2.1 Samples

Tissue samples from M. coucha were taken from our tissue collection obtained in previous studies,^{14, 19-21} where skin biopsies and tumors were cut longitudinally with scalpels, fixed in 4% formalin, and embedded in paraffin (FFPE). M. coucha are naturally infected with the cutaneous MnPV, which like HPVs belongs to the family Papillomaviridae. The animals were chronically exposed to UV radiation (312 nm) to induce cutaneous SCCs with different histopathological properties in virus-positive skin, namely, well-differentiated keratinizing SCCs (KSCC) and poorly differentiated non-keratinizing SCCs (nKSCC).¹⁴

2.2 Sample collection and lysis

FFPE material was cut into 8 µm (for microdissection) or 3 µm (for tissue stainings) thick sections and deparaffinized by submersion in 100% xylene (5 min, twice), 100% ethanol (10 min), 90% ethanol (5 min), 70% ethanol (5 min), and phosphate-buffered saline (PBS). The target tissue was microdissected under a microscope using a 25G cannula based on previously hematoxylin and eosin (HE) or immunostained consecutive sections. A new cannula was used for each sample to exclude possible contamination. The tissue was not allowed to dry out completely. Microdissected tissue was directly transferred to a low binding tube (Eppendorf) with 25 μL lysis buffer (100 mM tris(hydroxymethyl)aminomethan [Tris] pH 8.0, 100 mM dithiothreitol [DTT], 4% sodium dodecyl sulfate [SDS]) and vortexed for 5 s before heating at 100°C for 20 min. The tube was centrifuged for 1 min at 1000g for tissue sedimentation and vortexed again for 5 s. Then, the sample was incubated at 60°C for 2 h with regular vortexing (5 s) and centrifugation (1 min at 1000g) every 30 min. The tube was centrifuged for 10 min at 15,000g (4°C) to pellet cellular debris. The supernatant was transferred to a fresh 0.5 mL tube and stored at −80°C or directly subjected to further sample preparation and SP3 purification.

2.3 Sample preparation for proteome profiling

Samples were thawed, vortexed, and subsequently transferred to adaptive focused acoustics (AFA)-TUBE TPX polymerase chain reaction (PCR) stripes (Covaris, Inc.) for AFA ultrasonication in a LE220plus Covaris device. Here, the peak incidence power (PIP) was set to 450, the duty factor (DF) to 50%, the cycles per burst (CPB) to 600, and the time to 300 s per TPX PCR stripe. Dithering of the AFA focus was applied with a 3 mm z-offset. Samples were centrifuged at 15,000g for 10 min before proceeding with a protein quantification assay (Pierce, ThermoFisherScientific). For each sample, 10 μg of extracted protein was transferred to a 96-well PCR plate for processing with autoSP3 on an Agilent Bravo liquid handling system.¹⁷ In brief, the plate was prepared with each sample in a total volume of 12 μL 1% SDS, 100 mM ammonium bicarbonate (ABC). Proteins were automatically reduced and alkylated by the addition of 10 mM tris(2-carboxyethyl)phosphine (TCEP), 40 mM chloroacetamide (CAA), 1× protease inhibitor cocktail (PIC) in 100 mM ABC, and incubation for 5 min at 95°C. Subsequently, proteins were immobilized on paramagnetic, carboxylate-modified SP3 beads by establishing a > 50% organic environment with acetonitrile (ACN). Bead:protein conjugates were extensively washed twice with 200 μL 80% ethanol and once with 100% ACN. Proteins were digested with trypsin (1:60 protease-to-protein ratio) for 16 h in 100 mM ABC. Upon overnight digestion, the reaction was quenched by acidification to 0.5% trifluoroacetic acid (TFA). The peptide-containing supernatant was recovered to a new 96-well plate without the transfer of residual beads. MS injection-ready samples were stored at −20°C until data acquisition.

