. Retrospective identification of PEComa-NOS, classic AML (cAML), eAML, LAM, and clear cell sugar tumors was performed from 6 institutions and authors’ files from 1991 through 2022. Patients with TSC were identified and any associated neoplasms were documented. Clinicopathologic features identified from medical records included age, sex, tumor location, tumor size, original pathologic diagnosis, margin status, recurrence and metastatic status, treatment, last known disease status, and length of follow-up from the date of initial diagnosis. Multiple concurrent tumors at the same site with similar histologic features were noted as multiple but counted as one tumor. Two perirenal AMLs were included under the anatomic site of kidney rather than soft tissue.

All available pathology reports, original hematoxylin-eosin and immunohistochemistry-stained slides, and molecular-cytogenetic data were reviewed. For 11 cases, only pathology reports were available. The diagnoses were confirmed, with detailed methods for establishing each PEComa diagnosis summarized in Table 1. Of note, renal AMLs with 80% or more epithelioid component on resection were designated as eAMLs.¹⁴  All other renal AMLs, including those with predominantly epithelioid morphology but more than 20% of any other component (adipose tissue, spindled, large vessels) and AMLs diagnosed by biopsy only, were designated cAMLs. Tumors with predominantly epithelioid morphology outside the kidney were designated as PEComa-NOS.

When slides were available, a detailed morphologic review was performed, noting the presence of adipose tissue, thick-walled muscularized vessels, spindled cells, and epithelioid cells; the predominant histologic component; and the presence of multinucleation, extramedullary hematopoiesis, cytologic atypia, vascular invasion, necrosis, or infiltration of surrounding tissue. Mitotic count per 50 high-power fields (HPFs) was recorded, with more than 1 mitosis/50 HPFs considered increased.

Immunohistochemical stains reported and/or reviewed included melanocytic markers (HMB45, Melan A/MART1, and microphthalmia-associated transcription factor [MITF]) and muscle markers (desmin, caldesmon, and smooth muscle actin [SMA]). Additional stains used at study institutions included cathepsin K, S100, keratins, EMA, TFE3, PAX8, and CD34. Molecular and cytogenetic studies performed at the time of diagnosis on select tumoral tissue included fluorescence in situ hybridization (FISH) for TFE3 rearrangement and next-generation sequencing for fusion events and/or mutations, using clinically validated platforms.

For extrarenal and nongynecologic PEComas, high-risk features as defined by Folpe et al¹  were used to predict malignant behavior, including tumor size larger than 5 cm, infiltrative growth, hypercellularity, nuclear enlargement and hyperchromasia, mitotic activity more than 1/50 HPFs, and necrosis.^1,12  If fewer than 2 features were present, the tumor was classified as benign, and if 2 or more malignant features were present, the tumor was classified as malignant. Neoplasms 5 cm or larger with no other high-risk features or nuclear pleomorphism/multinucleated giant cells only were classified as uncertain malignant potential (UMP). With uterine PEComas, we used a modified algorithm established by Bennett et al.¹¹ 

eAMLs with resection material available were risk stratified using 2 existing criteria.^9,10  Based on the criteria established by Nese et al,¹⁰  eAMLs were stratified into low (0–1), intermediate (2–3), and high (4–5) risk based on the following 5 features: presence of TSC and/or concurrent AML, tumor size larger than 7 cm, carcinoma-like growth (morphologic pattern A); involvement of perinephric adipose and/or renal vein, and presence of necrosis.¹⁰  Using criteria established by Brimo et al,⁹  eAMLs were also stratified into either benign or malignant based on the presence of 4 atypical features: 70% or more atypical epithelioid cells (defined by polygonal cells with abundant cytoplasm; enlarged and vesicular nuclei; and prominent nuclei), 2 or more mitoses per 10 HPFs, atypical mitotic figures, and necrosis. Tumors with 3 or more features were considered malignant.⁹ 

