Zonal Regulation of Collagen Types and Post-Translational Modifications In Prostatic Benign and Cancer Tissues By Imaging Mass Spectrometry

Abstract

Background:

The emergence of reactive stroma is a hallmark of PCa progression and a potential source for prognostic and diagnostic markers of PCa. Collagen is a main component of reactive stroma and changes systematically and quantitatively to reflect the course of PCa, yet has remained undefined due to a lack of tools that can define collagen protein structure. Here we use a novel collagen-targeting proteomics approach to investigate zonal regulation of collagen type proteins in PCa prostatectomies.

Methods:

Prostatectomies from 9 patients were divided into zones containing 0, 5, 20, 70–80% glandular tissue and 0, 5, 25, 70% by mass of prostate cancer tumor following the McNeal model. Tissue sections from zones were graded by a pathologist for Gleason score, percent tumor present, percent prostatic intraepithelial neoplasia and/or inflammation. High resolution accurate mass (HRAM) collagen targeting proteomics was done on a select subset of tissue sections from patient-matched tumor or nontumor zones. Imaging mass spectrometry was used to investigate collagen type regulation corresponding to pathologist defined regions.

Results:

Complex collagen proteomes were detected from all zones. COL17A and COL27A increased in zones of inflammation compared to zones with tumor present. COL3A1, COL4A5, and COL8A2 consistently increased in zones with tumor content, independent of tumor size. Collagen hydroxylation of proline (HYP) was altered in tumor zones compared to zones with inflammation and no tumor. COL3A1 and COL5A1 showed significant changes in HYP peptide ratios within tumor compared to zones of inflammation (2.59 ± 0.29 p-value 0.015; 3.75 ± 0.96 p-value 0.036, respectively). By imaging mass spectrometry COL3A1 showed defined localization and regulation to tumor pathology. COL1A1 and COL1A2 showed gradient regulation corresponding to PCa pathology across zones. Pathologist defined tumor regions showed significant increases in COL1A1 HYP modifications compared to COL1A2 HYP modifications. Certain COL1A1 and COL1A2 peptides could discriminate between pathologist defined tumor and inflammatory regions.

Conclusions:

Site-specific post-translational regulation of collagen structure by proline hydroxylation may be involved in reactive stroma associated with PCa progression. Translational and post-translational regulation of collagen protein structure has potential for new markers to understand PCa progression and outcomes.

Introduction

For prostate cancer (PCa), the regions defined by tumor stroma and more normal regions adjacent to tumor stroma have emerged as novel sources for prognostic and diagnostic tissue biomarkers ^1–9. In PCa tissues, stroma morphologically and functionally changes to become reactive stroma with a significantly altered gene regulatory network compared to normal prostate stroma ^8,10,11. Reactive stroma is a maladaptation of normal tissue repair with altered extracellular matrix remodeling, growth factors, cytokines, and immune recruitment ^8,10–12. Reactive stroma emerges concomitant with pre-malignant prostate intraepithelial neoplasia, co-evolves with PCa throughout the processes of hyperplasia, inflammation, and tumor expansion, and contributes to therapeutic resistance ^{10,11,13–15}. Collagen regulation is a key histopathological feature that distinguishes reactive stroma from normal stroma and correlates with PCa progression ^7,16–19. PCa progression is accompanied by quantitative and systematic pathological changes in collagen content during PCa progression with increasing alignment of collagen fibers and increased transcription of specific collagens, particularly fibrillar collagens type 1 and type 3 ^{8,10,16,20–22}. However, it is unknown how the translational and post-translational regulation of the molecular changes in collagen structure contribute to PCa pathogenesis in the tumor microenvironment.

In many cancer types, translational regulation of very specific collagen types contributes to expansion of the tumor microenvironment, promotes tumor cell migration into the surrounding tissues, alters cell adhesion, and induces endothelial mesenchymal transition ^23–26. Multiple dynamically regulated collagen post-translational modifications (PTMs) that alter cell signaling, including non-enzymatic glycation, lysine crosslinking, N- and O-linked glycosylation, and tyrosine sulfation ^27–29. A main collagen PTM is 4-hydroxylation of proline (HYP), regulated by co-factors of iron, oxygen, and ascorbic acid, and dependency on nutritional status and oxidative stress ^28,30–33. The enzymes prolyl hydroxylases P4HA1, P4HA2, P4HA3 drive HYP modification on collagen and expression patterns of the enzyme vary by cancer cell type^33,34. Regulation of the prolyl hydroxylases are linked to metastasis, may alter cancer stemness properties and can increase sensitivity of cancer cells to drug treatment ^35–39. Variability in HYP site localization and regulation within the collagen structure alters protein folding, changes collagen triple helical configuration and affects protease access to the collagen structure ^28,29,31,40. Mammalian cells recognize very specific sites of collagen hydroxylated prolines through cell-fibril interactions with integrins and tyrosine kinase discoidin domain receptors, and this alters cell processes of adhesion, proliferation and migration^41–44.