2.4 MS

Liquid chromatography with tandem MS (LC-MS/MS) analyses were carried out on an Ultimate 3000 UPLC system (ThermoFisherScientific) directly connected to an Orbitrap Exploris 480 MS for a total of 120 min. Dried peptides were reconstituted before measurement and online desalted on a trapping cartridge (Acclaim PepMap300 C18, 5 µm, 300 Å wide pore; ThermoFisherScientific) for 3 min using 30 µL/min flow of 0.05% TFA in water. The analytical multistep gradient (300 nL/min) was performed using a nanoEase MZ Peptide analytical column (300 Å, 1.7 µm, 75 µm × 200 mm, Waters) using solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in ACN). For 102 min, the concentration of B was linearly ramped from 4% to 30%, followed by a quick ramp to 78%, after 2 min the concentration of B was lowered to 2% and a 10 min equilibration step appended. Eluting peptides were analyzed in the MS using data-depend acquisition (DDA) mode. A full scan at 120k resolution (380–1400 m/z, 300% AGC target, 45 ms maxIT) was followed by up to 2 s of MS/MS scans. Peptide features were isolated with a window of 1.4 m/z, fragmented using 26% NCE. Fragment spectra were recorded at 15k resolution (100% AGC target, 54 ms maxIT). Unassigned and singly charged eluting features were excluded from fragmentation and dynamic exclusion was set to 35 s.

2.5 Analysis of MS data

Data analysis was carried out by MaxQuant (version 1.6.14.0) (www.coxdocs.org)²² using databases extracted from www.uniprot.org under default settings (Mus musculus containing 55 341 entries from January 03, 2022, M. coucha 55 entries from September 06, 2022; MnPV sequences with seven entries). Identification false discovery rate (FDR) cutoffs were 0.01 on peptide level and 0.01 on protein level. The match between runs (MBR) option was disabled. Label-free quantification (LFQ) was done using an approach based on the MaxLFQ algorithm.²³ A minimum of two quantified peptides per protein was required for protein quantification. MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE²⁴ partner repository with the data set identifier PXD042196.

2.6 Statistics

The following filtering, normalization, and imputation were performed individually for each statistical contrast. Adapted from the Perseus recommendations²⁵ protein groups with valid LFQ values in 70% of the samples of at least one condition were used for statistics. For the LFQ values, no further normalization was performed. Adapted from the Perseus recommendations,²⁵ missing LFQ values being completely absent in one condition, were imputed with random values drawn from a downshifted (2.2 standard deviation) and narrowed (0.3 standard deviation) intensity distribution of the individual samples. Despite its benefits in the contrast wise comparisons, the design of the imputation setup with its condition-specific elements might lead to artificially increased distances in between the conditions. For missing values with no complete absence in one condition, the R package missForest was used for imputation.²⁶ The statistical analysis was performed with the R-package “limma.”²⁷ The p-values were adjusted with the Benjamini–Hochberg method for the multiple testing.²⁸ For the principal component analysis (PCA) plots, the filtering, normalization, and imputation as described above were applied to the whole set of samples of the HE staining and E1^E4 staining set, respectively. The sample positions in the two-dimensional grid were calculated via the “plotMDS” function of the R-package “limma.”

2.7 Network and functional analyses

Regulatory networks and functional analyses were generated using ingenuity pathway analyses (IPA) software (Qiagen Inc.). A “core analysis” was conducted using all data sources available in IPA. All networks and node types were selected to be displayed. To specify results, only mouse species-related entries were used for analyses. A z-score of ±2.0 as a cut-off for a reliable prediction was automatically set by IPA.²⁹ Protein networks were visualized with STRING database (www.string-db.org).³⁰ Genes and proteins were named based on the standard nomenclature for describing mouse genes and proteins.³¹

