Purpose:

Pleomorphic dermal sarcoma (PDS) is a rare malignant cutaneous tumor with an unknown cell of origin. Locally defined tumors can be treated by curative excisions, whereas advanced stages of the disease are difficult to treat, using standard regimens.

Experimental Design:

We performed whole-exome sequencing on a cohort of 28 individuals and corresponding transcriptomic analysis on 21 patients, as well as quantitative IHC image analysis on 27 patients.

Results:

PDS exhibits a universally high mutational load (42.7 mutations/mega base) with an inflamed, immunogenic tumor microenvironment. Three cases of PDS showed response to immune checkpoint blockade. Local mutation rate variation together with mRNA expression data demonstrate that PDS form a distinct entity, with PDGFRB as a lineage marker. In addition, we found that PDS is of mesenchymal, fibroblastic differentiation.

Conclusions:

PDS is of fibroblastic differentiation and exhibits a strong susceptibility to immunotherapy, including a high mutational burden and an inflamed tumor microenvironment.

This article is featured in Highlights of This Issue, p. 5541

Translational Relevance

Pleomorphic dermal sarcomas (PDS) uniformly exhibit a high mutational burden, with loss of TP53 and CDKN2A/B in almost all cases. By performing three independent comparative analyses between common skin tumors (gene expression analysis, spatial mutational density on DNA level, and IHC), we show that PDS is of mesenchymal, fibroblastic differentiation. As PDS exhibit a less favorable prognosis than squamous cell carcinomas (SCC) of the skin, but diagnostic biomarkers are limited, we propose PDGFRB as a lineage marker to discriminate PDS from SCC. We report response to immune checkpoint inhibition in three individuals and highlight a proinflammatory tumor microenvironment with specific targetable genomic and cellular alterations in PDS.

Pleomorphic dermal sarcomas (PDS) are cutaneous malignant tumors arising in UV-exposed locations and are typically diagnosed in elderly individuals. While the majority of PDS can be treated by curative excisions, local recurrences and distant metastases occur in about one-third of these patients (1–3). However, no standard treatment regimens are currently available for late-stage PDS.

Although the prevalence of PDS remains largely unknown, these tumors are frequently observed within the group of cutaneous sarcomas. Therefore, PDS is generally believed to be of mesenchymal lineage (4). However, recent data demonstrate similarities between PDS and cutaneous squamous cell carcinoma (SCC; refs. 5, 6). This may suggest that PDS originates from dedifferentiated SCCs of the skin (6). Nevertheless, the PDS cell of origin still remains elusive. Clinically, there is a strong rational to discriminate PDS from SCC, as PDS exhibits a less favorable prognosis in comparison with SCC (7, 8).

So far, little is known about molecular alterations in PDS (9, 10). Thus, there is a substantial clinical need to uncover systemic treatment options for patients with advanced stages of PDS.

Sample collection and clinical data

We analyzed samples from a total of 28 individuals of primary PDS [23 males and five females; median age: 80 years (58–94 years)] from UV-exposed locations (head/neck, n = 26; shoulder, n = 2). All cases were treatment naïve. Clinically, 23 of 28 patients (82.1%) had no tumor progress (local or distant relapse) over a time period of 31.8 ± 18.6 months, whereas five patients (17.9%) showed a disease progression after a mean of 17.0 ± 13.0 months (t test, two-sided; $P\ = \ 0.067$⁠). The study protocol was conducted in accordance with the ethical guidelines of the 1975 Declaration of Helsinki. All investigations were performed after approval of the institution's human research review committee of the University of Cologne, Germany (registration No. 15-307). Informed written consent was obtained from each subject.

Macrodissection and DNA/RNA isolation

Six 10-μm sections were each macrodissected for DNA and RNA isolation using hematoxylin-eosin stained slides with marked tumor and normal regions as a reference. Regions were marked by a pathologist (A. Quaas, D. Helbig, R. Buettner, S. Klein). DNA and RNA were isolated using the Maxwell LEV DNA FFPE Kit or Maxwell RSC RNA FFPE Kit, then further processed on the semiautomated Maxwell LEV Instrument and Maxwell RSC Instrument, respectively (Promega) according to manufacturer's instruction. RNA extraction included DNAse digestion. DNA was quantified using Qubit dsDNA BR Assay Kit and both Qubit RNA BR Assay Kit and NanoDrop spectrophotometer was used for RNA quantification (All from Invitrogen).

Whole-exome sequencing

Whole-exome sequencing (WES) was performed from genomic DNA, using the SureSelect Human All Exon V6 (Agilent) according to the manufacturer's instructions. Obtained exome libraries were paired-end sequenced on a NovaSeq (2 × 150 bp) platform (Illumina). We achieved a median sequencing depth of 84 in the tumors and 86 in corresponding normal tissues (Supplementary Table S3).