Seventy PEComa-family tumors were identified in 64 patients (median age, 15 years; range, 2–25 years; 45 male, 19 female), diagnosed during a 3-decade period (1991–2022). Their clinical and histologic, immunohistochemical, and germline/molecular features are summarized in Tables 2, 3, and 4, respectively. Thirty-four patients (34 of 64) had confirmed TSC, and 3 had suspected TSC mosaicism. Two had Li-Fraumeni syndrome (LFS), and 1 had neurofibromatosis type 1 (NF1). The most common diagnosis was cAML (n = 49), followed by eAML (n = 8), PEComa-NOS (n = 6), and LAM (n = 5). PEComas-NOS presented exclusively in non-TSC patients, and most (4 of 6) were TFE3 rearranged. No clear cell sugar tumors were identified.

Among patients with confirmed TSC or suspected mosaicism, neoplasms included cAML (n = 36), eAML (n = 2), LAM (n = 4), and “PECosis” (n = 1) (Figure 1, A through H). No PEComas-NOS were identified in the TSC cohort. Pulmonary LAMs occurred in 4 women aged 17 to 23 years. All were multifocal and showed no malignant features (Figure 1, E and F). The case of PECosis¹⁵  presented in a young female child with multifocal perivascular lesions in the vasa vasorum of the aorta, leading to aortic degeneration (Figure 1, G and H).

cAMLs occurred predominantly in female patients (26 of 36; 72%), and were found in the kidney (n = 33) or liver (n = 3). These tumors often presented as multifocal tumors in the same anatomic site (15 of 36; 42%), and presented at ages ranging from 3 to 23 years (median, 15 years). The histologic features are summarized in Table 1. Two cAMLs demonstrated high-grade malignant transformation, characterized by an abrupt transition from bland-appearing spindled or epithelioid cells to malignant cells with hyperchromatic, pleomorphic nuclei, tumor necrosis, and increased mitotic activity. One demonstrated angiosarcomatous transformation (Figure 2, A through D), with ERG, CD34, and CD31 highlighting the malignant endothelial component. This patient died from disease 1 month after tumor resection. The other patient showed malignant transformation of the epithelioid component (Figure 2, E through H), and was alive with no distant metastasis at 103 months after diagnosis.

Two renal eAMLs occurred in patients with TSC. The first presented as an 8-cm mass in an adolescent female with concurrent cAMLs in the nephrectomy resection specimen. The mass demonstrated atypical epithelioid cells, moderate cytologic atypia, necrosis, and renal vein invasion. Given these features, it was considered malignant per Nese et al¹⁰  criteria, although the tumor did not meet criteria for malignancy based on the Brimo et al⁹  criteria. The other renal eAML presented as a 0.5-cm mass in a female child and showed no malignant features.

Patients with confirmed negative TSC germline testing or lacking clinical TSC symptoms (n = 24), including those with non-TSC syndromes (LFS = 2, NF1 = 1) as well as patients without clinical TSC history available (n = 5), were included in this cohort. Tumors included cAML (n = 13), eAML (n = 6), PEComa-NOS (n = 6), cutaneous AML (n = 1), and LAM (n = 1). The case of LAM occurred in a young adult female with no family history or clinical evidence of TSC. cAMLs occurred in the kidney (n = 12) and liver (n = 1) in patients aged 9 to 24 years (median, 18.5 years). Most cAMLs occurred in females (11 of 13; 85%) and were rarely multifocal (2 of 13; 15%). None had malignant features. One cutaneous AML involving the foot presented in an adolescent male with NF1. Morphologically, it was composed of adipose tissue with spindle cell predominance with a minimal epithelioid component, and was therefore designated as a cutaneous AML rather than PEComa-NOS.

