Clinicopathological and Genomic Profiles of Atypical Fibroxanthoma and Pleomorphic Dermal Sarcoma Identify Overlapping Signatures with a High Mutational Burden

Atypical fibroxanthoma (AFX) and pleomorphic dermal sarcoma (PDS) are rare tumors developing in chronically sun-exposed skin. Clinicopathological features are similar, but they differ in prognosis, while PDS has a more aggressive course with a higher risk for local recurrence and metastases. In current clinical practice, they are diagnosed by exclusion using immunohistochemistry. Thus, stringent diagnostic criteria and correct differentiation are critical in management and treatment for optimal outcomes. This retrospective single-center study collected clinicopathological data and tumor samples of 10 AFX and 18 PDS. Extracted genomic DNA from tumor specimens was analyzed by a next-generation sequencing (NGS) platform (FoundationOne-CDx™). Among 65 identified mutations, TP53 inactivating mutations were observed in all tumor specimens. In both AFX and PDS, the known pathogenic gene alterations in CDKN2A, TERT promoter, and NOTCH1 were frequently present, along with high mutational burden and stable Micro-Satellite Instability status. The mutational profiles differed only in ASXL1, which was only present in AFX. Further differences were identified in likely pathogenic and unknown gene alterations. Similarities in their genomic signatures could help to distinguish them from other malignancies, but they are not distinguishable between each other using the FoundationOne-CDx™ NGS panel. Therefore, histological criteria to determine diagnosis remain valid. For further insight, performing deep tumor profiling may be necessary.


Introduction
Atypical fibroxanthoma (AFX) and pleomorphic dermal sarcoma (PDS) are rare malignant cutaneous neoplasms of fibrohistiocytic mesenchymal origin. They both typically arise in chronically sun-exposed skin of elderly patients and show male predominance. AFX and PDS share similarities in clinicopathological features and their treatment; however, they differ in the disease course [1,2]. Thus, rigid diagnostic criteria are essential to ensure appropriate disease management and follow-up.
Clinically, AFX and PDS are both characterized as rapidly growing nodules. Both tumors can demonstrate ulceration and bleeding in the disease course. Histologically, both malignancies present proliferation of pleomorphic spindle and epithelioid cells or multinucleated giant cells, which often display atypical mitotic figures and hyperchromatism. Neoplastic cells are located under an occasionally thinned or ulcerated epidermis cases where tumor thickness was ≥1 mm and extended beyond the basal excision margin, tumor thickness was designated as ≥1 mm. Demographic patient data, as well as clinicopathological and histological features of the tumors, were obtained from electronic medical records.
Out of 32 identified patients, 23 were included in the final NGS analysis through F1CDx. Of 32 enrolled patients, we collected 40 Formalin-fixed paraffin-embedded (FFPE) tissue samples. More than one sample was available from five patients. One patient diagnosed with PDS was excluded from the study due to insufficient tumor material, and samples from eight patients (total 11 biopsies) with tumor content below 20% were also excluded from the study. In total, 10 samples from 10 patients diagnosed with AFX and 18 samples from 13 patients with PDS (one patient with four longitudinal samples within the timeframe of eight months, one patient with three longitudinal samples collected within nine months) underwent further analysis with NGS through F1CDx (Figure 1).
Tumor samples were histologically re-evaluated for diagnosis of AFX and PDS, tumor infiltration depth, and ulceration status. Tumor infiltration depth was measured in millimeters from the granular layer of the epidermis to the deepest point of invasion. In cases where tumor thickness was ≥ 1 mm and extended beyond the basal excision margin, tumor thickness was designated as ≥ 1 mm. Demographic patient data, as well as clinicopathological and histological features of the tumors, were obtained from electronic medical records.