2.8 Tissue stainings

FFPE tissue sections were deparaffinized as described above for subsequent stainings. HE staining was performed using a standard protocol.³² For immunostainings heat-induced epitope retrieval (HIER) was achieved by boiling the sections in a steamer for 10 min in citrate buffer pH 6.0 or ethylenediaminetetraacetic acid (EDTA) buffer pH 9.0. Sections were blocked with 5% goat serum and incubated overnight at 4°C with primary antibodies against keratin 10 (anti-K10; BioLegend: PRB-159P; rabbit immunoglobulin (IgG), 1:1000), keratin 14 (anti-K14; BioLegend: PRB-155P; rabbit IgG, 1:1000), keratin 15 (anti-K15; Invitrogen: MA5-11344; mouse IgG2a, 1:200), keratin 17 (anti-K17-Alexa488; BioLegend: 697207; rat IgG2b, 1:1000), E-cadherin (BD: 610181; mouse IgG, 1:500), vimentin (anti-VIM; Cell Signaling: #5741; rabbit IgG, 1:100), collagen alpha-1(I) chain (anti-COL1A1; Cell Signaling: #72026; rabbit IgG, 1:500), caspase-14 (anti-CASP14 (bio-techne: NBP2-33906; rabbit IgG, 1:2000), CD271 (cell signaling: #8238; rabbit IgG, 1:250), proliferating cell nuclear antigen (anti-PCNA; Cell Signaling: #13110; rabbit IgG, 1:4000), MnPV-E1^E4 (self-made antibody, mouse IgG2b, 1:100), MnPV-L1 (self-made antibody, Mastomys monoclonal IgG 1:175). After washing, slides were incubated with Alexa488- or Alexa594-conjugated anti-mouse or anti-rabbit (Invitrogen; 1:2000) or anti-rat IgG2b (BioLegend: 408210; 1:400) antibodies and nuclei were stained with DAPI. Sections were mounted with Dako Faramount Aqueous Mounting Medium (Dako North America, Inc) and imaged with a Keyence BZ-9000 Microscope (Keyence Deutschland GmbH) or a Hamamatsu NanoZoomer S60 (Hamamatsu).

3.1 Establishing a spatial proteomic workflow for skin tissue

Based on HE staining, every FFPE nKSCC sample obtained from chronically UV-irradiated M. coucha¹⁴ was microdissected into well-differentiated and poorly differentiated tumor areas (Figure 1A). For complementary analyses, also epidermis from adjacent phenotypically normal skin — either uninfected or infected with MnPV — and benign MnPV-induced tumors were included (Figure 1A). For the latter, microdissection was based on positive staining on consecutive sections against the MnPV E1^E4 protein, translated from the most abundant viral transcript (Figure 1B). To avoid potential contamination by different keratins abundantly present, especially in outer epithelial layers of MnPV-induced tumors, epidermal parts were carefully separated from the keratinized material (Figure 1C) before lysis.

In preceding calibration tests, we realized that a minimum tissue volume of approximately 0.15 mm³ is required to obtain satisfactory MS data. Since high SDS concentrations of the lysis buffer would interfere with MS measurements, lysates were purified via SP3 paramagnetic bead technology (outlined in Figure 1D) that facilitates ultrasensitive proteomic analyses of protein mixtures.^{17, 18} Purified samples were further digested and resulting peptides were separated and examined in a mass spectrometer. Resulting MS spectra were identified and quantified using an LFQ approach based on the MaxQuant software²³ and mouse-specific databases extracted from uniprot.org.

Using this workflow, the following questions were addressed: (i) what are the characteristics of the tumor phenotypes with special focus on well-differentiated and poorly differentiated areas of UV-induced tumors, and (ii) which changes of the protein pattern are induced by MnPV when uninfected and infected epidermis as well as MnPV-induced tumors are compared?