The 6 renal eAMLs identified in this cohort arose in 4 female and 2 male patients, aged 6 to 22 years (median, 13 years). None met criteria for malignancy based on the risk stratification of Brimo et al.⁹  Using the criteria established by Nese et al,¹⁰  2 tumors were classified as benign and showed no evidence of disease at 7453 and 391 days. Two were classified as intermediate risk (scores of 2), one based on tumor size and presence of necrosis, and the other based on presence of necrosis and carcinoma-like growth. Two were considered malignant because of tumor size, necrosis, renal vein invasion, and at least focal carcinoma-like growth. One case designated as malignant based on the Nese et al¹⁰  criteria showed no evidence of disease at 12 months. Follow-up data were not available for the other tumors that were classified as malignant or intermediate risk based on these criteria.

The 6 PEComas-NOS arose in the uterus, appendix, liver, cecum, and proximal calf soft tissue. These tumors occurred predominantly in female (4 of 6) adolescents (age range, 9–17 years; median, 13.5 years). Two patients had confirmed LFS, and tumors from the other 4 patients harbored somatic TFE3 gene rearrangements (Table 4; Figure 3, A through I). All showed predominantly epithelioid morphology, and, when compared with other neoplasms, demonstrated more frequent cytologic atypia (2 of 6), necrosis (2 of 6), and increased mitoses (3 of 6). One met the Folpe et al¹  criteria for UMP and 3 were classified as malignant. One malignant PEComa-NOS arose in the cecum of a teenage female with negative germline TFE3 and TSC testing. The patient underwent a right colectomy, and 2 months later presented with abdominal pain and imaging concerning for diffuse liver metastases. After completing chemotherapy, she underwent a liver transplant. Histologic sections from the liver metastases showed purely epithelioid morphology, increased mitosis (up to 4 per 10 HPFs), nuclear pleomorphism, and necrosis (Figure 3, A through C). Targeted next-generation sequencing demonstrated SFPQ-TFE3 fusion. The patient died from disease 25 months after initial diagnosis. The second malignant PEComa-NOS occurred in the uterus of a teenage female with LFS who presented with a concurrent TFE3-rearranged renal cell carcinoma (RCC). The uterine PEComa demonstrated nuclear pleomorphism, brisk mitotic activity (>40 per 10 HPFs), and necrosis (Figure 4, A through F). Molecular testing of the tumor showed TSC2 mutation in addition to TP53. The patient developed metastasis to the lungs and died from disease 17 months after diagnosis. The third malignant PEComa occurred in the proximal calf of a teenage female with unknown TSC status. The 8.5-cm tumor demonstrated necrosis, vascular invasion, increased mitosis, and mild cytologic atypia. Follow-up data were not available for this patient. The tumor classified as UMP presented as a 3.1-cm liver mass in a male child with LFS. Although the tumor demonstrated moderate nuclear atypia with giant tumor cell and symplastic changes (Figure 5, A and B), it lacked severe nuclear pleomorphism and hyperchromasia, and there were no other features of malignancy.

Of the 6 PEComas-NOS, 4 of the 5 tested harbored TFE3 rearrangements, of which 2 were malignant. Three had the SFPQ-TFE3 fusion, which included the malignant tumor of the cecum with liver metastasis discussed above, a benign perivaginal tumor with areas of increased sclerosis in a teenage female (Figure 3, D through F), and a benign “sclerosing PEComa” in the appendix of an adolescent male (Figure 3, G through I). The TFE3 rearrangement for the fourth tumor was confirmed by FISH but the TFE3 fusion partner was not determined. This tumor showed malignant histology and presented in the soft tissue of the left proximal anterior calf in a teenage female; however, the patient was lost to follow-up following resection.

The patient with LFS and malignant uterine PEComa was diagnosed with a concurrent TFE3-rearranged RCC and underwent a partial nephrectomy at the time of uterine resection. Four additional patients (2 with TSC, 1 suspected mosaicism, 1 non-TSC) were found to have concurrent RCC and AMLs, the latter of which were usually multifocal (3 of 4). RCC types included eosinophilic solid and cystic (n = 2), papillary (n = 1), and unclassified (n = 1).