Out of 32 identified patients, 23 were included in the final NGS analysis through F1CDx. Of 32 enrolled patients, we collected 40 Formalin-fixed paraffin-embedded (FFPE) tissue samples. More than one sample was available from five patients. One patient diagnosed with PDS was excluded from the study due to insufficient tumor material, and samples from eight patients (total 11 biopsies) with tumor content below 20% were also excluded from the study. In total, 10 samples from 10 patients diagnosed with AFX and 18 samples from 13 patients with PDS (one patient with four longitudinal samples within the timeframe of eight months, one patient with three longitudinal samples collected within nine months) underwent further analysis with NGS through F1CDx ( Figure 1). Total genomic DNA was extracted from FFPE tumor tissue samples with DNA isolation kits on a Promega Maxwell RSC. Isolated DNA and gene alterations were analyzed by a validated hybrid capture-based next-generation sequencing (NGS) platform (Foun-dationOne-CDx™) at the Department of Pathology and Molecular Pathology of the University Hospital Zurich, Switzerland. The methods of F1CDx have been previously described [19]. The current F1CDx gene panel (https://www.foundationmedicine.com (accessed on 29.11.2020)) includes cancer-related genes, 309 genes with full coding exons, 21 genes with selected intronic regions, one gene with a promotor region (TERT, telomerase Total genomic DNA was extracted from FFPE tumor tissue samples with DNA isolation kits on a Promega Maxwell RSC. Isolated DNA and gene alterations were analyzed by a validated hybrid capture-based next-generation sequencing (NGS) platform (FoundationOne-CDx™) at the Department of Pathology and Molecular Pathology of the University Hospital Zurich, Switzerland. The methods of F1CDx have been previously described [19]. The current F1CDx gene panel (https://www.foundationmedicine.com (accessed on 29 November 2020)) includes cancer-related genes, 309 genes with full coding exons, 21 genes with selected intronic regions, one gene with a promotor region (TERT, telomerase reverse transcriptase), and one non-coding RNA gene (TERC, telomerase RNA component). Alterations were classified into known, likely and unknown pathogenicity by the F1CDx assay pipeline.
Patient primary tumor samples were stratified according to the histological diagnosis of AFX and PDS. For each tumor, the number of pathogenic mutations, variants of unknown significance, or both were counted per gene. In the case of multiple simultaneous mutations in a gene of a patient, the mutation count of the gene was incremented only once. The difference between the number of mutations of a gene in AFX and PDS was assessed by Fisher's exact test using R (version 4.0.4). Multiple testing adjustment was made using Benjamini-Hochberg FDR (false discovery rate) correction. Differences in the TMB in the patient subgroups were analyzed through Wilcox-Rank sum test. A FDR adjusted p-value of <0.05 was considered statistically significant.

Study Cohort
The AFX cohort included 10 patients with 10 representative AFX tumors, while the PDS cohort included 13 patients with 18 tumors. Five PDS tumor specimens represented local relapses and were obtained from two different patients with PDS.
The majority of patients with AFX (9/10, 90%,) and PDS (9/13, 69%) were male. The median age at diagnosis was 80.5 years (range: 61-90 years) in patients with AFX and 83.5 years (range: 71-92 years) in patients diagnosed with PDS. Two patients with PDS (15%) developed metastatic disease, with one patient developing nodal and one patient in-transit metastasis. No metastasis was observed in the AFX group. Local tumor relapse was present in one (10%) patient with AFX and three (23%) patients with PDS. Local relapses occurred after complete surgical excision, except for one patient with PDS who underwent incomplete excision.

Clinicopathological Features
Representative clinical and histological images of AFX and PDS from our cohort are shown in Figure 2. Most AFX and PDS primary tumors were located on the head (22/23, 96%), as represented in Figure 3. One (8%) PDS tumor was located in the lower leg. Baseline clinicopathological features of AFX and PDS of our study cohort are presented in Table 1. Upon IHC, neoplastic cells were generally negative for melanoma, cytokeratin, and muscle markers. In nine tumors, vimentin staining was used and showed positivity in 8 cases. CD10 staining was performed in two AFX tumors and was positive on neoplastic cells in both cases. (Supplemental Table S1). Table 1. Clinicopathological features of atypical fibroxanthoma (AFX) and pleomorphic dermal sarcoma (PDS).

