Abstract graphic file with name ao4c11051_0004.jpg We employed laser microdissection to selectively harvest histology-resolved tumors and stroma from formalin-fixed, paraffin-embedded head and neck squamous cell carcinoma (HNSCC) tissues. Peptide digests from the LMD-enriched HNSCC tissue were analyzed by quantitative mass-spectrometry-based proteomics using a data independent analysis approach. In paired samples, excellent proteome coverage was achieved, having quantified 6668 proteins with a median quantitative coefficient of variation under 10%. Significant differences in relevant functional pathways between the tumor and the stroma regions were observed. Extracellular matrix (ECM) was identified as a major component enriched in the stroma, including many cancer-associated fibroblast signature proteins in this compartment. We demonstrate the potential for comparative deep proteome analysis from a very low starting input in a scalable format. Correlating such results with clinical features or disease progression will likely enable the identification of novel targets for disease classification and interventions. Introduction Head and neck squamous cell carcinoma (HNSCC) is the sixth most common cancer worldwide accounting for approximately 890,000 new cases and 450,000 deaths (4.6% of cancer deaths) annually.^[41]1,[42]2 HNSCC cells first invade the basement membrane of native epithelium; a large proportion of patients are identified at the time of diagnosis to have lymph node metastasis, which is associated with poor survival.^[43]2−[44]4 Overall, the response to the available treatments has been moderate. Linking cellular phenotypes to functional proteome states of the HNSCC tumors, including its microenvironment, will add to our understanding of the pathophysiology of the disease and potentially identify drivers of metastasis. Previously, laser microdissection (LMD) of HNSCC FFPE samples in combination with quantitative mass spectrometry (MS) was introduced by us to characterize normal and tumor lesions.^[45]5 Recent deep proteomic analysis of tumor and matched normal adjacent tissues, followed by integrated proteogenomic analysis of MS-based proteomic data with genomics and transcriptomics, has identified molecular subtypes with treatment potential.^[46]6 However, the proteome coverage of early-stage tumors remains insufficient, in part, due to the limited availability of surgical tissues for molecular analysis at this early stage of the disease. Likewise, the spatial resolution of the available HNSCC proteomic data and its correlation with clinicopathological features are very limited. Recent advances in the LMD technology and MS-based techniques allow deep proteome analysis from very low input peptide samples.^[47]7−[48]12 Motivated by these advancements, in this study, we selectively harvested tumor and stroma regions from HNSCC FFPE samples using LMD for independent MS-based proteomics analysis using a data independent analysis (DIA) approach. We achieved deep proteome coverage, identified significant differences in functional pathways of tumor and stroma, and demonstrated the potential for comparative deep proteome analysis in a scalable format. Results and Discussion dia-PASEF LC-MS Analysis of Tumor and Stroma Tumor and stroma were independently harvested by LMD from five HNSCC patient FFPE specimens for analysis by DIA MS ([49]Figure [50]1 A). We obtained an average of 0.97 μg of tryptic peptides per mm^2 of tumor and 0.48 μg/mm^2 of stroma. The samples were analyzed in triplicate (500 ng of peptide digest per injection) on a timsTOF HT MS (40 min run time, equating to an ∼2 h MS analytical time per sample). Excellent chromatographic reproducibility was observed over the entire sample batch run ([51]Figure [52]1B,C). A total of 8759 protein groups were identified from 98,005 peptide matches, among which 8498 proteins were from LMD-enriched tumor epithelium and 8396 proteins from the stromal region, with 94% of proteins coquantified in both tumor and stroma. Of these proteins, 7825 were identified with two or more peptide matches in at least one sample ([53]Supporting Table S1A). We observed similar protein coverage (range 7543–7709) in each sample ([54]Figure [55]1D). Across all patient samples, 6668 proteins were coquantified with the median coefficients of variation (CVs) consistently under 10% ([56]Figure [57]1E). Figure 1. [58]Figure 1 [59]Open in a new tab Representative (A) H&E stained FFPE tissue sections on PEN membrane slides before (left) and after LMD harvest of tumor (middle) and stroma (right). Base peak chromatograms showing the overlay of triplicate runs of an enriched (B) tumor and (C) paired stroma sample. (D) Graphical representation of the number of proteins identified with 2 or more peptide matches in the enriched tumor and stroma samples. The median is indicated by a line. (E) Schematic representation of the CV value spread of quantitation for each sample; median (bold line) and upper and lower quartiles (thin line) are indicated. Differential Proteome in Tumor and Stroma Unsupervised hierarchical cluster analysis of the proteins quantified revealed that the stroma and tumor samples clustered independently ([60]Figure [61]2 A). Differential analysis identified 2655 significantly altered proteins between tumor epithelium and stroma ([62]Figure [63]2B and [64]Supporting Table S1B). Gene ontology molecular functional pathway enrichment analysis of these proteins ([65]Supporting Table 2) identified several functional groups relevant to the stromal and tumor regions ([66]Figure [67]2C,D). Notably, extracellular matrix (ECM) structural constituent (GO:0005201), collagen binding (GO:0005518), glycosaminoglycan binding (GO:0005539), and heparin binding (GO:0008201) were enriched in the stroma. Comparatively, nucleic acid catalytic activity, DNA binding, ATP hydrolysis, and helicase and ribosome constituents were enriched in tumor epithelium. In the top 100 tumor-enriched proteins, we identified candidates with functional and physical association with cell–cell adhesion (DSC2, DSP, JUP, KRT18, PKP3, and TRIM29), cadherin binding (EPS8L1, EVPL, F11R, JUP, KRT18, LAD1, PKP1, PKP3, PPL, TRIM29), and calcium-dependent protein binding (S100A14, S100A7, S100A7A, S100A8, S100A9). These proteins are highlighted in the volcano plot ([68]Figure [69]2B, right side). Figure 2. [70]Figure 2 [71]Open in a new tab Functional association of upregulated proteins. (A) Unsupervised hierarchical cluster analysis of stroma and tumor region proteins based on abundance. (B) Volcano plot of protein abundance. In volcano plot, the box dot plots of significantly changed proteins between the regions are indicated, and selected candidates are in red. Selected enriched functional pathways (String analysis) in stroma (C) and tumor (D) are shown with enrichment score and false discovery rate (FDR). Extracellular Matrix The extracellular matrix (ECM) is composed of structural proteins such as laminins, collagens, proteoglycans, and fibronectin.^[72]13,[73]14 Several components of ECM structural constituent, including BGN, COL11A1, COL11A2, COL12A1, COL14A1, COL1A1, COL1A2, COL2A1, COL3A1, COL6A1, COL6A2, COL6A3, COL8A1, COMP, DCN, EMILIN1, FBLN1, FBLN2, FN1, HSPG2, LUM, MXRA5, PRELP, and VCAN, were among the top 100 most abundant proteins in stroma ([74]Figure [75]2 C and [76]Supporting Tables 1B and [77]2), and these are highlighted in the volcano plot ([78]Figure [79]2B, left side). Several of these stromal-enriched proteins (e.g., COL11A1, COMP, FN1, POSTN, SULF1, and THBS2) are signature markers of COL11A1-expressing cancer-associated fibroblasts (CAFs)^[80]15 and are detectable in several cancers, including HNSCC.^[81]16 COL11A1 CAFs are a common driver of aggressive cancer behavior,^[82]15 and human sulfatase 1 (SULF1)-positive CAFs were recently demonstrated to promote colorectal cancer development.^[83]17 CAFs are key component of the tumor microenvironment due to their ability to synthesize and remodel the ECM and perform diverse functions.^[84]18−[85]20 The CAF-mediated ECM remodeling includes deposition of collagen, synthesis of proteoglycans, and overexpression of remodeling enzymes, e.g., matrix metalloproteinases (MMPs).^[86]21,[87]22 We observed that collagens (listed above), proteoglycans, and MMPs (specifically, MMP2, 3, 10, and 14) were more abundant in stroma. The CAF-mediated ECM remodeling is dynamic, providing signals that regulate the tumor cells and their interaction with the microenvironment and impacts their migration and metastatic potential.