PEComas encompass several entities that have been known by different names over the years and unified under a concept initially proposed by Bonetti and colleagues.¹⁶  AMLs and LAMs, with predilection for the kidney and lung respectively, were among the first PEComas to be recognized and continue to be widely referred to by their original names. PEComas have since also, far more rarely, been recognized in diverse sites including visceral organs, soft tissue, bone, and cutaneous sites. Although known by many terms in the past, including clear cell sugar tumor, primary extrapulmonary sugar tumor,¹⁷ abdominopelvic sarcoma of perivascular epithelioid cells,¹⁸  and clear cell myomelanocytic tumor of the falciform ligament/ligamentum teres,¹⁹  these are now recognized as morphologic variants of non-AML, non-LAM PEComas, and are referred to as PEComas-NOS, or simply PEComas.¹ 

Most PEComas harbor alterations in TSC1 or TSC2 genes. A subset of PEComas, most notably AML and LAM, show strong association with TSC, an autosomal dominant disorder involving TSC1 and TSC2 germline mutations.⁵  Most TSC patients experience renal AMLs that are often multiple and bilateral.^20,21  However, fewer than half of reported patients (all ages) with renal AML have TSC.²  Outside the TSC context, sporadic AML and LAM frequently demonstrate somatic loss of heterozygosity of chromosome 16p (containing TSC2)^22,23  and show upregulation of the mTOR pathway.²⁴  Among the much rarer PEComas-NOS, many demonstrate TSC2 alterations, whereas a subset show distinct molecular mechanisms of pathogenesis associated with abnormalities in TFE3, a member of the microphthalmia-associated transcription factor (MiT) family of transcription factor proteins.^6–8  Although the vast majority of cAMLs and LAMs are benign, the clinical behavior of PEComas-NOS and eAMLs can be difficult to predict and varies from benign to highly aggressive.

As has been described in the general population, the PEComas in our young cohort appear to fall into 2 distinct clinicopathologic groups. One group includes cAMLs and LAMs, which are strongly associated with TSC and typically benign. In our cohort, these tumors showed clinicopathologic trends that have been well described. They too were strongly associated with TSC and were mostly benign.^2,25–28  Among AMLs, those related to TSC were larger (median, 5 cm versus 3.9 cm), were more frequently multiple (42% versus 14%) compared with sporadic AML,^20,21  and reflected the primary challenge in the management of AMLs among patients with TSC, whereas the LAMs showed a typical preponderance in the lungs of older females (median age, 20 years).

The other group includes the PEComas-NOS that are far less frequent, usually present in the absence of TSC, at diverse anatomic sites, with wide-ranging behavior, and harbor either TSC gene mutations or TFE3 rearrangements.^2,8  Accordingly, PEComas-NOS in our series were rare, with only 6 cases identified across our 6 institutions, many of which serve as tertiary referral centers with active consultative practices. They occurred in a wide variety of sites, all in the absence of TSC, and ranged from benign to malignant. Four harbored TFE3 rearrangements and showed epithelioid-predominant morphology typical for TFE3-rearranged PEComas.^29,30  Interestingly, the 2 PEComas-NOS without TFE3 rearrangement occurred in patients with LFS, a cancer predisposition syndrome characterized by TP53 germline alterations.

Our findings support the continued subclassification of cAMLs and LAMs within the PEComa family, as is currently reflected in the World Health Organization classification,^14,31  as they demonstrate clinicopathologic and molecular/genetic characteristics distinct from PEComas-NOS. In our cohort, however, renal eAMLs, also known as renal epithelioid PEComas, were challenging to place firmly within either group. They showed a prevalence of known TSC (2 of 8) that was lower compared with cAMLs (36 of 49), but higher than that in PEComas-NOS (0 of 6). Of 8 tumors, 2 were classified as malignant by the Nese et al¹⁰  criteria, whereas there were none by the Brimo et al⁹  criteria (further discussed below). These findings precluded clear conclusions and reflect longstanding uncertainties and controversies regarding their definition and behavior.^9,10,32 

Although the variety and many trends in our young cohort reflect those that have been described in all age groups, interesting age-related differences relating to germline abnormalities and associated oncogenic pathways are apparent.