NGS Analysis
To investigate the molecular landscape, we attempted to identify common and different driver events in primary tumors of the two patient subgroups by F1CDx analysis. Across 23 analyzed primary tumor samples a total of 65 mutations with known and likely somatic impact were identified. (Supplemental Table S2)

NGS Analysis
To investigate the molecular landscape, we attempted to identify common and different driver events in primary tumors of the two patient subgroups by F1CDx analysis. Across 23 analyzed primary tumor samples a total of 65 mutations with known and likely somatic impact were identified. (Supplemental Table S2).
All PDS and AFX tumors carried TP53 (tumor protein 53) inactivating mutation (Figure 4a). A frequent driver gene evidenced in both subgroups was CDKN2A (cyclindependent kinase inhibitor 2A) gene loss or inactivating mutation, present in six (60%) AFX and 11 (85%) PDS samples (Figure 4b). Moreover, NOTCH1 (NOTCH receptor 1) activating mutation was present in half (5/10) of the AFX and eight (62%) PDS tumors (Figure 4c). Further, TERT promoter activating mutation was frequently evidenced in six (60%) AFX and in nine (69%) PDS samples. However, no significant differences in the most frequent alterations were identified between the subgroups. ( Figure 5 and Supplemental Table S3). In contrast, alterations in CDH1 (cadherin 1) and in CD22 (cluster of differentiation-22) were mostly observed in PDS samples. CDH1 was evidenced in five (38%) PDS samples and was absent in the AFX subgroup (p = 0.053). Gene alteration in CD22 were predominant in PDS (7/13, 54%) and present in one (10%) AFX sample (p = 0.0791). (Figure 6 and Supplemental Table S4).  Among 18 PDS tumors, four longitudinal samples of one and three longitudinal samples of another patient with PDS underwent NGS analysis. Comparing recurrent specimens to primary PDS tumors among the same patient, we identified a consistency in the gene alterations. Most of the mutations showed a trend towards an increase in the variant allele frequencies. Additionally, a tendency to increase TMB was likewise observed.  We have observed a non-statistically significant difference for ASXL1 (additional sex combs-like 1) alteration between the subgroups. ASXL1 mutation was detected in 3 AFX (30%) samples but in none of the PDS samples (p = 0.0593). (Supplemental Table S3).
GNAS (GNAS Complex Locus) was previously reported as a significantly mutated gene in PDS. In our analysis, alterations in GNAS were assigned in mutations of unknown significance and were evidenced in both subgroups (AFX: 4/10, 40% and PDS: 4/13, 30%, p = 0.6734).
Among all analyzed AFX and PDS tumors, including the longitudinal samples, most tumors (27/28, 96.4%) had high (≥10 mt/mb) TMB, while one PDS tumor had a low TMB (2.52 mt/mb). Mean measured TMB was 54.59 mt/mb in AFX and 69.49 mt/mb in PDS samples (p = 0.6484). MSI status was stable in all analyzed tumor samples. (Table 2 and  Supplemental Table S6).     Among 18 PDS tumors, four longitudinal samples of one and three longitudinal samples of another patient with PDS underwent NGS analysis. Comparing recurrent specimens to primary PDS tumors among the same patient, we identified a consistency in the gene alterations. Most of the mutations showed a trend towards an increase in the variant allele frequencies. Additionally, a tendency to increase TMB was likewise observed.

Discussion
We have identified similarities in the demographic, clinicopathological, and genomic features of AFX and PDS. NGS analysis through F1CDx identified TP53 inactivating mutation in all investigated primary AFX and PDS tumors. Known gene alterations in CDKN2A, TERT, and NOTCH1 were frequently evidenced in AFX and PDS. Both subgroups harbored high TMB and had a stable MSI status. Differences between AFX and PDS were identified in likely and unknown gene alterations, although they were not statistically significant.
Our study included a cohort of 23 patients, highly representative for AFX and PDS cases considering their rarity among cutaneous malignant tumors. For AFX a prevalence rate of 0.24% was estimated in a retrospective study of over 42 000 skin tumors; however, the incidence of PDS and AFX is unknown [20]. The clinico-pathological characteristics of AFX and PDS in our cohort were as expected, showing consistency with previous reports [1,18].
Advances in NGS technology have encouraged clinicians to implement NGS analysis in guiding their daily clinical practice. This type of molecular profiling may provide insights into disease pathogenesis. Therefore, we screened for molecular signatures that could improve the distinction of AFX and PDS in addition to the current clinical practice.