Analysis of collagen sub-types and structural features of collagen in tissue sections can be challenging in that the stroma and extracellular matrix components are largely insensitive to standard trypsin-based proteomic analysis. There is a lack of antibody reagents specific to collagen sub-type or collagen-post-translational modifications. We have recently developed a collagenase-based imaging mass spectrometry analysis workflow that specifically targets analysis of stroma and extracellular matrix composition in formalin-fixed paraffin-embedded tissues from tissue microarray cores, biopsies and full tissue sections^45–47. In the current study, we hypothesized that clinical and pathological features of PCa prostatectomy tissues may be characterized by translational and post-translational regulation of collagen types. To test this, prostatectomies were divided by zone of tumor grade and inflammation. Thin tissue sections from the selected prostatectomies were annotated by a pathologist for tumor content, pre-malignant prostate intraepithelial neoplasia (PIN) and inflammation. Selected tissue sections were analyzed by high mass accuracy liquid chromatographic proteomics to measure and sequence collagen type proteomic signature. Regulation of collagen structure by pathology was tested using collagen targeting imaging mass spectrometry on the same tissue sections ^45–48. Major findings are that additional collagen type proteins beyond the primary fibrillar collagens type I and III are regulated between high grade PCa and inflammatory zones. We further identify that distinct sites of the collagen post-translational modification hydroxylation of proline uniquely regulated within the prostate tumor and not in adjacent tissue. Defining pathological regulation of collagen type expression provides new information towards understanding PCa.

Methods

Materials and Reagents

Xylenes, 200 proof ethanol, methanol, HPLC grade water were purchased from Fisher Scientific (Pittsburgh, PA, USA). Hematoxylin (Gill’s Hematoxylin Type 2 and Eosin Y, intensified, Cancer Diagnostics) purchased through Thermo Fisher Scientific (Waltham, MA, USA). Collagenase type III (C. histolyticum) was purchased from Worthington (Lakewood, NJ, USA). Acetonitrile, α-Cyano-4-hydroxycinnamic acid (CHCA), trifluoroacetic acid (TFA), ammonium bicarbonate, ammonium phosphate monobasic, calcium chloride, and Trizma® base were purchased from Sigma-Aldrich (St. Louis, MO, USA).

Tissue Procurement

Tissue sections were procured through the Biorepository & Tissue Analysis Shared Resource (BTASR) at the Hollings Cancer Center at the Medical University of South Carolina (MUSC) with Institutional Review Board approval. Tissues came from prostatectomies divided into zones containing 0, 5, 20, 70–80% glandular tissue and 0, 5, 25, 70% by mass of prostate cancer tumor following the McNeal model ⁴⁹. Data for each sample included Gleason grade, percent tumor present, percent PIN and/or inflammation, and serum PSA levels at time of diagnosis (Figure 1).

Figure 1.

Pathological features of the prostatectomy zones used in the study. Each tissue is annotated by a KKU and same SKUs indicate that the tissue is from the same patient. For this study, collagen targeting sequencing proteomics was done on a select group of tumor or nontumor tissue (blue square). Collagen targeting imaging mass spectrometry was done on all tissue sections. INF- inflammation, PIN-prostate intraepithelial neoplasia.

Tissue Preparation for Imaging Mass Spectrometry

Each selected formalin-fixed, paraffin-embedded (FFPE) tissue block was sectioned into 5-μm thick sections by MUSC BTASR and two serial sections stained with hematoxylin and eosin. Stained tissue sections were digitally scanned (Nanozoomer, Hamamatsu). Serial sections were annotated for pathology and one section with pathological annotation was retained for reference. For the remaining serial section, coverslips were removed by overnight incubation in xylenes. The same FFPE sections of prostate tissue were prepared for imaging mass spectrometry as previously described ^45–48. Briefly, tissues were heated, dewaxed, antigen retrieved at pH 3 and sprayed with a recombinant glycosidase (PNGase F Prime™, N-zyme Scientifics) to deglycosylate proteins and improve enzymatic access to collagens ⁴⁸. Released N-glycans were removed from the tissues using an aqueous high pH, low pH strategy ⁵⁰. Tissue was then antigen retrieved using 10 mM Tris buffer pH 9, and a thin layer of collagenase type III (collagenase) was applied by automated sprayer (M3 TM-Sprayer, HTXImaging, Chapel Hill, NC, USA). Collagenase spraying used parameters of 45°C, 10 psi, 25 μL/min, 1200 velocity, and 15 passes with a 3.0 mm offset. Samples were collagenase digested in ≥80% relative humidity at 37.5°C for five hours followed by application of CHCA matrix prepared as 7 mg/mL in 50% acetonitrile, 1% TFA with a spiked standard of 200 femtomole/microliter [Glu1]-Fibrinopeptide B human (Glufib) (Sigma-Aldrich, St. Louis, MO, USA). CHCA was applied by automated sprayer with parameters of 79°C, 10 psi, 70 μL/min, 1300 velocity, and 14 passes with a 2.5 mm offset. Slides were rapidly dipped (<1 s) in cold 5 mM ammonium phosphate, monobasic to limit matrix cluster formation ⁵¹ and dried in a desiccator prior to imaging mass spectrometry data acquisition.

Imaging Mass Spectrometry Acquisition and Image Analyses

PCa tissue sections were analyzed using a MALDI-FT-ICR (solariX™ Legacy 7.0 Tesla, Bruker Scientific, LLC, Bremen, Germany) in positive ion mode, collecting 200 laser shots per pixel. Transients of 512 kilowords were acquired in broadband mode over mass to charge (m/z) 600–3,000 with a calculated on-tissue mass resolution at full width half maximum of 81,000 at m/z 1400. Lockmass (GluFib peptide m/z 1570.6768) was maintained at 10 ppm during tissue imaging. All data were analyzed using linear interpolation without denoising normalized to total ion current using SCiLS Lab software 2019 Pro Build 7.02.10901 (Bruker Scientific, LLC, Bremen, Germany). Digital images of H&E stained tissue sections marked by the pathologist were uploaded and co-registered to the imaging mass spectrometry data. Comparisons were made based on overall tissue characterization or pathological regions of tumor, prostatic intraepithelial neoplasia or inflammation. Statistical significance of extracted peak intensities compared between regions was determined using exact p-values calculated by the Mann-Whitney test (GraphPad, Prism; version 8.3.0). Percent composition tumor, prostatic intraepithelial neoplasia, or inflammation (INF) was calculated based on pathologist annotated area compared to total area using ImageJ⁵². Peptide identifications were made by high mass accuracy (≤ 3 ppm) matches from LC-proteomic data done on the same tissue sections.

