Spatial transcriptomics of a giant pilomatricoma

Pilomatricomas (PMs) are common benign adnexal tumors that show a predilection for the head and neck region and are characterized at the molecular level by activating mutations in the beta‐catenin (CTNNB1) gene. Giant PMs are a rare histopathological variant, according to the World Health Organization, which are defined by a size greater than 4 cm and are reported to show upregulation of yes‐associated protein compared to PMs of typical 1–3 cm size. We describe the case of a 67‐year‐old man with an 8 cm giant PM involving his temporal scalp, whose PM we characterized by 10X spatial gene expression analysis. This revealed five total transcriptomic clusters, including four distinct clusters within the giant PM, each with a unique transcriptional pattern of hair follicle‐related factors, keratin gene expression, and beta‐catenin pathway activity.


| INTRODUCTION
Pilomatricomas (PMs) are common benign adnexal tumors that arise from hair follicle matrix cells. 1 Their age of incidence is bimodal, with one peak in childhood and early adolescence and a second peak in later adulthood. 2PMs typically present as hard solitary nodules measuring 1-3 cm in the head and neck region, although the extremities and trunk are also common sites. 3Overlying skin may appear normal or show red or bluish discoloration with ulceration.Giant PMs are a rare World Health Organization (WHO)-acknowledged PM variant that is greater than 4 cm in size, frequently ulcerate and are more likely to simulate a malignant neoplasm on clinical exam. 4st PMs harbor mutations in the CTNNB1 gene 5 leading to constitutive beta-catenin/lymphoid enhancer binding factor (LEF) target gene activation to control proliferation, differentiation, and survival. 6,7multaneous gain of chromosome 18 harboring the BCL2 apoptosis regulator (BCL2) gene is also a recurrent abnormality accompanied by BCL2 protein overexpression. 8,9Finally, one study has shown a positive correlation between yes-associated protein (YAP) expression and PM size. 10atial gene expression analyses allow for the measurement of gene expression while retaining spatial information, enabling visualization and statistical comparisons of gene expression differences across tissue regions. 11,12Here, we report 10X Genomics Visium Spatial Gene Expression analysis of a giant PM.As part of a larger institutional review board-approved study, we subjected a representative region of the patient's giant PM (Figure 2A) to 10X Genomics Visium Spatial Gene Expression analysis using the Space Ranger v1.3.1 data pipeline. 12We visualized and analyzed the resulting data in the 10X Genomics Loupe Browser v6.0.0 and applied t-distributed stochastic neighbor embedding (t-SNE) 13,14 which revealed five statistically distinct transcriptomic clusters (Figure 2B,C), one of which (Cluster 5) constituted benign   4.6x10 -9 2.3x10 -2 1.4x10 -2 3.1x10 -2 2.7x10 -10 1.9x10 -6 F I G U R E 3 Gene enrichment within spatial gene expression clusters.(A) The most significantly enriched genes within each cluster.(B) Total number of genes showing significantly higher or lower transcript number in each clusters when compared to all other clusters.Box indicates clusters used for all remaining analyses of tumor spatial transcriptomics, in which Cluster 5 representing stromal elements was excluded.(C-E) Gene ontology (GO) enrichment analysis for biological processes applied to the top 250 most significantly highly expressed genes from each cluster, for which enough significantly expressed transcripts were available to analyze.stroma.The remaining four clusters (Clusters 1-4) represented giant PM and followed the characteristic maturation pattern from basaloid component resembling hair matrix cells toward heavily keratinized layers lacking cell nuclei but still harboring detectable mRNA transcripts (Figure 2D).

| CASE REPORT
The Loupe Browser's gene comparison tool enabled the identification of transcripts that were significantly elevated or reduced within each cluster (Figure 3A,B).Statistical comparisons were performed using either the negative binomial exact test (Sseq) 15 or a fast asymptotic beta test (edgeR), 16