Dideoxy sequencing

Two primers were used to detect DDX31 R72K mutations: DDX31-R72K-F1 (5′-TACAGCTGGAGCCGTTCCG-3′), DDX31-R72K-R1 (5′-CCATGGTCTGCGTGGGTG AC-3′), DDX31-R72K-F2 (5′-CGGGACGAGTCCGTACAGC-3′), DDX31-R72K-R2 (5′- GGTGCGCAGACTGCTGGG-3′). PCR was conducted using Platinum Taq DNA polymerase (Invitrogen) in C1000 Touch Thermal Cycler (Bio-Rad). PCR products were purified using QIAquick PCR Purification Kit (Qiagen) and sent for sequencing (Eurofins Scientific).

Comparison with published skin malignancies

Somatic single-nucleotide variants (SNVs) of cutaneous melanoma (CM), cutaneous SCC, and cutaneous basal cell carcinoma (BCC) were obtained from a recently published study (11).

Analysis of WES data

Sequencing reads were aligned to the human reference genome (NCBI build 37/hg19) by using version 0.7.13-r1126 of the BWA mem aligner. Corresponding read pairs were then masked as potential PCR duplicates and areas of overlapping pairs of reads were excluded. Somatic alterations (substitutions, insertions, deletions, copy numbers), as well as cellularity estimates were determined by our in-house cancer genome analysis pipeline (12, 13).

Signature fitting with the catalogued mutational signatures

To estimate the contribution of previously validated mutational signatures in PDS, melanoma, BCC, and SCC tumor types, we used the catalogued signatures extracted with SigProfiler from WES data (11). The SNV counts were classified into 96 categories based on their immediate 5′ and 3′ sequence context (14). Normalization was performed as such that for each of the 96 categories, the counts were multiplied to a ratio between the exome and the genome occurrence of the respective trinucleotidic class. In our model, the SNV count matrix $A$ is represented as product of two nonnegative matrices $W$ and $H$⁠: $A \approx W \cdot H$⁠. Here, $H$ is the mutational signature matrix and $W$ is the so-called emission matrix reflecting the number of mutations associated to each signature of every sample. To fit the catalogued signatures $H$ to the normalized SNV counts $A$⁠, the emission matrix $W$ was imputed. The optimization of $W$ was performed by minimizing the Frobenius norm $\parallel A - W \cdot H\parallel {{\rm{\ }}_F}$⁠, while maintaining the constraint ${W_{ij}} \ge 0$ (⁠$i\ = \ 1, \cdots ,{n_g}$ and $j\ = \ 1, \cdots ,{n_k}$⁠). The set of reference signatures best explaining $A$ was selected such the Frobenius norm was notably reduced.

Spatial mutational density

Spatial mutational density (SMD) correlates to the cancer type reflecting its cell of origin. For highly mutated cancer types also, exonic region from the WES data can be used (15). We compared the SMD of PDS, CM, BCC, and SCC tumors. For each sample, a collection of 10 Mb windows (n = 295) across the genome were used by counting the number of point mutations in each window. This number was then normalized to the total number of mutations. A feature selection step was performed using a restricted Boltzman machine, from which we extracted 150 features. The selected features were then embedded and plotted in two dimensions using t-stochastic neighbor embedding (t-SNE; ref. 16).

Quantification of spatial immune cell distribution

For quantification of CD4 and CD8 inflammatory cells, we digitalized IHC slides using a NanoZoomerS360 Hamamatsu whole-slide-image (WSI) slide scanner at 40× magnification. Subsequent images were segmented, and quantification of immune cell populations were performed on WSI using a watershed algorithm. For cellular classification, a neural network was trained.

IHC

IHC stainings were performed using the BOND MAX from Leica) according to the manufacturer's protocol. All immunostainings were scored independently by trained pathologists (A. Quaas, R. Buettner, S. Klein), as well as a dermatopathologist (D. Helbig).

Mutational landscape of PDS

Similar to other tumors with chronic UV exposure, we found that PDS exhibit an extremely high tumor mutational burden (TMB) with an average of 42.7 nonsynonymous variants per mega base (Fig. 1A). This particularly high mutational burden ranks among the most highly mutated cancers, even exceeding the mutation frequency of CM and at a comparable level to cutaneous SCC (Supplementary Fig. S1).

Figure 1.

High mutational burden and loss of TP53 and CDKN2A/B are common in PDS. A, Commonly mutated genes of PDS, where corresponding type of mutation is shown within the legend below. The dotted line represents the average of variants per mega base. The mutational frequency of PDS is compared with cutaneous SCC, cutaneous BCC, and CM, where the bars indicate their relative prevalence (percentage, right). Mutational signature analysis of each sample is shown as a heatmap below the commonly mutated genes. Cross entity comparison of mutational signatures shown below the right panel of mutational frequencies. Clinically, locally or systemically progressed tumors are indicated with a black bar. B, Copy-number variation of PDS tumors. The dotted line indicates a threshold of 0.01 FDR. Blue indicates deletions, red indicates amplifications. Chromosomes of interest are highlighted.