3.2 Spatial MS analysis of tumor tissue reveals strong intra- and intertumoral heterogeneity

To assess the similarity between the proteomes of in total 25 samples (i.e. control epidermis, MnPV-induced benign tumors, well- and dedifferentiated areas of UV-induced SCCs), microdissection was based on HE stainings. The resulting MS data covered 3752 proteins, of which 866 proteins were shared between all samples. Intriguingly, subjecting these proteome data to a PCA (Figure 2A), the individual samples clustered in strikingly distinct areas. This clearly shows that the overall resemblance of the individual proteomes deriving from the same tissue type is principally higher than to those of other tissue origins. Spatial proximity revealed that proteomes of well-differentiated tumor areas still preserved a higher similarity to morphologically normal epidermis, while proteomes obtained from MnPV-induced tumors and, most prominently, dedifferentiated areas of SCCs were more distinct (Figure 2A). As depicted in the Venn diagram (Figure 2B) based on the proteins shared between all samples, the high similarity between phenotypically normal epidermis and well-differentiated SCC areas is reflected by a low number ( = 97 proteins; see Figure 2B, red circle) of significantly (adj.p < 0.05) differentially expressed proteins. Conversely, dedifferentiated SCC areas ( = 240 proteins) and MnPV-induced tumors ( = 571 proteins) revealed substantially more differences from normal epidermis. Of note, these differentially abundant proteins were not equally altered, but each comparison revealed prominent outliers (Figure 2C). Some of those, for instance, keratin 10 (K10; log2FC = –6.41, adj.p < 0.001) and K2 (log2FC = –9.71, adj.p < 0.001), marking postmitotic suprabasal and spinous keratinocytes, respectively, are exceedingly changed in dedifferentiated tissue when compared with normal epidermis (Figure 2D). All significantly changed proteins are summarized in Supporting Information: STables 1–6.

Phenotypic alterations microscopically seen after HE staining (Figure 1A), point to an EMT which is characterized by a switch of E-cadherin to N-cadherin expression.³³ However, N-cadherin (expressed from the Cdh2 gene) could not be detected in any tissue and E-cadherin (expressed from the Cdh1 gene) was only found in one comparison at all (MnPV-induced tumor versus dedifferentiated SCC; log2FC = 2.87, adj.p < 0.001). Nonetheless, the loss of E-cadherin is accompanied by decreased expression of keratins, considered as epithelial differentiation markers.^{34, 35} When microscopically dedifferentiated areas were compared with well-differentiated areas, 10 significantly reduced keratins (K1, 2, 5, 6a, 10, 14, 16, 17, 71, and 79; log2FC between –1.73 and –6.30) were found. Additionally, compared with normal epidermis, the mesenchymal marker vimentin (VIM) was tendentially upregulated in well-differentiated (log2FC = 0.72, adj.p = 0.89) and in dedifferentiated SCC areas (log2FC = 1.56, adj.p = 0.064). Also linked with loss of E-cadherin during an EMT, an increase of VIM confers migratory abilities to cancer cells^{34, 35} An EMT could lead to an increase of cancer stem cells (CSCs)³⁶ that are characterized by coexpression of major hyaluronan (HA) receptor CD44 and nerve growth factor (NGF) receptor CD271 (also known as p75^NTR).³⁷ While the latter was not detected in any tissue type analyzed, CD44 was significantly decreased in MnPV-induced tumors (log2FC = –0.79, adj.p = 0.038) when compared with normal epidermis. Conversely, the CD44 levels were tendentially elevated in well-differentiated SCC areas (log2FC = 0.55, adj.p = 0.72) and significantly enriched in dedifferentiated SCC areas (log2FC = 0.96, adj.p = 0.036) when compared with normal epidermis. The difference between well-differentiated and dedifferentiated SCC areas was statistically not significant (log2FC = 0.42, adj.p = 0.56).

These data indicate that the proteome of well-differentiated SCC tissue is relatively similar to morphologically normal epidermis, while MnPV-induced benign tumors and dedifferentiated SCC areas are rather distinct. Furthermore, phenotypic changes of well-differentiated and dedifferentiated areas of SCCs, among others, are indeed based on a dedifferentiation of altered tumor cells during EMT and the formation of a CSC subpopulation. Consequently, well-differentiated SCC could be considered an intermediate step between normal epidermis and dedifferentiated SCCs.