Chromatographic Proteomics

After imaging, matrix was removed from the tissue sections using ethanol washes ⁵⁰. Tissue sections were scraped into centrifuge tubes and digested with collagenase overnight in solution as previously described ⁴⁵. Collagenase peptides were dried down and desalted by solid phase extraction using C18 StageTip (Thermo Scientific). Peptides were separated and analyzed on an EASY nLC 1200 System (ThermoScientific) in-line with the Orbitrap Fusion Lumos Tribrid mass spectrometer (ThermoScientific) with instrument control software v. 4.2.28.14. Peptides were pressure loaded at 1,180 bar, and separated on a C18 reversed phase column (Acclaim PepMap RSLC, 75 μm x 50 cm (C18, 2 μm, 100 Å), ThermoFisher) using a gradient of 2% to 35% B in 180 min (Solvent A: 0.1% FA; Solvent B: 80% ACN/ 0.1% FA) at a flow rate of 300 nL/min. The column was thermostated at 45 °C. Mass spectra were acquired in data-dependent mode with a high resolution (60,000) FTMS survey scan, mass range of m/z 375–1620, followed by tandem mass spectra (MS/MS) of the most intense precursors with a cycle time of 3 s. The automatic gain control target value was 4.0e5 for the survey MS scan. Fragmentation was performed with a precursor isolation window of 1.6 m/z, a maximum injection time of 50 ms, and HCD collision energy of 35%; the fragments were detected in the Orbitrap at a 15,000 resolution. Monoisotopic-precursor selection was set to “peptide”. Apex detection was not enabled. Precursors were dynamically excluded from resequencing for 20 sec and a mass tolerance of 10 ppm. Advanced peak determination was not enabled. Precursor ions with charge states that were undetermined or > 7 were excluded. For protein analysis, tandem mass spectra were searched using both MASCOT (Version 2.4.01) and SEQUEST HT via Proteome Discoverer 1.4 (Thermo Scientific, Waltham, MA, USA) against a subset of human protein sequences downloaded on May 7, 2017 from UniProtKB (SwissProt) containing 1,783 entries (keywords used: collagen, elastin, aggrecan, gelatin, osteonectin, perlecan, plasminogen, and fibronectin). Search parameters were unspecified enzyme, precursor mass tolerances of ± 20 ppm, and fragment mass tolerances ± 0.8 Da. Methionine oxidation, asparagine and glutamine deamidation were used as variable modifications. Data were uploaded into Scaffold v4.8.1 (Proteomesoftware, Portland OR, USA) with protein probabilities ≥99% including two or more peptides. Hydroxyproline site modification probabilities were reported using MaxQuant version 1.6.3.3 ⁵³ searched against the same human database. Protein localization and function relative to collagen regulation was determined using Uniprot ⁵⁴ via incorporated databases (Reactome) and primary publications. Accurate mass comparisons between image data and proteomic data (≤ 3 ppm) were done by calculating accurate mass of a peptide using Protein Prospector version 5.22.0 (Baker, P.R. and Clauser, K.R. http://prospector.ucsf.edu, University of California at San Francisco, San Francisco CA, USA). Gene Ontology⁵⁵ or manual literature searches were used to annotate non-collagen extracellular matrix proteins for potential role in collagen regulation.

Results

Overview.

Pathology tissue blocks of prostatectomies from nine PCa patients were evaluated for localization and regulation of collagen types. Tissues were selected to represent pathological variation by PCa progression. Data from each tissue section included an H&E stain with pathological annotation of tumor, prostatic intraepithelial neoplasia (PIN), and inflammation (Supplemental Figure 1). Prostate specific antigen (PSA) levels were recorded. Chromatographic proteomics was used to quantify collagen by type on a select subset (the top highest-grade tumors and 4 glandular zones with or without inflammation). Collagen imaging mass spectrometry data was used to measure signals from collagen peptides that localized with specific pathologies.

Pathological characterization of prostatectomies.

A cohort of prostatectomies from 9 consented patients was examined for localization and regulation of collagen types. Median age was 65.9, 95% CI [62.2, 69.3]. Race was primarily Caucasian (8 Caucasian, 1 African-American). Prostatectomies were divided by % tumor composition and glandular zone following the McNeal model ⁴⁹. One zone representing high grade tumor and one zone representing lower grade tumor or no tumor were selected for analysis. PCa is a heterogenous cancer type and PCa tumors carry multiple grades of cancer cells accompanied by changes in glandular morphology. Pathological grading of PCa was done using the 2014 World Health Organization grading system with five prognostic grade groups ⁵⁶. This system reports Gleason scoring by primary (x) and secondary (y) cancer cell grading with five grades possible dependent on the sum of x+y. In the selected cohort, pathologies from all 2014 WHO grade groups were represented (Figure 1) with the majority having a grade of 2 (3+4 =7). Median percent tumor size was 18.4% of total area, 95% CI [10.8, 26.0]. Prostatic intraepithelial neoplasia, a precursor of malignant cells ⁵⁷, was present in five tissue sections (10360B, 10191B, 9161A, 10173A, and 10304A). Three tissues had no tumor and regions of inflammation (10304B, 10172B, and 10582B, percent of inflammation 7.5, 4.9, and 19.1, respectively). One tissue showed no tumor and no inflammation with compacted glandular stroma (9161B). Serum levels of prostate specific antigen (PSA) are reported as an additional pathological readout. PSA levels > 4.0 ng/mL have been used to recommend prostate biopsies⁵⁸, although research has demonstrated that PCa may be present with low levels of serum PSA or may not be present in men with higher levels of serum PSA ⁵⁹. In the cohort, serum PSA varied from 2.0 to 7.7 ng/mL (median 5.5, 95% CI [3.6–7.4]. Notably, a PSA reading of 2.0 ng/mL was reported for a tissue with a >3 cm tumor present (10191B). In summary, the cohort represented suitable pathological variation towards probing collagen type and distribution across the disease severity spectrum of PCa.