Figure 1.

High mutational burden and loss of TP53 and CDKN2A/B are common in PDS. A, Commonly mutated genes of PDS, where corresponding type of mutation is shown within the legend below. The dotted line represents the average of variants per mega base. The mutational frequency of PDS is compared with cutaneous SCC, cutaneous BCC, and CM, where the bars indicate their relative prevalence (percentage, right). Mutational signature analysis of each sample is shown as a heatmap below the commonly mutated genes. Cross entity comparison of mutational signatures shown below the right panel of mutational frequencies. Clinically, locally or systemically progressed tumors are indicated with a black bar. B, Copy-number variation of PDS tumors. The dotted line indicates a threshold of 0.01 FDR. Blue indicates deletions, red indicates amplifications. Chromosomes of interest are highlighted.

Close modal

By determining genes that were significantly enriched for point mutations across our set of 28 whole-exome sequenced PDS, we detected damaging, loss-of-function mutations of TP53 in all cases (Fig. 1A). In addition, 93% of the lesions harbored either a CDKN2A/B mutation (68%) or deletion (25%), or both (46%), while 7% of the 28 cases showed even a biallelic loss. In comparison, other skin tumors (SCC, BCC, CM) showed a significantly lower mutation frequency of CDKN2A/B (Fisher exact test; PDS vs. SCC: $P\ = \ 8\ \times \ {10^{ - 5}}$⁠; PDS vs. BCC: $P\ = \ 3\ \times \ {10^{ - 20}}$⁠; PDS vs. CM: $P\ = \ 7\ \times \ {10^{ - 12}}$⁠). Further significantly mutated genes in PDS also included DNHD1, GNAS, RTN1, RTL1, ZBTB7A, NCKAP5L, and FAM200A. All of these genes were also found to be mutated in SCC tumors, but at a significantly different distribution (Wilcoxon matched-pairs signed-rank test; $P\ = \ 5\ \times \ {10^{ - 4}}$⁠). Both, BCC and CM showed barely any mutations in these genes (Fig. 1A). Furthermore, we found a recurrent p.R72K mutation affecting DDX31 in five cases, which was confirmed by dideoxy sequencing (Supplementary Fig S2A). In an independent validation cohort of PDS, we found one of 25 cases harboring the DDX31R72K mutation. The recurrent mutation DDX31R72K was also found in three of 68 SCC tumors (4.4%), as well as in BCC (1%) and CM (0.17%; Supplementary Fig S2B). As this mutation occurs outside of a known functional domain, its biological consequence remains elusive.

We further identified alterations of genes regulating the PDGFR pathway. These included amplifications of PDGFRA/KIT in three cases and mutations in one sample affecting the kinase domain of KIT (Supplementary Table S1). Additional recurrent copy-number alterations in PDS included the amplification of 2p25.3 (Fig. 1B), harboring TRAPPC12. TRAPPC12 mRNA expression was associated with a significantly higher gene expression in tumors with 2p25.3 amplifications, when compared with unaltered cases (⁠$P\ = \ 0.0236$⁠; Supplementary Fig. S3A). In addition, high gene expression levels of TRAPPC12 were associated with worse survival in CM available in the The Cancer Genome Atlas dataset (Supplementary Fig. S3B). In addition to homozygous deletions in CDKN2A/B in two cases, we found recurrent deletions in 8p23.3-4 (Fig. 1B).

Molecular distinctions between PDS and other skin tumors

By investigating the mutational signatures of PDS, we found both UV-induced signatures 7a and 7b (Fig. 1A). In addition, three cases of PDS revealed signature 44, which has been associated with defective DNA mismatch repair (Supplementary Fig. S3C). To further investigate the similarities and differences between other UV-induced cutaneous malignancies, we compared the mutational signatures of PDS, SCC, BCC, and CM (Fig. 1A). While signature 44 was more prominent in SCC, both signatures 7a and 7b were found at a similar proportion in PDS. Neither SCC, BCC nor CM showed a similar presence of both UV-induced signatures 7a and 7b in our analysis.