^[88]23 Characterization of the stromal microenvironment is valuable,^[89]18,[90]24,[91]25 and the pathology-guided LMD enrichment method coupled with the high sensitivity of liquid chromatography with tandem mass spectrometry (LC-MS/MS) in HNSCC tissues will provide clues connecting the disease pathology to outcomes. One limitation of this study is a small sample set, which limits our ability to perform rigorous informatic analyses of specific disease association. Nonetheless, KEGG pathway enrichment and protein–protein interaction prediction of ECM proteins were performed to obtain snapshot characterizations of biological processes and protein components within the HNSCC tumor microenvironment. [92]Table [93]1 lists the proteins that are associated with the ECM–receptor interaction, protein digestion and absorption, signaling, and pathways in cancer. Table 1. KEGG Pathway Enrichment of ECM Proteins That Changed Significantly between Stromal and Tumor Regions. KEGG ID description gene count strength FDR matching proteins hsa04512 ECM–receptor interaction 26/88 1.56 3.89 × 10^–28 ITGB4, COMP, LAMB1, COL1A1, LAMA4, LAMC1, THBS1, VWF, TNC, COL6A3, COL1A2, COL6A2, LAMB2, THBS4, FN1, COL4A2, COL6A1, THBS2, THBS3, HSPG2, COL4A1, COL2A1, LAMA1, LAMA2, TNXB, DAG1 hsa04974 protein digestion and absorption 23/100 1.45 7.81 × 10^–23 COL1A1, COL21A1, COL8A1, COL5A3, COL6A3, CPA3, COL1A2, COL14A1, COL6A2, COL3A1, COL12A1, ELN, COL18A1, COL4A2, COL6A1, COL11A1, COL5A1, COL16A1, COL5A2, COL15A1, COL4A1, COL2A1, COL28A1 hsa04610 complement and coagulation cascades 16/82 1.38 8.09 × 10^–15 F9, F12, VWF, F13A1, SERPING1, FGB, F2, C1QB, A2M, FGG, SERPINC1, C1QC, C1QA, SERPINA1, KNG1, FGA hsa04510 focal adhesion 24/195 1.18 2.47 × 10^–18 ITGB4, COMP, LAMB1, COL1A1, LAMA4, LAMC1, THBS1, VWF, TNC, COL6A3, COL1A2, COL6A2, LAMB2, THBS4, FN1, COL4A2, COL6A1, THBS2, THBS3, COL4A1, COL2A1, LAMA1, LAMA2, TNXB hsa05222 small cell lung cancer 9/92 1.08 5.91 × 10^–06 LAMB1, LAMA4, LAMC1, LAMB2, FN1, COL4A2, COL4A1, LAMA1, LAMA2 hsa04151 PI3K-Akt signaling pathway 24/349 0.92 2.69 × 10^–13 ITGB4, COMP, LAMB1, COL1A1, LAMA4, LAMC1, THBS1, VWF, TNC, COL6A3, COL1A2, COL6A2, LAMB2, THBS4, FN1, COL4A2, COL6A1, THBS2, THBS3, COL4A1, COL2A1, LAMA1, LAMA2, TNXB hsa04611 platelet activation 8/122 0.9 0.00043 COL1A1, VWF, COL1A2, COL3A1, FGB, F2, FGG, FGA hsa04350 TGF-β signaling pathway 5/91 0.82 0.024 DCN, THBS1, FBN1, FMOD, LTBP1 hsa04926 relaxin signaling pathway 6/126 0.76 0.0181 MMP2, COL1A1, COL1A2, COL3A1, COL4A2, COL4A1 hsa05205 proteoglycans in cancer 8/194 0.7 0.0072 DCN, MMP2, COL1A1, THBS1, LUM, COL1A2, FN1, HSPG2 hsa05200 pathways in cancer 12/515 0.45 0.0261 MMP2, LAMB1, LAMA4, LAMC1, LAMB2, F2, FN1, COL4A2, COL4A1, LAMA1, LAMA2, KNG1 [94]Open in a new tab A predicted protein–protein association map (deduced from the STRING database analysis) of the stromal CAF-associated proteins is presented in [95]Figure [96]3 A. Many biological processes are mediated by the protein–protein interaction between stromal CAFs and other tumor-associated cells. One of the CAF proteins, SULF1, edits 6-O-sulfation of heparan sulfate proteoglycans (HSPGs).^[97]26 This affects the interaction of HSPGs with ligands and signaling molecules, e.g., fibronectin, Wnts, TGF-β1, and FGFs.^[98]27−[99]29 In an earlier study, we reported that SULF1 RNA expression is significantly higher in fibroblasts compared with that in tumor epithelial cells and has strong correlation with the expression of CAF1 marker genes.^[100]16 We have shown that SULF1 RNA and protein are elevated in multiple cancer tissues and that increased expression is associated with poor survival outcomes.^[101]16 A recent study demonstrated that SULF1+ CAFs portend the poor survival of colorectal cancer patients.^[102]17 Figure 3. [103]Figure 3 [104]Open in a new tab Protein–protein association map of stromal compartment upregulated (A) CAF markers and (B) heparin binding proteins. In this study, we observed the enrichment of SULF1 in the LMD-enriched stromal microenvironment. In addition to SULF1, a protein with high affinity for heparin/heparan sulfate,^[105]30 other heparin binding proteins were significantly enriched in the stroma, including many CAF-associated proteins, e.g., COL11A1, COMP, FN1, and POSTN. STRING analysis of the heparin binding proteins ([106]Figure [107]3 B) revealed that a large majority of the stromal-enriched heparin binding proteins are predicted to have functional and physical association, suggesting that there is likely a heparan sulfate proteoglycan interacting network that mediates functional signaling in the tumor microenvironment (TME). Overall, we demonstrate excellent proteome coverage of the HNSCC tumor and stromal microenvironment from LMD-enriched tissue samples using a highly sensitive and quantitative dia-PASEF LC-MS/MS workflow. Though our sample set was limited, we quantified proteomic alterations between the HNSCC stroma and tumor using a minimal 1.5 μg of isolated tryptic peptides for triplicate analysis per tissue-specific region. The methodology allows comparative deep