Complexity of collagen type protein varies with PCa pathology.

Collagen targeting proteomics was used to determine regulation of collagen types in PCa. The top four high grade tissues were compared with three tissue sections containing inflammation and the single tissue section that represented compacted glandular stroma and no tumor or inflammation. A total of 45 proteins were identified with 100% probability and ≥2 unique peptides. Within the identified protein population, a total of 19 collagen type proteins were represented (Figure 2A, Supplemental Table 1). Additional extracellular matrix proteins included vimentin (VIM), lumican (LUM), decorin (DC), biglycan (BGN), transforming growth factor beta induced (TGFBI), and ADAM Metallopeptidase Domain 15 (ADAM15). Bioinformatics from gene ontology biological processes and primary literature searches reported that the detected non-collagen proteins were involved in collagen regulation with primary processes of collagen fiber regulation and collagen binding (Figure 2 B, Supplemental Table 2). Collagen composition was clustered by percent composition and varied for each tissue sections (Figure 2C). Primary collagens were the fibrillar collagens collagen alpha-1(I) chain (COL1A1), collagen alpha-2(I) chain (COL1A2) and collagen alpha-1(III) chain (COL3A1). Collagen type IV (α1, α2, α3), collagen type V (α1, α2, α3), and collagen type VI (α1, α2, α3) were the next abundant collagen types and changed in abundancy by patient tissue and potentially within a single prostate. For instance, SKU 9152 (A&B) and 10582 (A&B) represent two patient pairs. The high-grade tumor 9152A (55.4% tumor) showed increased composition of collagen type VI compared to pair 9152B that had very low tumor content (4.2% tumor). Similarly, SKU pair 10582 A (15.7% tumor) and 10582 B (no tumor and 4.9% inflammation by area) showed variation in percent composition of collagen type 1(alpha XI) α1, α2 (COL11A1 and COL11A2) with unique peptides from COL11A2 detected in non-tumor 10582B. Collagen alpha-1 (II) (COL2A1), a cartilaginous collagen associated with mineralization, was represented by an average of 6.3 exclusive unique peptides per tissue (tumor or inflammation) and average total spectral counts over 264. Trends for regulation of collagen type proteins were compared between high grade tumor (n=4) and tissue sections with inflammation (n=3) (Figure 2 D–F). Collagen types XVIIα1 (COL17A1) and XXVII α1 (COL27A1) trended to decreasing in PCa compared to tissue with inflammation (p-values 0.06, 0.004 and 0.07). Collagen types COL3A1, COL4A5, and collagen α2(VIII) COL8A2 trended to increasing in PCa (p-values 0.07, 0.08, and 0.01 respectively). Overall, the data depicts that collagen type composition changes by pathological grading while data from SKU-pairs suggests that collagen composition is dynamically regulated across the prostate during PCa.

Figure 2.

Collagen composition from high mass accuracy collagen sequencing proteomics. The normalized total spectra from unique peptides is used to illustrate composition. A) Total cellular localization of collagen and other proteins identified by the study. B) Non-collagen proteins participate in binding and regulation of collagen, contributing to tissue homeostasis. C) Collagen type distribution by hierarchical sunburst analysis which arranges collagen types by abundance. Center of each sunburst reports SKU, grade of tumor, or percent inflammation. Zones with inflammation shows more abundance of different collagen types compared to tumor zones. D) Collagen types with trends that decrease in PCa as evaluated by t-test <0.01 when comparing tumor to inflammatory zones. E) Collagen types with trends that increase in PCa as evaluated by t-test <0.01 when comparing tumor to inflammatory zones. F) Evaluation of fractional composition by zone type.

Collagen hydroxylation of proline (HYP) is regulated in PCa.

HYP is an important post-translational regulation of collagen that stabilizes collagen structure, protects against degradation, and facilitates cell-fiber recognition through integrins and discoidin domain receptors (DDRs)^28,33,42,43. Changes in HYP site regulation may alter collagen organization to facilitate cell processes, e.g., invasion, migration, remodeling⁴⁴. While previous studies have shown PCa upregulation of prolyl hydroxylases, the specific site regulation of HYP within the collagen structure has not been reported in PCa. Secondary database searches for HYP demonstrated that the number of HYP peptides varied by patient and pathological status of the tissue (Figure 3, Supplemental Table 3). Comparison of number of peptides with HYP modifications to unmodified peptides showed that per patient, primary collagens COL1A2 and COL3A1 had ratios of HYP/unmodified peptides ≥1.5 (Figure 3A). COL3A1 had significant increases in ratios of HYP/unmodified peptides in tumor versus nontumor tissue (2.59 ± 0.29 tumor versus 1.92 ± 0.10 no tumor, p-value 0.015). COL5A1 also showed an increase in ratios of HYP peptides to unmodified peptides (3.75 ± 0.96 tumor versus 1.94 ± 0.58 no tumor, p-value 0.036). Alignment of full-length collagen sequence with collagen helical domain (pfam PF01391) and fibrillar collagen C-terminal domain (pfam PF01410) showed that the triple helical domain had significant sequence coverage (Figures 4B). Many identified peptides showed more than one HYP site per peptide and HYP varied by patient and tumor status. Previous reports suggest that collagen types transcriptional regulation may be associated with PSA levels⁶⁰. Analyses between collagen type peptides and PSA levels reported there were no identified peptides with a positive correlation to PSA levels, while three peptides showed a significant negative correlation to PSA levels (Table 1). Interestingly, all three of the peptides were modified by hydroxylation of proline and all three were within the triple helical regions. Together, this data suggests that HYP regulation of the collagen structure is a dynamic and likely patient specific response in PCa with a potential influence on collagen structure by PSA status.