Of the selected genes shown in Fig. 1A, PDS shared the most similarity with SCC. Therefore, we performed transcriptome sequencing of SCC (n = 6) and PDS (n = 21) to assess differences in their gene expression levels (Supplementary Table S2). By hierarchical clustering of the gene expression profile, we clearly could confirm a separation of both entities (Fig. 2A; Supplementary Fig. S4A). To further show that PDS is a distinct entity with a defined mutational landscape in comparison with other skin malignancies (BCC, CM, and SCC), we analyzed local mutation rate patterns. It has recently been shown that mutation rate patterns can be used to accurately predict the tissue type of the respective tumor (15, 17). Therefore, we adapted the concept of local mutation rate variation to WES data, which then allowed us to compare a variety of acquired somatic mutations of PDS tumors between other skin malignancies. We confirmed that PDS indeed clustered as a distinct entity, separating from tumors of epithelial origin (SCC, BCC) and neural crest originated entities (CM; Fig. 2B). As the cluster of SCC was adjacent to PDS (Fig. 2B), we then performed a differential gene expression analysis of these two entities, followed by a gene set enrichment analysis of differentially and highly expressed genes (⁠$P\$<$\ 0.05;{\rm{\ adjusted\ for\ multiple\ testing}}$⁠). Using the ARCHS tissue-specific gene sets this analysis yielded a mesenchymal, fibroblastic differentiation of PDS, in contrast to the epithelial SCC (Fig. 2C). To further validate these findings, we performed NanoString of a subset of samples used for transcriptome sequencing due to tissue limitations (SCC: n = 6; PDS: n = 10; Supplementary Fig. S4B–S4D). This analysis confirmed the previous finding of the mesenchymal differentiation of PDS.

Figure 2.

Differential gene expression analysis in PDS reveals mesenchymal expression signatures with enrichment of PDGFRA/B expression. A, Heatmap and hierarchical clustering of transcriptomic data of PDS and cutaneous SCC as indicated. B, t-SNE plot of SMD analysis of PDS, cutaneous SCC, cutaneous BCC, and CM. The number of variants is represented by dot size according to the legend. C, Gene set enrichment analysis of significantly upregulated genes in PDS compared with SCC using ARCHS4 (36) tissue expression (enrichr; ref. 37). Cell types that best reflect the representation of highly differentially expressed genes (P < 0.05; adjusted for multiple testing) in PDS compared with SCC are shown in red (PDS, n = 21; SCC, n = 7).

Figure 2.

Differential gene expression analysis in PDS reveals mesenchymal expression signatures with enrichment of PDGFRA/B expression. A, Heatmap and hierarchical clustering of transcriptomic data of PDS and cutaneous SCC as indicated. B, t-SNE plot of SMD analysis of PDS, cutaneous SCC, cutaneous BCC, and CM. The number of variants is represented by dot size according to the legend. C, Gene set enrichment analysis of significantly upregulated genes in PDS compared with SCC using ARCHS4 (36) tissue expression (enrichr; ref. 37). Cell types that best reflect the representation of highly differentially expressed genes (P < 0.05; adjusted for multiple testing) in PDS compared with SCC are shown in red (PDS, n = 21; SCC, n = 7).

Close modal

PDS shows a mesenchymal fibroblastic differentiation with PDGFRB as a linage marker to discriminate PDS from SCC

To directly characterize the differentiation of PDS, we made use of the large variation of tumor purity across the samples (Supplementary Fig. S4E) to extract tumor-specific genes from mRNA expression data. To this end, we ranked each gene by its correlation between expression and tumor purity and selected statistically significant genes (⁠$P\ >\ 0.05$⁠; adjusted for multiple testing) for gene set enrichment analysis. We found that fibroblasts showed the highest enrichment score (Fig. 3A). Altogether, these results strongly support the hypothesis that PDS originate from fibroblasts.

Figure 3.

PDGFRB expression discriminates between PDS and SCC. ARCHS4 tissue gene set enrichment analysis (A), where genes highly correlating to tumor purity (⁠$P\ > \ 0.05$⁠; adjusted for multiple testing) were used as input. B, ARCHS4 kinase coexpression for the same genes. Representative pictures of protein expression of PDGFRB using IHC in PDS and (scale bar, 500 μm; C) in SCC (scale bar, 250 μm; D). E, Assessment of PDGFRB expression using IHC on PDS and SCC. The boxes indicate cases, where brown boxes indicate positivity, while gray boxes show negativity. Undifferentiated SCCs, including spindle cell morphology are highlighted (u, undifferentiated).

Figure 3.

PDGFRB expression discriminates between PDS and SCC. ARCHS4 tissue gene set enrichment analysis (A), where genes highly correlating to tumor purity (⁠$P\ > \ 0.05$⁠; adjusted for multiple testing) were used as input. B, ARCHS4 kinase coexpression for the same genes. Representative pictures of protein expression of PDGFRB using IHC in PDS and (scale bar, 500 μm; C) in SCC (scale bar, 250 μm; D). E, Assessment of PDGFRB expression using IHC on PDS and SCC. The boxes indicate cases, where brown boxes indicate positivity, while gray boxes show negativity. Undifferentiated SCCs, including spindle cell morphology are highlighted (u, undifferentiated).

Close modal

To find potential linage-specific genes in PDS, we repeated the same procedure from the previous paragraph, but used the ARCHS gene set for kinase coexpression in the enrichment analysis. This yielded PDGFRA/B as being the most enriched kinase (Fig. 3B). Furthermore, we could confirm this result in our NanoString dataset (Supplementary Fig. S4D). To show that this difference is also seen at the protein level, we stained for PDGFRA/B. While all PDS lesions (n = 28) were positive for PDGFRB (Fig. 3C and E), only a fraction of cases showed PDGFRA expression (Supplementary Fig. S4G). The universal PDGFRB expression was further validated in an independent cohort of 13 additional PDS tumors (Fig. 3E). In contrast, all tested SCC tumors (n = 85) were negative for PDGFRB (Fig. 3D and E), including five cases of undifferentiated SCC (Supplementary Fig S4F). This strongly suggests that PDGFRB could be used as a linage marker to specifically discriminate SCC from PDS.