3.3 In-depth IPA analyses reveal differential enrichment of signaling pathways

Next, IPAs were performed to get insight into the biological context of the obtained datasets. Considering the top 50 enriched diseases and bio functions, several were linked to mechanisms involved in enhanced cell survival and decreased cell death. Figure 3A shows a selection of 30 diseases and bio functions ordered by relevance based on a z-score calculated by IPA.²⁹ In fact, cell death-associated processes belonging to the categories “Cancer, Organismal Injury and Abnormalities,” such as necrosis and apoptosis were deactivated (Figure 3A, see also Supporting Information: STables 7-13). Conversely, cell viability and cell survival were activated, especially when MnPV-induced tumors were compared with morphologically normal epidermis (Figure 3A).

Analyses of canonical signaling pathways (Figure 3B, see also Supporting Information: STables 14-20) additionally showed deactivated CXCR4 signaling in MnPV-induced benign tumors, contrasting its activation in dedifferentiated SCC areas. The latter is in line with reports suggesting increased CXCR4 signaling to initiate cell proliferation and outgrowth of colon carcinoma micro-metastases, respectively.^{38, 39} Moreover, spliceosomal cycle and transfer RNA (tRNA) charging are strongly activated in MnPV-induced tumors, suggesting an intracellular permissive environment with high loads of transcriptionally active viral genomes, the constraint to process polycistronic viral transcripts and to produce virion progeny. In contrast, most other pathways (e.g., gonadotropin-releasing hormone [GNRH] signaling and mammalian target of rapamycin [mTOR] signaling) are deactivated when compared with normal epidermis, while well-differentiated SCC areas differ only with respect to the downregulated Oxidative Phosphorylation and Xenobiotic Metabolism PXR Signaling pathway (Figure 3B), indicating intertumoral heterogeneity. Strikingly, the opposite situation, namely, an activation of nearly all top canonical pathways in dedifferentiated SCC areas further reflects intratumoral heterogeneity. However, differences in (de)activation of various pathways finally led to distinct clustering of the different tissue types in the PCA (Figure 2A).

Interestingly, although peptides derived from NGF receptor CD271 were not detected by MS profiling, IPA analyses predicted an upregulation of NGF signaling in dedifferentiated SCC areas (Figure 3B) that may favor the maintenance of stem-like properties and tumorigenicity.⁴⁰ Of potential interest, NGF signaling is linked to several enriched bio functions (e.g., metastasis and growth of tumor) (Figure 3A).⁴¹ Consistent with this assumption, phosphatase and tensin homolog (PTEN) PI3K/AKT signaling and mTOR pathway were—compared with morphologically normal epidermis—activated in dedifferentiated SCCs areas, thereby promoting tumor cell growth, metastasis and invasion to new healthy tissues.^{42, 43} In line with this, pathways related to EMT (e.g., ephrin receptor signaling) and metastasis (e.g., colorectal cancer metastasis signaling) were also predicted to be activated. Conversely, these pathways are deactivated in MnPV-induced tumors (Figure 3B).

3.4 Activity of upstream regulators depends on tissue origin

Next, upstream regulator analyses were performed with IPA to identify effector molecules that may explain the up- or downregulation of proteins in the MS datasets (Figure 3C, see also Supporting Information: STables 21-27). Generally, well-known functional regulators associated with cancer (e.g., TP53) were obtained.

In our system, the predicted activity of the gene product of Nfe2l2 (NRF2, nuclear factor erythroid 2-related factor 2) was strongly increased in MnPV-induced benign tumors when compared with normal epidermis (z-score = 4.596). Intriguingly, NRF2 activation increases the expression of cytoprotective proteins, thereby contributing to DNA damage prevention.⁴⁴ Moreover, comparing dedifferentiated SCC areas with normal epidermis, IPA also predicted lower activity of RB1 (Retinoblastoma 1, z-score = -2.137) and BCL6 (B-cell lymphoma 6; z-score = -2.646) while STAT1 (Signal transducer and activator of transcription 1; z-score = 2.377) activity was increased. This may underscore the special importance of such factors during the dedifferentiation process, since these changes were not detected when comparing other tissue areas.