In patients with AMLs, we found confirmed TSC or suspected TSC mosaicism to be far more prevalent (72%) in our cohort than the approximately 20% to 25% prevalence reported in the general population.³³  Although this could be due in part to the presence of TSC specialty clinics in tertiary referral centers, this finding is consistent with prior observation that TSC-associated AMLs usually present at a younger age.³⁴ 

Among the 6 PEComas-NOS, 4 (67%) harbored TFE3 rearrangements, the most common alternative oncogenic pathway, which is typically found in approximately 25% of PEComas-NOS in the general population.⁶  This apparent enrichment of TFE3-rearranged PEComas-NOS is consistent with the trend noted in a prior review of malignant PEComas in children,¹²  and with several reported cases of TFE3-rearranged PEComas in children younger than 18 years.^29,35–37  Interestingly, the only 2 PEComas-NOS without TFE3 rearrangement occurred in patients with LFS, a cancer predisposition syndrome characterized by TP53 germline alterations. These 2 cases add to rare reports of both clinically benign and malignant PEComas in young adult patients with LFS,^12,38,39  including concurrent PEComas in a pair of siblings.³⁸  A possible synergistic role of TP53 in the pathogenesis of PEComas is further supported by identification of concurrent somatic TP53 mutations in TSC-altered PEComas,^6,40,41  and in 48% to 63% of malignant PEComas.^6,40,42  The potential relationship between LFS and PEComas raises important clinical considerations in concurrent somatic and germline testing for young patients with PEComas-NOS.

The molecular and genetic landscape of PEComas in the young thus demonstrates a relative enrichment of TSC among patients with AMLs, as well as LFS and TFE3 rearrangements among patients with PEComas-NOS. These findings are consistent with the fact that somatic mutations occur over time and lead to tumor formation. In a young affected population, the development of PEComas is more likely to be accelerated owing to TSC germline or somatic mosaicism, TP53 germline alterations in LFS, or activation of alternative oncogenic pathway in TFE3 rearrangements.

The frequency of malignancy among our young PEComa cohort was not insignificant. Although cAMLs are mostly thought to be benign, 2 in our series (4% of cAMLs) demonstrated malignant transformation of a component (1 epithelioid sarcomatous and 1 angiosarcomatous), both in the context of TSC. To the best of our knowledge, only rare cases of TSC-related malignant AMLs, both eAMLs, have been reported in young patients.^43,44  Additionally, rare reports of angiosarcomatous, liposarcomatous, or leiomyosarcomatous transformation of AMLs have been documented in adults without TSC.^45–47  The 2 cases of malignant transformation in our series expand the spectrum of TSC-related malignant AMLs in young patients and suggest that although malignant transformation of TSC-related cAMLs is rare, vigilance is warranted.

Three of 6 PEComas-NOS showed malignant histology. Two of these patients developed metastatic disease, and the third was lost to follow-up. Although the precise determinants for malignancy are evolving, this spectrum of behavior and frequency of malignancy falls within the range reported in other studies of PEComas in all age groups,^1,8,11  and is in agreement with the rare reports of malignant PEComas in young patients without TSC arising in a variety of locations.^{12,13,18,48–52} 

The assessment of malignancy within our eAML cohort was problematic. In the general literature, there is reported propensity for more aggressive behavior, which would relate them to PEComas-NOS; however, this has been controversial, with some studies suggesting this not to be the case. In our cohort of 8 tumors, 2 were classified as malignant by the Nese et al¹⁰  criteria, whereas there were none by the Brimo et al⁹  criteria. Among the 4 cases with at least 12 months of follow-up data, none have recurred or metastasized, although no follow-up information was available for 1 of 2 tumors that were classified as malignant by the Nese et al¹⁰  criteria. This limited sample size and minimal follow-up data prevent us from drawing definitive conclusions. However, the findings may suggest that renal eAMLs are heterogeneous. Definitely, increased sample sizes from numerous pediatric institutions are necessary to further define these tumors and refine the predictors of aggressive tumor behavior.