PDS tumors mostly exhibit an inflamed, proimmunogenic tumor microenvironment

The high mutational burden of PDS suggests a potential susceptibility to immunotherapeutic approaches. Thus, patients with advanced stages of PDS might benefit from treatment with programmed cell death 1 (PD1)/programmed cell death ligand 1 (PD-L1) inhibitors, such as pembrolizumab, nivolumab, the anti–CTLA-4 antibody ipilimumab, or a combination of these agents. While two of our patients treated with pembrolizumab still remain in remission (patient No. 1, complete response, 21-month follow-up; patient No. 2, complete response, 4-month follow-up), another individual with locally progressed PDS responded well (patient No. 3, local tumor meltdown after single infusion) to the same treatment (Supplementary Fig. S5). Collectively, these three cases suggest a high efficacy of immune checkpoint blockade in advanced stage PDS. Furthermore, we quantified CD4- and CD8-positive T cells, for marginal and central tumor zones, as well as PD-L1 expression on tumor cells (Fig. 4AC). This analysis revealed only six cases with low amounts of both CD4/CD8-positive cells and low PD-L1 expression (immunologically cold tumors), while the remaining 22 cases showed either inflamed or immunogenic characteristics (immunologically hot tumors). We could not detect a significant difference of TMB between immunologically hot and cold tumors (⁠$P\ = \ 0.7983$⁠; Fig. 4C). A further differential gene expression analysis between immunologically hot and cold tumors revealed upregulation of TIGIT in the immunologically hot tumors (Fig. 4D). In addition, we found a substantial correlation between TIGIT and CD8A (Fig. 4E).

Figure 4.

Immune profiling of PDS tumors reveals distinct immunologic subgroups. Representative heatmap of an inflamed PDS case showing CD8 cell density within the tumor compartment (A), as well as in an immunologic cold PDS case (B). C, Heatmap of inflammatory cells (CD4 positive and CD8 positive), according to their distribution inside the tumor (center and infiltration zone), as well as PD-L1 expression status on tumor cells (tumor proportion score). The clusters are divided into “hot” and “cold.” D, Volcano plot of differentially expressed genes between immunologic “hot” versus “cold” tumors. Significant genes are highlighted with text and red dots. E, Correlation of CD8A and TIGIT on mRNA expression level using Pearson correlation coefficient.

Figure 4.

Immune profiling of PDS tumors reveals distinct immunologic subgroups. Representative heatmap of an inflamed PDS case showing CD8 cell density within the tumor compartment (A), as well as in an immunologic cold PDS case (B). C, Heatmap of inflammatory cells (CD4 positive and CD8 positive), according to their distribution inside the tumor (center and infiltration zone), as well as PD-L1 expression status on tumor cells (tumor proportion score). The clusters are divided into “hot” and “cold.” D, Volcano plot of differentially expressed genes between immunologic “hot” versus “cold” tumors. Significant genes are highlighted with text and red dots. E, Correlation of CD8A and TIGIT on mRNA expression level using Pearson correlation coefficient.

Close modal

Besides TP53 mutations in all PDS cases analyzed (n = 28), we found alterations of CDKN2A/B in about 90% of the cases (Fig. 1A). Loss of CDKN2A/B can be detected in many cancer entities and has been associated with disease progression in SCC and CM (24–27). It has been shown that tumors with a loss of CDKN2A/B may benefit from CDK4/6 inhibitors, such as palbociclib, abemaciclib or ribociclib, all approved for the treatment of metastasized breast cancer (24, 28–30). Further therapeutically relevant alterations in PDS included amplifications of PDGFRA [with confirmed PDGFRA expression on protein level (Supplementary Fig. S4F)] and mutations within the kinase domain of KIT (Fig. 1A). Clinically, several drugs have proven to induce long-term remissions in PDGFR-expressing cancers, such as gastrointestinal stromal tumors, dermatofibrosarcoma protuberans, or myeloid malignancies (31–35).

Recently, it has been demonstrated that variations of the local mutation rate are specific to the cell of origin (15, 17). By applying this method to a collection of PDS, SCC, CM, and BCC tumors, we could show that PDS constitutes a distinct entity that is compatible with a fibroblast differentiation (Figs. 2B, C, and 3A). Moreover, PDGFRB appeared uniformly expressed in PDS (n = 41) but lacked expression in all well-differentiated and undifferentiated SCC tumors (n = 85; Fig. 3C–E). Therefore, PDGFRB could specifically discriminate between PDS and SCC, implying its diagnostic potential as a future linage marker.