Figure 3.

Analyses of collagen hydroxylation of proline by proteomic sequencing. Peptides with negative log probability scores ≥ 100 were used in comparison. A) Comparison of ratio of peptides having one or more hydroxylated proline to unmodified peptides per collagen type. COL3A1 and COL5A1 show higher amounts of HYP in tumor zones than in inflammatory zones, p-value is Mann-Whitney test. B) Sequence coverage of COL1A1, where green highlights hydroxylated prolines. Boxed in region highlights differential hydroxylated proline sites spanning triple helical domains and fibrillar collagen c-terminal domains from same-patient zones with high amount of tumor (9152B, 55.4% tumor) and low amount of tumor 9152A (4.2% tumor). C) COL1A2 sequence coverage where green highlights hydroxylated proline sites. Boxed in region highlights potential changes between zones with high amount of tumor versus low tumor in the same patient. D) COL3A1 sequence coverage where green indicates hydroxylated proline site. COL3A1 has 145 potential sites of 4-hydroxyproline (source, uniprot). Black arrows highlight regions differentially modified with high tumor content (9152B) compared to low tumor content (9152A). Zones were grouped by SKUs as follows: High grade tumor = 9152B (55.4% tumor), 9152A (4.2% tumor), 10360A (13.5% tumor) and 10582A (15.7% tumor with intraductal features). Inflammation = 10191A (19.1% inflammation), 10172B (7.5% inflammation), 10582B (4.9% inflammation). No tumor and no inflammation = SKU 9161B.

Figure 4.

Translational and post-translational regulation of COL3A1 structure across PCa pathology by imaging mass spectrometry. Hydroxylation of proline is annotated by site probability in parenthesis from the highest scoring peptide. Abbreviation AA indicates amino acid residues that are sampled from the collagen structure. INF- inflammation, PIN-prostate intraepithelial neoplasia. A, D, G, J) COL3A1 per pathological region show significant changes in expression between tumor and PIN. Inflammation expression is more diffuse and may vary dependent on distance to the tumor and grade of tumor within the prostatectomy. B, E, H, K) Overall patterns of each peptide with tumor, PIN or inflammation annotated. C,F,I L) Comparison of high grade tumor (9152) to high grade tumor with intraductal features (SKU 10582). SKU 10582 shows increased expression levels of each peptide within the tumor site and with gradients extending from the tumor dependent on what portion of the COL3A1 structure is sampled. J-L) highlight that it is specific prolines that are differentially modified by pathology. Proline 749 is not modified, while proline 755 has the highest observed site probability at 95%; proline 760 presents with a low site probability at 4.8%.

Table 1.

Peptides showing negative correlation with PSA values. Precursor mass under 3 parts per million (PPM) was used to match between high mass accuracy chromatography sequences and the high mass accuracy image data.

Collagen Type	Sequencea	Modifiedb	AA position	Peptide Scorec	Theoretical M+H	Observed m/z	Mass Error (PPM)	Correlation Coefficient	p-valued
COL3A1	GQRGEP(1)GP	371	365–373	102.0	941.4435	941.4428	2.1	−0.77	0.01
COL3A1	GEP(1)GGP(1)G ADGVPGKDGPR	746;749	744–761	94.4	1651.7671	1651.7650	1.3	−0.71	0.03
COL1A1	GESGREGAP(1) GAEGSP(1)GRD GSP(1)	1018; 1024; 1030	1010–1030	176.5	1974.8384	1974.8370	0.7	−0.72	0.03

Abbreviations: AA- amino acid.

^a-hydroxyproline site probabilities listed in parenthesis;

^b-Amino acid site modification;

^c-peptide score is −log probability

^d-Spearman’s Correlation; p-value of the correlation test; Mass error in parts per million (PPM).

COL3A1 post-translational modification of hydroxylated proline increases in tumor zones.