Another interesting upstream regulator, DEC1 (“Differentiated embryo-chondrocyte expressed gene 1”) was predicted to be activated in dedifferentiated tumor tissue when compared with morphologically normal epidermis (z-score = 3.130). Transcription regulator DEC1 stimulates PI3K/AKT/mTOR activation and causes a prosurvival and prometastatic phenotype in breast cancer.^{45, 46} In dedifferentiated SCC areas, its activation may explain increased mTOR signaling (Figure 3B).

Previous studies have shown that MAP4K4 promotes pancreatic cancer⁴⁷ and acts as an upstream regulator of the ERK/MAPK pathway required for adenocarcinoma maintenance.⁴⁸ Consistently, the predicted activity of MAP4K4 in our comparisons (Figure 3C) correlates with (de)activation of ERK/MAPK signaling in canonical pathway analyses (Figure 3B).

With respect to the human TP53 gene (or its murine and Mastomys homolog Trp53), IPA predicted that its regulatory activity affects 24 differentially expressed proteins in well-differentiated but 66 proteins in dedifferentiated SCC areas when compared with phenotypically normal epidermis (Supporting Information: STables 22 and 23). This is in line with the previous finding that Trp53 mutations—potentially altering protein interactions—are more frequent in poorly differentiated nKSCCs than in well-differentiated KSCCs.¹⁴ Consequently, reflecting tumor heterogeneity on a molecular level, the protein networks obviously differ when visualizing the connections of these hits (Figure 3D) using STRING database.³⁰

Taken together, these analyses identified pathways and regulators involved in the multistep process of skin carcinogenesis and SCC dedifferentiation.

3.5 Validation of proteomic data using tissue stainings

To verify the proteomic profiles, immunostainings against up/downregulated proteins (Figure 2D) or those related to differentiation and viral infection were performed (Figure 4). In agreement with the MS data, these stainings showed a significant increase of viral infection markers (i.e., E1^E4, log2FC = 9.96, adj.p < 0.001 and L1, log2FC = 10.07, adj.p < 0.001) in MnPV-induced tumors when compared with phenotypically normal epidermis. Here, confirming the proteomics, no visible change in K14 (log2FC = -0.47, adj.p = 0.47), but diminished levels of K10 (log2FC = –1.92, adj.p = 0.025) and K15 (log2FC = –2.18, adj.p = 0.013) could be detected. K15 could be visualized in progenitor basal cells in hair follicles, which were found more frequently in normal epidermis than in tumor areas including MnPV-induced tumors.

The staining pattern of well-differentiated areas of SCCs and normal epidermis were relatively similar and no significant differences in levels of K14, K17, VIM, caspase-14 (CASP14) and collagen alpha-1(I) chain (COL1A1) could be discerned by MS. Despite being weaker expressed when compared with normal epidermis, K15 levels in well-differentiated SCC areas (log2FC = –1.11, adj.p = 0.27) were not as low as in MnPV-induced tumors (log2FC = –2.18, adj.p = 0.013). Conversely, in UV-induced SCCs, K10 (log2FC = -3.68, adj.p = 0.023) was only stained in keratinized layers, which were radically removed by microdissection before MS analyses (Figure 1C). Thus, staining intensity of these outer cornified layers did not correlate with low K10 levels within the tumor cells.

Dedifferentiated SCC areas showed a contrary pattern, since all stained keratins were significantly reduced compared with normal epidermis (log2FC between –1.77 and –6.41), while VIM (log2FC = 1.56, adj.p = 0.064) was positively stained, but could not be visualized in normal epidermis.