Clusters 1 2 3 4
F I G U R E 4 Expression patterns of genes implicated in hair follicle biology and pilomatricoma pathogenesis.(A) The expression level of select transcripts related to hair follicle biology and pilomatricoma pathogenesis across pilomatricoma spatial transcriptomic clusters.Transcripts from a subset of genes were not detected in sufficient numbers to enable spatial comparisons.(B) Spatial overlays of relative transcript quantity for select genes are plotted in panel (A).N.S., not statistically significant.
5 represented the benign stromal elements and unsurprisingly had a unique transcriptional signature, with elevated transcripts associated with inflammation and lymphocyte biology.However, our analysis was limited to comparisons of transcript levels between clusters, and we determined that drawing conclusions about stromal gene signatures via comparisons to giant PM gene expression was not informative.Instead, we excluded Cluster 5 and focused the remaining analyses on gene expression differences between giant PM clusters (Figure 3B).
Each of the giant PM expression clusters harbored unique highly expressed gene sets (Figure 3B, boxed region).Clusters 1, 2, and 4 also showed sets of transcripts with significantly reduced expression.As most clusters harbored thousands of highly expressed transcripts, we used gene ontology (GO) analysis [17][18][19] to parse the top 250 most highly expressed transcripts from Clusters 1, 2, and 4 to identify the top 10 most upregulated biological processes (Figure 3C-E).This was not possible for Cluster 3, which harbored few significantly expressed transcripts.Cluster 1, corresponding to the deep basal cell layer, was enriched in processes related to DNA replication and cell division (Figure 3C).However, we also detected highly expressed genes related to the process of salivary gland cavitation, which is the developmental process by which salivary gland ducts form lumens. 20 Cluster 2 corresponded to the layers of cells immediately superficial to the basal layer and was enriched in processes related to cell-cell adhesion and communication, as well as cell cornification and establishment of the skin barrier (Figure 3D).Cluster 4 corresponded to enucleated "ghosted" keratin layers, and only six significantly enriched processes were identified, which included intermediate filament organization, keratinization, the hair cycle, and negative regulation of endopeptidase activity (Figure 3E).
As PMs are hypothesized to arise from hair follicle matrix cells, we analyzed a panel of genes related to hair follicle development as summarized in recent review articles (Figure 4), 21,22 excluding keratin transcripts, which were analyzed separately.These include genes implicated in PM pathogenesis, including CTNNB1 and LEF1, and select targets of their regulation. 5Cluster 1 was enriched for sonic hedgehog and negative regulators of Wnt signaling, including wnt inhibitory factor-1 (WIF1), Dickkopf Wnt signaling pathway inhibitor 4 (DKK4), TLE family member 2 (TLE2), and naked-1 (NKD1).There was a trend toward elevated expression of CTNNB1 in Cluster 1, but this was not statistically significant.Cluster 2 showed reduced expression of Wnt pathway inhibitors such as WIF1 and DKK4, although transcripts for other Wnt pathway inhibitors such as NKD2 became more prevalent.This transition to Cluster 2 corresponded with elevated transcript levels of Wnt/beta-catenin target genes MYC, LEF1, AXIN2, and bone morphogenetic protein 4 (BMP4).Unfortunately, while YAP expression is reportedly elevated in giant PMs, 10 we were unable to detect sufficient YAP transcripts to enable spatial comparisons.