proteome analysis in a scalable format and allows the connection of the enriched proteomes to pathologic features of the tumor tissues. While the peptide yield from the stromal regions was lower, the sensitivity of the LC-MS/MS workflow allows reproducible proteome analysis with coverage comparable to that of the enriched tumor regions. Future expansion of this study in a large sample cohort using this workflow will allow for further identification and validation of proteins relevant to disease progression and survival. Materials and Methods Experimental Design In this study, proteomic profiling of five patient samples was conducted. From each tissue specimen, the tumor and stroma-enriched regions were independently harvested by LMD. LC-MS analysis of these paired samples was performed in triplicate to demonstrate the technical reproducibility. Statistical analysis was performed to determine the coefficient of variation (CV) for each sample analysis. The relative protein difference between the two paired sample groups, i.e., enriched tumor (n = 5) and enriched stroma (n = 5), were compared, and Student’s t test was used to compute p-values. Patient Specimens Study participants were enrolled between 1995 and 2020 at the Department of Otolaryngology-Head and Neck Surgery, MedStar Georgetown University Hospital, under a protocol approved by the MedStar Health Research Institute-Georgetown University Oncology Institutional Review Board. Thin-sectioned formalin-fixed paraffin-embedded (FFPE) HNSCC tumor specimens mounted on poly(ethylene naphthalate) (PEN) membrane slides were prepared by Histology & Tissue Shared Resource at Georgetown University Medical Center. The tissue slides were stained with hematoxylin and eosin (H&E), imaged by microscopy, and reviewed by a board-certified pathologist for annotation of the tumor and stroma areas to inform LMD harvest. Laser Microdissection An average of 42 mm^2 (range 13–54 mm^2) and 32 mm^2 (range 15–50 mm^2) of tumor epithelium and stroma, respectively, was harvested by LMD as recently described.^[108]7 Representative micrographs before and after LMD harvest were imaged using an Aperio AT2 scanner (Leica Microsystems, Wetzlar, Germany). The LMD-enriched samples were digested using trypsin and pressure cycling technology (PCT), and the peptide samples were desalted using C18 cartridges. LC-MS Analysis A timsTOF HT mass spectrometer connected to a nanoElute 2 LC system via a CaptiveSpray 2 source (Bruker, Billerica, MA) was used for dia-PASEF LC-MS/MS analysis. Each sample was analyzed in triplicate (except one sample in duplicate, which was used for the initial method optimization and availability was limited). The peptides were separated by a C18 IonOpticks column (particle size 1.6 μm, 75 μM μm ID, 25 cm length) at a flow rate of 0.25 μL/min; solvent A (0.1% formic acid in water), solvent B (0.1% formic acid in acetonitrile), gradient of 0–28 min 5–23% B, 28–32 min to 30% B, 32–36 min to 90% B, and hold at 90% B to 40 min. For dia-PASEF analysis, the window scheme was calculated using the py_diAID tool ([109]https://github.com/MannLabs/pydiaid). The capillary voltage was set at 1600 V, stepping collision energy at 32, 40, and 50 eV, dia-PASEF scan range 100–1700 m/z in positive mode, and IMS service ramp time of 100 ms. Data Analysis Acquired mass spectrometry data was searched against Uniprot-Human-reviewed database (20,383 protein entries) using the directDIA+ workflow (Spectronaut 18 software) for protein identification and quantification using BGS default settings. Unsupervised hierarchical clustering analyses were done using Clustvis (v 1.2.0) in R package (v 3.6.2).^[110]31 Perseus software (v 2.0.11) was used to compare the relative abundance of each protein in two sample groups via performing Student’s t tests, log-fold changes, and Volcano plot analysis.^[111]32 All raw p-values were adjusted using the false discovery rate (FDR). The Gene Ontology (GO) molecular functional pathway enrichment analysis was performed using the STRING database (string-db.org, v 12.0). The proteins with statistically significant abundance variations between the two groups were uploaded along with the values for the analysis. KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment of specific protein groups was further extracted from the STRING output. The protein–protein interaction map was generated from the STRING functional and physical protein association network analysis. GraphPad Prism software (version 10.2.3) was used for data visualization. Acknowledgments