Finally, our series includes rare examples in the PEComa family. This includes an instance of PECosis in the context of TSC, which has not yet been reported in a pediatric patient. This rare phenomenon is characterized by the presence of multiple PEComas and/or microscopic proliferations of perivascular epithelioid cells, usually in patients with TSC.^26,53  Fewer than 10 cases are reported in the literature,²⁶  and to our knowledge it has not been described in pediatric patients previously. Additionally, it includes an example of a cutaneous AML arising in the foot of a young male patient with NF1. These demographics fall in line with reports that cutaneous AMLs do not arise in the TSC context, have a male predominance, and are considered biologically distinct from cAMLs. In fact, in spite of their striking morphologic similarities to cAMLs, some have questioned their relationship to PEComas.^54,55 

It is noteworthy that 5 patients had concurrent RCCs, characterized as eosinophilic solid and cystic (n = 2), papillary, unclassified, and TFE3 rearranged. Except for the TFE3-rearranged RCC, the RCCs with concurrent renal AMLs are on the spectrum of RCCs associated with TSC/MTOR genetic alterations, which are associated with TSC and may occur sporadically as well.^56,57  This reflects shared pathways between some RCCs and AMLs, and reinforces the importance of considering RCCs in the differential diagnosis of AMLs. The concurrence of TFE3-rearranged RCC and malignant uterine PEComa-NOS in one of the LFS patients is of interest. Although association of TFE3-related neoplasms with LFS has yet to be reported, studies have found that DNA damage activates TFE3 and TFEB in a p53-dependent and mTORC1-dependent manner.³⁴  Together with the known phenomenon of translocation carcinomas of the kidney developing after chemotherapy,³⁴  this suggests a possible role for TP53 activation and mutations in the development of TFE3-rearranged tumors.

The main limitation of our study is the incomplete follow-up data available, most significantly among eAMLs. A few PEComas-NOS lacked follow-up, although those with follow-up information corroborated the histologic findings. Additionally, although we found unique examples of rare entities, for example angiosarcomatous transformation of an AML and PECosis, our cohort did not include some members of this very large family, for example spindle-predominant PEComas or the recently described fibroma-like PEComas.⁵⁸  Finally, this all-comer PEComa series allows us to consider the sheer breadth of this diverse family, bringing to view open questions: How do different subtypes relate to each other and what, for example, drives the differences between classic renal AMLs and PEComas-NOS? Do they develop from different cells of origin, states of differentiation, or are they subject to different influences at distinct anatomic sites? Clarification of the mechanisms that underlie their diversity may help better predict their biology and develop more precise therapeutic options for these rare and fascinating tumors.

In conclusion, PEComas in a young population display a similar spectrum and share many clinical, histologic, and molecular features that characterize PEComas in older adults. However, there are certain important differences. The primary age-associated differences identified in our cohort included (1) higher prevalence of TSC, (2) notable association with LFS, and (3) relative enrichment of TFE3 rearrangements. These findings reflect the relative importance of germline alterations or activation of alternative oncogenic pathways in TFE3 rearrangements in tumor formation given that somatic mutations tend to occur over time, and emphasizes the value of both germline and somatic molecular testing for young patients with PEComas.

The authors would like to thank Megan Troxell, MD, PhD, for her assistance in classifying Stanford’s renal cell carcinoma.