High (TMB) has been associated with favorable response to immune checkpoint blockade (18). While two of our patients treated with pembrolizumab (19) still remain in remission, a third individual showed response after treatment with an anti-PD1 antibody (Supplementary Fig. S5). Interestingly, we did not observe any difference in the mutational burden of PDS tumors after classifying them into immunologically hot and cold tumors, based on CD8 infiltration and PD-L1 expression (Fig. 4C). Therefore, additional immunogenic markers, including PD-L1 and CD8, might be necessary to identify inflamed PDS tumors that might be susceptible for immune checkpoint inhibition. Furthermore, we also observed deleterious PTEN mutations in three cases (Fig. 1A), which have been associated with intrinsic resistance to immune checkpoint inhibitors in other entities (20–22)—reasoning to assess these alterations prior to potential immunotherapeutic interventions in PDS. Two of our three cases that showed complete response to pembrolizumab were also characterized using WES and image analysis. These two patients could be assigned to the “immunologically hot category,” while no PTEN mutations were detected.

In addition, we found a strong correlation between CD8A and TIGIT in PDS (Fig. 4E). This may further provide a rational to explore an additional treatment option in advanced PDS, as TIGIT is currently being investigated as novel immune checkpoint in clinical trials (23).

In summary, we performed the first comprehensive characterization of PDS and found that these tumors have an extremely high mutational load. Our analyses revealed a strong indication that PDS shows a fibroblastic differentiation with PDGFRB as a specific linage-specific marker. Furthermore, PDS showed a high susceptibility to immune checkpoint blockade, which is supported by the treatment response to an anti–PD-1 antibody (pembrolizumab) in three cases. Finally, our results pinpoint TIGIT as a potential novel target to treat advanced-stage PDS.

H.C. Reinhardt reports grants from Deutsche Forschungsgemeinschaft and German Ministry for Education and Research during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

S. Klein: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. A. Quaas: Conceptualization, formal analysis, investigation, methodology, writing-original draft. K.-W. Noh: Formal analysis, validation. M. Cartolano: Data curation, software. N. Abedpour: Data curation, software, formal analysis. C. Mauch: Conceptualization, investigation. J. Quantius: Data curation, investigation. H.C. Reinhardt: Resources, writing-original draft. R. Buettner: Conceptualization, formal analysis, supervision, project administration. M. Peifer: Conceptualization, resources, data curation, software, formal analysis, supervision, investigation, visualization, methodology, writing-original draft, writing-review and editing. D. Helbig: Conceptualization, validation, writing-original draft.

We thank Wiebke Jeske and Susann Zupp for technical assistance performing the TMA and IHA staining.