To understand where collagen structure regulation corresponding with PCa pathological localization, imaging mass spectrometry was used to map primary collagen peptides across the tissue, comparing by pathological region. Annotated pathological regions included 5 regions of inflammation (INF) and six regions of prostatic intraepithelial neoplasia (PIN). Inflammation region on SKU 10173 was surrounded by tumor and intensity expression levels correlated with those of the tumor (Spearman’s correlation 0.979 p-value <0.001) suggesting that the region of inflammation was under the influence of the tumor field (Supplemental Figure 2). COL3A1 peptide intensities matched by high mass accuracy appeared with a diffuse gradient that increased in tumor regions compared to nontumor regions (Figure 4). This correlated with LC-proteomics data where tissues with tumor had elevated expression levels of COL3A1. Certain COL3A1 peptides showed differential patterns between zones, suggesting differential regulation of COL3A1 structure throughout prostatic tissue. For instance, COL3A1 amino acid positions 444–452 (peptide 444–452) appeared lower in regions surrounding high grade tumor (Figure 4B, C). However, in SKU 10582 which had tumor with intraductal features, COL3A1 peptide 444–452 appeared elevated in the primary tumor region with increased expression in regions of inflammation. Other COL3A1 peptides also increased in intensity localized to tumor (Figure 4C, F, I, J). The intensity distribution of these peptides varied in tissue adjacent to the tumor (Figure 4C compared to 4F). For tissue with inflammation, COL3A1 increases appeared to be adjacent to regions of inflammation. Generally, both inflammation and PIN regions had significantly lower COL3A1 intensity than in tumor (Table 2). However, the tumor with intraductal features showed overall increased intensity compared to other high grade tumors. This suggests that in PCa tissue, COL3A1 expression may vary dependent on distance from the tumor and, potentially, the type of tumor. HYP peptides matched by mass accuracy in this study had similar pathological distributions to unmodified COL3A1 peptides. As an example, the unmodified peptide at amino acid sequence 747–761 with triple prolines (Figures 4G–I) had a positive correlation with expression levels of the double HYP modified peptide (Spearman’s correlation 0.706, p-value <0.001; Figure 5J–L). The combined LC and imaging proteomics data suggested that HYP site localization on the collagen structure is very specific. The sequencing data from LC-proteomics showed that HYP at amino acid position 755 had an overall higher probability of site modification (0.95) compared to prolines at position 749 (null probability of HYP modification) and proline at position 761 (0.048 probability of modification) (Figure 4K–L). This appeared with intensity in the tumor with intraductal features and adjacent to high grade tumor regions (SKU 9152, Figure 4L). Overall, COL3A1 is regulated in defined pathological regions at the translational level and at the post-translational level through hydroxylation of proline.

Table 2.

COL3A1 peptides mapping to image data used for comparison between tumor and non-tumor (where non-tumor is prostatic intraepithelial neoplasia, PIN). Peptide peak intensities were extracted from area of pathology annotation (tumor n=11; inflammation n= 4; PIN =5). Precursor mass error under 3 parts per million (PPM) was used to match between high mass accuracy chromatography sequences and the high mass accuracy image data.

Collagen α−1(III) chain (COL3A1)
Sequencea	Modifiedb	AA position	Peptide Scorec	Theoretical M+H	Observed m/z	Mass Error (PPM)	PIN/Tumor Ratio	p-valued
GPRGERGEA	None	444–452	121.6	928.4595	928.4620	−2.7	2.98 ± 0.13	0.054
GKSGDRGESGPA	None	1059–1070	168.6	1117.5232	1117.5246	−1.3	1.96 ± 0.09	0.280
GGPGADGVPGKDGPR	None	747–761	237.5	1336.6604	1336.6574	2.2	3.25 ± 0.10	0.002
GRP(1)GERGLP(1)	239;245	237–245	170.1	970.5056	970.5037	2.0	4.38 ± 0.18	0.022
GRDGNPGSDGLP(1)	1022	1011–1022	130.1	1157.5182	1157.5145	2.8	2.85 ± 0.14	0.145
GSP(1)GERGETGP(1)P(1)	797;805;806	795–806	120.9	1188.5127	1188.5133	−0.5	2.49 ± 0.10	0.072
GGPGADGVP(0.952)GKDGP(0.048)R	755;760	747–761	116.2	1352.6553	1352.6592	−2.9	2.87 ± 0.11	0.018
GARGNDGARGSDGQP(1)	332	318–332	101.9	1430.6367	1430.6389	−1.5	2.73 ± 0.11	0.063

Abbreviations: AA- amino acid.

^a-hydroxyproline site probabilities listed in parenthesis;

^b-Amino acid site modification;

^c-peptide score is −log probability

^d-Mann-Whitney U

Figure 5.

Regulation of COL1A1 and COL1A2 hydroxylated prolines (HYP) localized to pathology. Heatmaps highlight changes in expression levels across tissue. Amino acid positions are indicated in parenthesis. A) Combined ion map for COL1A1 top ten most abundant HYP modified peptides. B) Combined ion map for COL1A2 top ten most abundant HYP modified peptides. C) Tumor localized comparison of HYP/ Unmodified peptide ratios. COL1A1 shows higher ratios of hydroxylated proline when compared to COL1A2. D) Analysis of COL1A1 and COL1A2 composition within same patient tumor or nontumor zones. Analysis uses image segmentation which COL1A1 and COL1A2 clusters confidently identified peptides based on intensity and localization. Clusters are shown by color, tumor is highlighted by spectral cluster with orange color sampled by 957 spectra.

Collagen α−1(I) and collagen α−1(II) structures are regulated with PCa pathology.

The primary fibrillar type collagen type 1 forms the structure of all human tissue and organs ⁶¹; tissue homeostasis is controlled by localized translational and post-translational regulation of the collagen structure that becomes imbalanced in disease. Localization of Collagen α−1(I) chain (COL1A1) and collagen α−1(II) chain (COL1A2) peptides were investigated as indicators of structural variation due to PCa pathology. By imaging mass spectrometry, distinctive gradient changes were observed for peptides from COL1A1 and COL1A2 structures (Supplemental Figure 3). Notably, these gradients were not all overlapping, suggesting very specific structural components of collagen 1 type proteins are modulated in PCa pathology. Confidentially identified HYP peptides from COL1A1 and COL1A2 further demonstrated differential patterning in the combined ion images (Fig. 5A&B). HYP modified peptides generally had gradients with more intensity localized around or in tumor sites rather than in nontumor zones. Quantification of identified HYP peptide/Unmodified peptides within tumor regions established that there were significant differences in ratios of tumor localized HYP peptides for COL1A and COL1A2 (COL1A1 1.77 ± 0.04; COL1A2 1.44 ± 0.18, p-value 0.029) (Figure 5C). By in silico analysis, putative hydroxyproline sites on each chain reported 129 sites for COL1A1 and 16 sites for COL1A2. The variation in potential sites compared to actual findings in PCa tissue suggests a potentially higher complexity in hydroxyproline site regulation within collagen 1 type structures that occurs within tumor localized regions. Segmentation mapping of all identified COL1A1 and COL1A2 peptides was used to cluster COL1A1 and COL1A2 peptides to pathology based on localization and intensity (Figure 5D). Here, clustering reported distinct patterns in specific tumor regions, particularly SKU 10582 (intraductal features) and 9161 (high cell density), which were not observed in the nontumor zones. Additionally, area under the receiver operating curve reported that certain COL1A1 and COL1A2 peptides could distinguish between pathologist define regions of inflammation and tumor (Table 3). The combined data implies that collagen type 1 structures are differentially regulated by PCA tumor and, with larger studies, could become useful in the development of markers reporting emergence of tumor.