NGS analysis of over 300 analyzed genes in the F1CDx panel, revealed high similarity in driver events (TP53, TERT promoter, CDKN2A, and NOTCH1), showing no significant difference between AFX and PDS. This is consistent with a previous report, where the authors pre-selected 11 genes for comparing the mutational profile of AFX and PDS [9].
Alterations in TP53, TERT promoter, CDKN2A, and NOTCH1 have been previously reported in cSCC. cSCC, as well as AFX and PDS, are known to frequently develop in chronically sun-exposed skin areas. This could explain why they share similar UV-induced gene alterations in their molecular profile [21,22].
A difference in the subgroups among the known alterations was observed in ASXL, with a p-value above the significance level of 0.05 (p = 0.0593). Few AFX patients showed ASXL1 alterations; whereas, these alterations have not been detected in patients with PDS. ASXL1 is located on chromosome 20 and belongs to the group of ASX-like genes (ASXL1-3). It encodes a nucleoprotein which enhances regulatory proteins and plays a role in the stability of gene expression. ASXL1 mutations have been previously observed in a variety of hematological malignancies in humans [23]. Even though ASXL1 could help distinguish the respective subgroups, its occurrence is restricted to a subset of patients with AFX. Therefore, ASXL1 cannot be used as an isolated distinct biomarker for AFX and explored in a larger cohort.
Including the unknown alterations as well as the known and likely mutations in our analysis, we noted potential differences in GATA4, PIK3CA, CUL4A, DAXX, MPL, CDH1, and CD22 among AFX and PDS. Alterations of unknown significance have an unclear effect on function. However, these alterations have been previously reported in the context of various cancer entities, which may indicate biological relevance.
Half of the AFX cohort harbored a GATA4 alteration of unknown significance. GATA4 (chromosome 8) encodes for GATA-4, a member of the zinc finger transcription factors family (GATA-1-GATA-6). It is involved in the differentiation of endoderm-and mesoderm originated tissues and has been reported in various cancers such as gastrointestinal, ovarian, and lung cancer and glioblastoma multiforme [24,25]. Based on TCGA (The Cancer Genome Atlas) one unknown alteration evidenced in our AFX cohort (GATA4 P359K) has been previously reported in lung squamous cell carcinoma.
Alterations in PIK3CA were present in almost one-third of the AFX cohort (3/10, 30%). One mutation (E542K) has already been described in the literature [26]. On the contrary, PIK3CA mutations were absent in the PDS group. PIK3CA encodes p110-α, a subunit of phosphatidylinositol 3-kinase (PI3K). PI3K pathway is involved in cell signaling, which regulates cell functions such as cellular growth, proliferation, differentiation, and motility and was described as activating mutation in breast cancer [27,28]. In PIK3CAmutated breast cancer targeted therapy with alpelisib, an α-selective phosphatidylinositol 3-kinase inhibitor is used for treatment [29]. Preclinical and clinical data in different tumor entities indicate that activating PIK3CA alterations may predict sensitivity to therapies targeting PI3K or AKT [30]. PIK3CA-mutated AFX tumors may therefore respond to mTOR inhibitors, including everolimus and temsirolimus, or an α-selective phosphatidylinositol 3-kinase inhibitor [31].
Comparable to PIK3CA, alterations in CUL4A, DAXX, and MPL were seen in 30% (3/10) of the AFX group and were not found in the PDS cohort. CUL4A (chromosome 13) is part of the cullin family. It acts as part of a ubiquitin-protein ligase complex and is associated with oncogenesis. Amplification and overexpression of CUL4A were reported in different tumor types (e.g., breast and prostate cancer, hepatocellular carcinoma) [32].
DAXX (chromosome 6) codes for death-associated protein 6. DAXX plays a role in tumorigenesis. Overexpression of DAXX has been reported to have an oncogenic function [33].
Myeloproliferative leukemia protein is encoded by MPL (chromosome 1). The role of activating MPL mutation has been reported in the context of myeloproliferative neoplasms [34].