This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation: SFB829 to D. Helbig and C. Mauch; SFB1399 to M. Peifer, R. Buettner, A. Quaas, and H.C. Reinhardt), the German Cancer Aid (Mildred-Scheel Professorship to M. Peifer), German Ministry for Education and Research (BMBF e:Med program: 01ZX1901 to M. Peifer and H.C. Reinhardt), and the Else Kröner-Fresenius Stiftung (2016_Kolleg.19 to S. Klein).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Miller
K
,
Goodlad
JR
,
Brenn
T
. 
Pleomorphic dermal sarcoma: adverse histologic features predict aggressive behavior and allow distinction from atypical fibroxanthoma
.
Am J Surg Pathol
2012
;
36
:
1317
26
.
2.
Persa
OD
,
Loquai
C
,
Wobser
M
,
Baltaci
M
,
Dengler
S
,
Kreuter
A
, et al
Extended surgical safety margins and ulceration are associated with an improved prognosis in pleomorphic dermal sarcomas
.
J Eur Acad Dermatol Venereol
2019
;
33
:
1577
80
.
3.
Tardio
JC
,
Pinedo
F
,
Aramburu
JA
,
Suarez-Massa
D
,
Pampin
A
,
Requena
L
, et al
Pleomorphic dermal sarcoma: a more aggressive neoplasm than previously estimated
.
J Cutan Pathol
2016
;
43
:
101
12
.
4.
Mentzel
T
,
Requena
L
,
Brenn
T
. 
Atypical fibroxanthoma revisited
.
Surg Pathol Clin
2017
;
10
:
319
35
.
5.
Lai
K
,
Harwood
CA
,
Purdie
KJ
,
Proby
CM
,
Leigh
IM
,
Ravi
N
, et al
Genomic analysis of atypical fibroxanthoma
.
PLoS One
2017
;
12
:
e0188272
.
6.
Koelsche
C
,
Stichel
D
,
Griewank
KG
,
Schrimpf
D
,
Reuss
DE
,
Bewerunge-Hudler
M
, et al
Genome-wide methylation profiling and copy number analysis in atypical fibroxanthomas and pleomorphic dermal sarcomas indicate a similar molecular phenotype
.
Clin Sarcoma Res
2019
;
9
:
2
.
7.
Brantsch
KD
,
Meisner
C
,
Schonfisch
B
,
Trilling
B
,
Wehner-Caroli
J
,
Rocken
M
, et al
Analysis of risk factors determining prognosis of cutaneous squamous-cell carcinoma: a prospective study
.
Lancet Oncol
2008
;
9
:
713
20
.
8.
Schmults
CD
,
Karia
PS
,
Carter
JB
,
Han
J
,
Qureshi
AA
. 
Factors predictive of recurrence and death from cutaneous squamous cell carcinoma: a 10-year, single-institution cohort study
.
JAMA Dermatol
2013
;
149
:
541
7
.
9.
Griewank
KG
,
Schilling
B
,
Murali
R
,
Bielefeld
N
,
Schwamborn
M
,
Sucker
A
, et al
TERT promoter mutations are frequent in atypical fibroxanthomas and pleomorphic dermal sarcomas
.
Mod Pathol
2014
;
27
:
502
8
.
10.
Griewank
KG
,
Wiesner
T
,
Murali
R
,
Pischler
C
,
Muller
H
,
Koelsche
C
, et al
Atypical fibroxanthoma and pleomorphic dermal sarcoma harbor frequent NOTCH1/2 and FAT1 mutations and similar DNA copy number alteration profiles
.
Mod Pathol
2018
;
31
:
418
28
.
11.
Alexandrov
LB
,
Kim
J
,
Haradhvala
NJ
,
Huang
MN
,
Tian Ng
AW
,
Wu
Y
, et al
The repertoire of mutational signatures in human cancer
.
Nature
2020
;
578
:
94
101
.
12.
Herling
CD
,
Abedpour
N
,
Weiss
J
,
Schmitt
A
,
Jachimowicz
RD
,
Merkel
O
, et al
Clonal dynamics towards the development of venetoclax resistance in chronic lymphocytic leukemia
.
Nat Commun
2018
;
9
:
727
.
13.
Cun
Y
,
Yang
TP
,
Achter
V
,
Lang
U
,
Peifer
M
. 
Copy-number analysis and inference of subclonal populations in cancer genomes using Sclust
.
Nat Protoc
2018
;
13
:
1488
501
.
14.
Alexandrov
LB
,
Nik-Zainal
S
,
Wedge
DC
,
Aparicio
SA
,
Behjati
S
,
Biankin
AV
, et al
Signatures of mutational processes in human cancer
.
Nature
2013
;
500
:
415
21
.
15.
Salvadores
M
,
Mas-Ponte
D
,
Supek
F
. 
Passenger mutations accurately classify human tumors
.
PLoS Comput Biol
2019
;
15
:
e1006953
.
16.
van der Maaten
L
,
Hinton
G
. 
Visualizing data using t-SNE
.
J Mach Learn Res
2008
;
9
:
2579
605
.
17.
Polak
P
,
Karlic
R
,
Koren
A
,
Thurman
R
,
Sandstrom
R
,
Lawrence
M
, et al
Cell-of-origin chromatin organization shapes the mutational landscape of cancer
.
Nature
2015
;
518
:
360
4
.
18.
Krieger
T
,
Pearson
I
,
Bell
J
,
Doherty
J
,
Robbins
P
. 
Targeted literature review on use of tumor mutational burden status and programmed cell death ligand 1 expression to predict outcomes of checkpoint inhibitor treatment
.
Diagn Pathol
2020
;
15
:
6
.
19.
Klein
S
,
Persa
OD
,
Mauch
C
,
Noh
KW
,
Pappesch
R
,
Wagener-Ryczek
S
, et al
First report on two cases of pleomorphic dermal sarcoma successfully treated with immune checkpoint inhibitors