Table 3.

Example of COL1A1 and COL1A2 collagen peptides mapping to image data. Area under the receiver operating curve was used to determine if the peptides could discriminate between pathologist defined tumor and inflammation. AUC ≥0.700 was used as a test result indicating high sensitivity and specificity in distinguishing between same patient inflammation and tumor regions, p-value was <0.001.

Collagen α−1(I) chain (COL1A1)
Sequencea	Modifiedb	AA position	Peptide Scorec	Theoretical M+H	Observed m/z	Mass Error (PPM)	AUC INF vs Tumor
GPVGPVGAR	No	1076–1084	167.3	809.4628	809.4617	1.4	0.743
GPRGETGPA	No	908–916	154.7	841.4163	841.4165	−0.29	0.751
GPAGERGEQ	No	626–634	168.3	900.4170	900.4181	−1.2
GPAGARGNDGATGAA	No	317–331	163.5	1242.5822	1242.5799	1.8	0.740
GPAGERGSP(1)	514	505–514	143.4	843.3955	843.3953	0.26	0.786
GPAGQDGRP(1)	565	556–565	116.4	870.4064	870.4074	−1.1	0.738
GKDGLNGLP(1)	1159	1150–1159	115.7	886.4629	886.4606	2.6	0.873
GPAGRP(1)GEVGP (0.003)P(0.997)	919;925	913–925	209.7	1122.5538	1122.5520	1.6	0.887
GPRGLP(1)GER	307	301–310	242.1	954.5116	954.5089	2.8	0.896
Collagen α−2(I) chain (COL1A2)
GPVGRTGEVGAV	No	826–837	152.3	1098.5902	1098.5925	−2.1	0.755
GERGLHGEF	No	571–579	121.6	1001.4799	1001.4829	−3.0	0.765
GPAGATGDRGEAGAAGPA	No	685–702	188.1	1482.6932	1482.6972	−2.7
GPAGARGSDGSVGPV	No	229–243	182.4	1283.6339	1283.6338	0.04	0.798
GLVGEP(1)GPA	348	343–351	139.3	1082.6317	1082.6347	−2.8	0.884
GSP(1)GERGEVGPA	711	709–720	112.9	1343.6186	1343.6217	−2.3	0.798

Abbreviations: AA- amino acid; I, m/z – mass to charge; Mass error in PPM- parts per million; INF- Inflammation.

^a-Sequences with hydroxyproline site probabilities listed in parenthesis if present;

^b-Proline site modification;

^c-peptide score is (-log probability).

Discussion

Prostate cancer has been linked to collagen regulation at the genetic and transcription level ^8,10,11 with few studies on translational and post-translational regulation of collagen. Stroma morphologically and functionally changes to become reactive stroma, a maladaptation of normal tissue processes that promotes the emergence and progression of prostatic cancer ^8,10. Reactive stroma is characterized by upregulation of the primary fibrillar collagens type 1 and 3 accompanied by increasing alignment of collagen fiber structures ^{8,10,16,20–22}. In the current study, we used our novel collagen targeting imaging method⁴⁵ that works on all types of formalin-fixed, paraffin-embedded tissue samples to investigate translational and post-translational modification of collagens across pathologically defined zones of prostatectomies. A major finding within this study is that the aggregate collagen type protein profiles of prostate tissue quantitatively vary by zone within same patient prostatectomies. We also report that post-translational regulation of the primary fibrillar collagens type I and III varies based on pathological localization within the prostatic gland. Additionally, the method reported extracellular matrix proteins interacting with the collagenome. These findings provide important information on heterogeneity within the prostatic gland and links to disease sensitive changes in collagen structure as potential biomarkers of PCa.