Comparing dedifferentiated and well-differentiated SCC areas, stainings further confirmed the reduction of K10 (log2FC = –2.70, adj.p = 0.009), K14 (log2FC = –3.06, adj.p = 0.001), and K17 (log2FC = –3.22, adj.p = 0.009), respectively. Despite insignificant MS data regarding VIM (log2FC = 0.84, adj.p = 0.59) and COL1A1 (log2FC = –1.59, adj.p = 0.31) (not shown), tissue stainings showed that both proteins were upregulated in dedifferentiated SCC areas. While E-cadherin was not detected by MS in most tissues, immunostaining clearly detected its loss in dedifferentiated SCC areas when compared with MnPV-induced tumors (log2FC = –2.87, adj.p < 0.001) or well-differentiated SCCs areas.

Of note, consistent with upregulated NGF signaling in dedifferentiated SCC areas, the stainings demonstrated islands of cells strongly positive for NGF receptor CD271 that is considered to favor the maintenance of stem-like properties and tumorigenicity.⁴⁰ Conversely, corresponding stainings were not found in well-differentiated SCC areas or normal epidermis, where only basal layers showed slight CD271 positivity. Similarly, MS data revealed a slight increase of CD44 in well-differentiated (log2FC = 0.55, adj.p = 0.72) and a significant enrichment in dedifferentiated SCC areas (log2FC = 0.96, adj.p = 0.036), when compared with normal epidermis. The coincidence of CD44 MS data and CD271 staining was also observed when MnPV-induced tumors were compared with well-differentiated SCC areas (log2FC = –1.24, adj.p = 0.021) or dedifferentiated SCC areas (log2FC = –1.71, adj.p < 0.001). This indicates that cancer cells acquire stem-like characteristics during EMT and malignization, which in MnPV-induced tumors are most likely mediated by the expression of MnPV genes. Remarkably, the island-like staining pattern in dedifferentiated areas may explain the insignificant difference when compared with well-differentiated areas (log2FC = 0.42, adj.p = 0.56), since during lysis negative cells “dilute” positive ones.

3.6 Analyses of MnPV-mediated effects on the tissue

Since cutaneous SCC development is a multistep process in which the viral status plays a substantial role, it was also of particular interest to compare infected and uninfected tissue to understand the impact of MnPV infection in early carcinogenesis. Due to a different physiological context, samples were analyzed independently from the specimen described above. This also applies for the MnPV-induced tumors, which were different from the ones examined before. Here, tissues were dissected based on the positivity/negativity of MnPV E1^E4 stainings and subsequent MS analyses covered 2414 proteins of which 304 proteins were shared among all samples.

The overall view on the proteomes by PCA showed a differential clustering of the 16 individual specimens depending on their infection status (Figure 5A). As visualized in a Venn diagram (Figure 5B), MnPV-infected epidermis differed (adj.p < 0.05) from uninfected epidermis in the expression of 121 detected proteins (Figure 5B, green circle), of which 102 were shared with MnPV-induced lesions. These differed from uninfected skin in 265 proteins, indicating that viral infection directly influences the proteome. This is in line with the fact that the viral load in a permissive environment of MnPV-induced lesions exceeds that in persistently infected epidermis.¹⁴ As expected, viral proteins (i.e., E2, L1, L2, and E1^E4), were detected significantly more often when infected epidermis or MnPV-induced tumors were compared with uninfected epidermis (Figure 5C), while no significant differences between infected epidermis and MnPV-induced tumors were found. Within the proteomes, the viral proteins were among the most dramatically changed (Figure 5D, see also Supporting Information: STables 28-30). A prominent example of the top five modulated cellular proteins is CASP14, which plays a role in keratinocyte differentiation⁴⁹ and was found to be strongly upregulated in the infected epidermis (log2FC = 6.63, adj.p < 0.001) and MnPV-induced tumors (log2FC = 6.90, adj.p < 0.001) (Figure 5E). Conversely, several collagens, for example, COL1A1, were strongly diminished in infected epidermis (log2FC = –3.34, adj.p = 0.027) and MnPV-tumors (log2FC = –6.62, adj.p < 0.001) (Figure 5E, Supporting information: STables 28-30). Viral infection also caused significant upregulation of 13 members of the family of 40 S ribosomal proteins when compared with uninfected epidermis, for example, RPS2 and RPS11 (Figure 5, Supporting Information: STable 28). In MnPV-induced tumors, even 29 RPS proteins were significantly increased when compared with uninfected skin (Supporting Information: STable 30). In this biological context, it makes sense that eukaryotic translation initiation factors, for example, EIF3I, EIF4A1, and EIF5A, were also upregulated in MnPV-induced tumors (Supporting Information: STable 30). Another interesting finding was that MnPV infection apparently caused a significant downregulation of caveolin-1 (encoded by Cav1; in MnPV-induced tumors: log2FC = –4.56, adj.p < 0.001, but unchanged in infected epidermis, log2FC = –0.69, adj.p = 0.20) (Figure 5E).