.
Oncoimmunology
2019
;
8
:
e1665977
.
20.
George
S
,
Miao
D
,
Demetri
GD
,
Adeegbe
D
,
Rodig
SJ
,
Shukla
S
, et al
Loss of PTEN Is associated with resistance to anti-PD-1 checkpoint blockade therapy in metastatic uterine leiomyosarcoma
.
Immunity
2017
;
46
:
197
204
.
21.
Rizvi
H
,
Sanchez-Vega
F
,
La
K
,
Chatila
W
,
Jonsson
P
,
Halpenny
D
, et al
Molecular determinants of response to anti-programmed cell death (PD)-1 and anti-programmed death-ligand 1 (PD-L1) blockade in patients with non-small-cell lung cancer profiled with targeted next-generation sequencing
.
J Clin Oncol
2018
;
36
:
633
41
.
22.
Koyama
S
,
Akbay
EA
,
Li
YY
,
Aref
AR
,
Skoulidis
F
,
Herter-Sprie
GS
, et al
STK11/LKB1 deficiency promotes neutrophil recruitment and proinflammatory cytokine production to suppress T-cell activity in the lung tumor microenvironment
.
Cancer Res
2016
;
76
:
999
1008
.
23.
Tundo
GR
,
Sbardella
D
,
Lacal
PM
,
Graziani
G
,
Marini
S
. 
On the horizon: targeting next-generation immune checkpoints for cancer treatment
.
Chemotherapy
2019
;
64
:
62
80
.
24.
Eilers
G
,
Czaplinski
JT
,
Mayeda
M
,
Bahri
N
,
Tao
D
,
Zhu
M
, et al
CDKN2A/p16 loss implicates CDK4 as a therapeutic target in imatinib-resistant dermatofibrosarcoma protuberans
.
Mol Cancer Ther
2015
;
14
:
1346
53
.
25.
Kusters-Vandevelde
HV
,
Van Leeuwen
A
,
Verdijk
MA
,
de Koning
MN
,
Quint
WG
,
Melchers
WJ
, et al
CDKN2A but not TP53 mutations nor HPV presence predict poor outcome in metastatic squamous cell carcinoma of the skin
.
Int J Cancer
2010
;
126
:
2123
32
.
26.
Pickering
CR
,
Zhou
JH
,
Lee
JJ
,
Drummond
JA
,
Peng
SA
,
Saade
RE
, et al
Mutational landscape of aggressive cutaneous squamous cell carcinoma
.
Clin Cancer Res
2014
;
20
:
6582
92
.
27.
Shain
AH
,
Yeh
I
,
Kovalyshyn
I
,
Sriharan
A
,
Talevich
E
,
Gagnon
A
, et al
The genetic evolution of melanoma from precursor lesions
.
N Engl J Med
2015
;
373
:
1926
36
.
28.
Shah
A
,
Bloomquist
E
,
Tang
S
,
Fu
W
,
Bi
Y
,
Liu
Q
, et al
FDA approval: ribociclib for the treatment of postmenopausal women with hormone receptor-positive, HER2-negative advanced or metastatic breast cancer
.
Clin Cancer Res
2018
;
24
:
2999
3004
.
29.
Sledge
GW
 Jr
,
Toi
M
,
Neven
P
,
Sohn
J
,
Inoue
K
,
Pivot
X
, et al
MONARCH 2: abemaciclib in combination with fulvestrant in women with HR+/HER2- advanced breast cancer who had progressed while receiving endocrine therapy
.
J Clin Oncol
2017
;
35
:
2875
84
.
30.
Turner
NC
,
Ro
J
,
Andre
F
,
Loi
S
,
Verma
S
,
Iwata
H
, et al
Palbociclib in hormone-receptor-positive advanced breast cancer
.
N Engl J Med
2015
;
373
:
209
19
.
31.
Nishida
T
,
Doi
T
,
Naito
Y
. 
Tyrosine kinase inhibitors in the treatment of unresectable or metastatic gastrointestinal stromal tumors
.
Expert Opin Pharmacother
2014
;
15
:
1979
89
.
32.
Metzgeroth
G
,
Walz
C
,
Erben
P
,
Popp
H
,
Schmitt-Graeff
A
,
Haferlach
C
, et al
Safety and efficacy of imatinib in chronic eosinophilic leukaemia and hypereosinophilic syndrome: a phase-II study
.
Br J Haematol
2008
;
143
:
707
15
.
33.
Rutkowski
P
,
Klimczak
A
,
Lugowska
I
,
Jagielska
B
,
Wagrodzki
M
,
Debiec-Rychter
M
, et al
Long-term results of treatment of advanced dermatofibrosarcoma protuberans (DFSP) with imatinib mesylate - The impact of fibrosarcomatous transformation
.
Eur J Surg Oncol
2017
;
43
:
1134
41
.
34.
Ugurel
S
,
Mentzel
T
,
Utikal
J
,
Helmbold
P
,
Mohr
P
,
Pfohler
C
, et al
Neoadjuvant imatinib in advanced primary or locally recurrent dermatofibrosarcoma protuberans: a multicenter phase II DeCOG trial with long-term follow-up
.
Clin Cancer Res
2014
;
20
:
499
510
.
35.
Fu
Y
,
Kang
H
,
Zhao
H
,
Hu
J
,
Zhang
H
,
Li
X
, et al
Sunitinib for patients with locally advanced or distantly metastatic dermatofibrosarcoma protuberans but resistant to imatinib
.
Int J Clin Exp Med
2015
;
8
:
8288
94
.
36.
Lachmann
A
,
Torre
D
,
Keenan
AB
,
Jagodnik
KM
,
Lee
HJ
,
Wang
L
, et al
Massive mining of publicly available RNA-seq data from human and mouse
.
Nat Commun
2018
;
9
:
1366
.
37.
Kuleshov
MV
,
Jones
MR
,
Rouillard
AD
,
Fernandez
NF
,
Duan
Q
,
Wang
Z
, et al
Enrichr: a comprehensive gene set enrichment analysis web server 2016 update
.
Nucleic Acids Res
2016
;
44
:
W90
7
.