On the contrary, our analysis revealed an opposite trend of alterations for CDH1 and CD22 within the explored neoplasms. CDH1 was only present in PDS (5/13, 38%). Inactivating mutation of CDH1 (chromosome 16), encoding for E-cadherin, are known to contribute to malignant cell detachment from the primary tumor [35,36].
Additionally, CD22 (chromosome 19) was evidenced in more than half of the PDS cohort and one AFX tumor. CD22 encodes for a surface molecule and is known to be expressed by most blasts of B-cell acute lymphoblastic leukemia [37].
Altogether, all listed alterations are known to play a role in cancer. However, alterations with unknown somatic impact are still poorly investigated in the context of AFX and PDS and require further in-depth studies. Moreover, the difference in the listed alterations among AFX and PDS was not significant. Thus, they are not suitable as potential biomarkers to discriminate the explored malignancies.
Our results showed that AFX exhibits high TMB, also known as a mutational load. The PDS subgroup also evidenced high TMB, which is in line with a previous report [11]. TMB is affected by different sources, including exposure to mutagenic agents such as ultraviolet light in melanoma and cigarette smoke in lung cancer [38,39]. Hence, we could hypothesize that the investigated neoplasms harbor high TMB due to chronic UV-exposure, given that they arise in high sun-exposed skin areas. A low TMB of 2.52 mt/mb was noted only in one biopsy taken from the lower leg, a site less chronically exposed to the sun.
In comparison to cSCC TMB levels were reported equally high (median TMB 45.2 mt/mb) [40]. In other cutaneous sarcoma subtypes, such as superficial angiosarcoma, high TMB (20.7 mt/mb) was lower than observed in our cohort [41].
High TMB is known to be associated with a better response to immunotherapy with checkpoint inhibitors (CPI) such as anti-PD-1 and anti-PD-L1 across diverse tumor entities [42]. Interestingly, a case study reported that two patients diagnosed with advanced PDS and high TMB have been successfully treated with anti-PD-1 (Pembrolizumab, (KEYTRUDA ® MSD Kenilworth, NJ, USA). One patient with a locally metastasized PDS (TMB 63.162 mt/mb) showed complete remission after 8 cycles of treatment with Pembrolizumab. Another patient with an inoperable relapsed PDS (TMB 77.997 mt/mb) treated with Pembrolizumab for four months, similarly showed a complete response. However, this patient received additional radiotherapy, introducing an additional factor potentially influencing treatment response [42]. Pembrolizumab is approved for other UV-induced skin tumors, such as malignant melanoma [43]. Therefore, we suggest that metastasized and/or unresectable AFX and PDS could be considered for immunotherapy with CPIs.

Conclusions
Our analysis highlights that mutations in TP53, TERT promotor, CDKN2A, and NOTCH1 are frequent in AFX and PDS, along with a high TMB. This signature could help to distinguish AFX and PDS from other malignancies, although this molecular profile is not unique for AFX and PDS.
Our data demonstrate that AFX and PDS are not distinguishable from each other with the current F1CDx panel. Thus, our findings did not reveal new rigid diagnostic criteria to distinguish AFX from PDS. The current clinical practice using histological criteria to determine the diagnosis and the prognosis remains valid. Nevertheless, an interesting trend in ASXL1 occurring more frequently in AFX has been noted. To further investigate the prevalence of ASXL1 in AFX, NGS analysis of a larger cohort is necessary. The inclusion of somatic alterations with unknown impact into our analysis showed a potential difference in GATA4, PIK3CA, CUL4A, DAXX, MPL, CDH1, and CD22 among the subgroups. Further studies with the inclusion of the unknown alterations may present significant differences between AFX and PDS. For further insight into the molecular pattern of the investigated malignancies, performing deep tumor profiling such as single-cell analysis may be required. This could identify new biomarkers unique for AFX or PDS. Furthermore, our data show that AFX and PDS harbor high TMB, which is expected from a UV-exposed anatomical site. Therefore, unresectable and/or metastasized AFX and PDS harboring high TMB could be considered for immunotherapy. Although immunotherapy could offer a new promising therapeutic approach, further investigations in randomized prospective trials of immunotherapy in AFX and PDS patients with high TMB are needed.