IPA predictions of the datasets suggested that viral infection deactivates functions related to cell death and necrosis, but increases those linked to cell viability, protein synthesis, protein metabolism, outgrowth of cells, and inflammation (Figure 5F, see also Supporting Information: STables 31-33). This was already visible in MnPV-infected epidermis and was further enhanced in MnPV-induced tumors. The molecular basis for these changes in tissue homeostasis seemed to be related to an enrichment of EIF2 signaling and decreased GP6 signaling (Figure 5F, see also Supporting Information: STables 34-36). IPA further predicted upstream regulators in the changed signaling pathways, some of which were key factors well known from HPV infection and cancer, like the gene products of Myc (proto-oncogene c-Myc; activated), Rictor (rapamycin-insensitive companion of MTOR; deactivated), Hras (Harvey rat sarcoma viral oncogene homolog, activated), Ctnnb1 (beta-catenin; deactivated), and Cebpa (CCAAT/enhancer-binding protein alpha; deactivated). Interestingly, the gene products of Mlxipl (carbohydrate-responsive element-binding protein ChREBP; activated), Mycn (N-myc proto-oncogene; activated), and Rock1 (Rho-associated coiled-coil containing protein kinase 1; deactivated) were also predicted to be differentially active in infected tissue and uninfected epidermis (Figure 5F, see also Supporting Information: STables 37-39).

3.7 Tissue stainings validate proteomic changes caused by MnPV infection

We also verified these proteome data using tissue stainings against several differentially expressed viral and cellular gene products, some of which were partially found among the top five up/downregulated candidates (Figure 5E).

Confirming the MS data, a significant (adj.p < 0.001) increase of viral proteins could be found in infected epidermis (E1^E4: log2FC = 7.50; L1: log2FC = 9.64) and MnPV-induced tumors (E1^E4: log2FC = 9.53; L1: log2FC = 9.55) when compared with uninfected epidermis.

While the proteome analyses were not clear, a stronger K17 staining could be observed in MnPV-induced tumors when compared with uninfected epidermis (log2FC = 1.29, p = 0.029, adj.p = 0.068). In the infected epidermis, elevated K17 expression is restricted to hyperproliferative areas that could progress to macroscopically visible lesions. Conversely, K15 levels were significantly reduced in infected epidermis (log2FC = –2.83, adj.p = 0.008) and MnPV-induced tumors (log2FC = –4.24, adj.p < 0.001), which could also be observed in tissue stainings (Figure 6). According to the MS data, K14 expression was independent from the infection status, which was confirmed by immunofluorescence. K10 levels — despite accumulated in keratinized layers (Figure 6) that were removed before MS analyses (Figure 1C) — were not changed in infected epidermis (log2FC = -0.75, adj.p = 0.41), but tendentially diminished in MnPV-induced tumors (log2FC = –1.49, adj.p = 0.088). Interestingly, in some MnPV E1^E4 positive areas, a clear reduction of K10 could be observed, indicating that MnPV is interfering with the physiological keratinocyte differentiation process as reported for cutaneous HPV8.⁵⁰ This interference also causes an increase in CASP14, one of the top upregulated proteins in MnPV-infected tissue as shown by MS (Figure 5) and tissue staining (Figure 6).