Changes in collagen type regulation observed between tumor zones and zones with inflammation involve a continuum of collagen signaling across initiation and progression of PCa. In PCa, initial prostatic stroma changes are accompanied by collagen deposition and closing of glandular regions by transformation of prostate stromal cells into myofibroblasts. A desmoplastic response, called reactive stroma, contributes to transformation of prostate stromal fibroblasts into myofibroblasts, in part through TGFβ1 driven pathways^10,20,62. Myofibroblasts in turn quantitatively increase collagen deposition and remodeling results in re-alignment of collagen fibers with increasing tumor grade ^7,63. Remodeling also results in the release and production of repair response, notifying the immune system to recruit inflammatory cells to the site of collagen deposition^10,62. Adaptive immune responses may in turn, result in specific remodeling facilitated by immune cells^64,65. Our proteomic analysis showed that collagen types of the prostatic gland are regulated across inflammation and tumor tissue. Zones of inflammation showed an abundant and complex mix of collagen types with trends of increasing type XVII (COL17A1) and XXVII (COL27A1). COL17A1 is a favorable prognostic marker in breast cancer ⁶⁶ with expression occurring in basal cell carcinoma of the skin ⁶⁷. The role of COL27A1 in cancer is poorly characterized, yet COL27A1 increases during processes of embryogenesis and in adult cartilage ^68,69. From these data, we suggest that COL17A1 and COL27A1 may be part of the reactive stroma response when tumor is present in the prostate tissue. The decreased collagen complexity in tumor zones was accompanied by increased collagens types III (COL3A1), IV (COL4A5), and VIII (COL8A2). Increased COL3A1 corresponds to previous reports of increased COL3A1 by prostate myofibroblasts through TGFβ1 pathways²⁰. Decreases in basement membrane expression of COL4A5 were previously observed in invasive prostate carcinoma⁷⁰. This contrasts with our finding that COL4A5 increases in tumor zones compared to zones of inflammation. It is possible that elevation of COL4A5 represents a later “stage” of reactive stroma and may have some dependence on the percent, type and grade of tumor found per zone. Additionally, remodeling could involve other types of post-translational modifications or truncation products not found in our study ⁷¹. We detected increases in collagen type VIII (COL8A2) in tumor zones compared to inflammation. Interestingly, collagen type VIII has been shown to be induced by physical cell to cell contact in PCa bone metastasis⁷². Recent work has shown that collagen type VIII is a potential biomarker of prostate cancer, and is elevated in the serum of prostate cancer patients compared to normal controls⁷³. We propose that reactive stroma may be staged by collagen type composition and regulation; identifying exact contributors of early stages compared to tumor adjacent stroma may lead to the most effective therapeutic targets. These specific proline sites are currently being assessed in larger prostate cancer tissue cohorts to evaluate their biomarker potential.

Collagen structure modulation includes diverse translational and post-translational regulation of various portions of the structure to form networks and fibrils that are essential for tissue homeostasis and are deregulated in disease ^28,61,74. In this study, we focused on PCa post-translational regulation of collagen by hydroxylation of proline (HYP). Significant increases in HYP site regulation were found on the fibrillar COL3A1 when measured in tumor zones compared to nontumor zones. This specific observation of overall HYP increased in tumor matches evidence that proly-4-hydroxylases are upregulated by decreases in oxygen tension, a potential tipping point in the cancer-regulated re-alignment of collagen fibers that promotes cancer fibroblast migration from the tumor site^30,35,75. HYP sites on fibrillar collagens COL1A1, COL1A2 and COL3A1 were localized to tumor regions with gradients extending outside of the tumor margin that may possibly follow oxygen tension gradients due to dense glandular stroma. An important finding in the current work is that certain collagen prolines but not all prolines were modified by hydroxylation of proline within the PCa tumor microenvironment. Furthermore, not all HYP sites were significantly regulated between tumor and nontumor tissue. This suggests that it is very specific prolines within the collagen structure that may be susceptible to disease modification, supporting that post-translational changes to collagen structure may be biomarkers useful for PCa.

Patient specific variability was observed both in stromal pathology and with the study on translation and post-translational modification of collagen. A main variability appeared to be with localization of the HYP site within the collagen structure. While this is likely dependent in some ways upon the amount and grade of tumor within the zone, it is possible that secondary factors play a role in which prolines are hydroxylated. Proline metabolism alone has a critical role in tissue homeostasis through wound healing, oxidative stress regulation and immune response ^76–79. By serum analysis, aggressive and non-aggressive PCa have been significantly linked to negative regulation metabolites within the proline metabolic pathway⁸⁰. Inclusion of proline in a metabolite panel with PSA has been suggested to improve PCa diagnosis⁸¹ and the current study found negative correlation of proline site regulation correlating with PSA measurements. HYP modification requires cofactors of iron, oxygen, and nutritional intake of ascorbic acid. Losses of collagen HYP lead to defects in normal tissue development & integrity, wound healing, and result in diseases such as scurvy ^37,82,83. The socioeconomic status of the patient thus plays a large role in tissue homeostasis through multiple points that regulate collagen proline, particularly through nutritional access, immune response, and life stress. Additionally, racial status is a significant factor in PCa progression. African American men show a high risk for PCa, clinically present with higher grade PCa at a younger age, and are twice as likely to die from PCa when compared to other races ^84–86. Although not specifically addressed in this first report, initial data suggests racial differences between collagen organization and proline hydroxylation in the prostate tumor microenvironment. A specific prostate cancer racial tissue cohort⁸⁷ is currently under evaluation for collagen and proline hydroxylation characterization aligned with clinical, socioeconomic and behavioral data.

Conclusions

A main issue in the treatment of prostate cancer is whether PCa will remain indolent for many years or become malignant. New biomarkers are continually being sought towards prognosis and diagnosis of PCa. Translational and post-translational collagen regulation and re-alignment is a key pathological feature that systematically changes during emerging reactive stroma to tumor growth yet remains mostly unexplored. We show here that very specific collagen types are deposited in inflammatory zones compared to tumor zones. We additionally demonstrate that certain portions of the primary fibrillar collagen structure are post-translationally altered by hydroxylation of proline in tumor zones and are quantitatively decreased in nontumor regions. This highlights that subtle changes in collagen structure contribute to disease. Hydroxylation of proline is a nexus of collagen organization that is responsive to social and economic factors through nutritional, oxidative stress and metabolic factors. It is quite possible that very specific sites of proline modification within the collagen structure could be markers of risk for aggressive PCa and contribute to progression. Multiple studies are ongoing to extend this pilot study and determine how collagen type proteins contribute to risk and influence PCa aggressiveness. Translational and post-translational collagen regulation is a rich source for a new understanding of PCa.

Abbreviations:

PCa -ECM

extracellular matrix

ECM IMS

extracellular matrix imaging mass spectrometry

LUAD

lung adenocarcinoma

MALDI IMS

matrix-assisted laser desorption / ionization imaging mass spectrometry

TMA

tissue microarray