| DISCUSSION
We report here the application of 10X Genomics Visium Spatial Gene Expression analysis to a giant PM, which identifies unique clusters of spatial gene signatures.These clusters largely overlap with distinct layers of pilar maturation, although Cluster 3 harbors few unique transcripts and is more challenging to definitively classify.Given that many uniquely regulated transcripts within Cluster 3 are keratins, it may reflect a transient intermediary stage during which factors mediating cell proliferation are lost prior to terminal keratinization.
Our report does have several shortcomings.For one, our report lacks data confirming that identified transcriptional differences carry through to the protein level.It would thus be interesting to perform a follow-on study utilizing IHC to measure the levels and distribution of target factors identified in our study, if possible, within a larger cohort of giant and typical PMs.Second, a component of the t-SNE algorithm is stochastic, and thus, there is some variability in output between reruns.We did not analyze other regions of our giant PM or perform similar analyses in PMs from other patients, but both could lead to increased confidence in our findings in addition to potentially identifying additional pathways regulating PM behavior.
The molecular drivers of cutaneous adnexal tumor development are poorly understood, and our study sheds light on the pathobiology of giant PM.These benign tumors are often surgically challenging, and rare malignant counterparts can behave aggressively.An improved understanding of the pathobiology of these tumors may inform future treatment options, and our study serves as a model that may be applied to better understand the pathophysiology of benign and malignant adnexal tumors.

AF
67-year-old man presented with a right temporal scalp mass that had been enlarging over the last 6 years.The mass had recently bled extensively after the patient accidentally traumatized it while combing his hair, resulting in a visit to the local emergency room.A physical exam revealed a soft, violaceous, subcutaneous mass with thin but intact overlying skin.CT imaging revealed an 8 cm mass limited to the soft tissues of the scalp.Complete excision was performed with flap reconstruction.Gross examination showed a circumscribed, white-colored mass abutting the deep surgical margin.Microscopic sections revealed solid nests of basaloid cells within the dermis undergoing abrupt keratinization, transitioning into ghost cells and loose keratin aggregates, with focal transepidermal perforation (Figure 1A,B).No invasive growth and no significant cytologic atypia were seen.Immunohistochemistry (IHC) performed for beta-catenin confirmed strong and diffuse nuclear expression in the tumor cells (Figure 1C,D).A diagnosis of PM was rendered, and the inked margins were negative for the tumor.Given the 8 cm size, the tumor met the criteria for a giant PM per the WHO.The patient experienced a complete recovery without tumor recurrence.I G U R E 1 Histopathology and beta-catenin immunohistochemistry (IHC) of giant pilomatricoma.(A, B) Low-and high-power images showing sections of pilomatricoma demonstrating characteristic non-invasive basal cell layer with an abrupt transition to enucleated keratin layers.Hematoxylin and eosin (A) Â200 and (B) Â400.(C, D) Low-and high-power images show staining the pattern of beta-catenin IHC, confirming strong and diffuse staining throughout, which is mostly localized to the nucleus and, to a lesser extent, the plasma membrane.DAB (C) Â200 and (D) Â600.

F
I G U R E 2 Localization of unique gene expression clusters within giant pilomatricoma.(A) Image of region selected for spatial gene expression profiling.Hematoxylin and eosin Â400.(B) Overlay of tissue regions corresponding to clusters.(C) t-distributed stochastic neighbor embedding (t-SNE) visualization of pilomatricoma spatial gene expression clusters.The transcriptomic data within each of the circular regions overlaid onto the tissue in panel (B) is visualized as a single point in two dimensions.Points with similar features tend to cluster near each other.(D) Solid-fill visualization of points from panel (C) with labels indicating select tissue types and an arrow showing maturation direction from basal layer to superficial ghosted keratin.
Patil, Jin Xu, Paul Weisman, and Daniel R. Matson collected and analyzed the data.Apoorva T. Patil, Daniel D. Bennett, and Daniel R. Matson prepared and edited the manuscript.ACKNOWLEDGMENTS Daniel R. Matson has received support from the NIH (T32HL007899 and K08DK127244) and The Hartwell Foundation.Additional funding for this work was provided by the University of Wisconsin-Madison Department of Pathology and Laboratory Medicine.The authors thank the University of Wisconsin Translational Research Initiatives in Pathology laboratory (TRIP), supported by the UW Department of Pathology and Laboratory Medicine, UWCCC (P30 CA014520), and the Office of The Director-NIH (S10 OD023526) for use of its facilities and services.