In this study, two publicly available datasets were utilized. The first dataset with identifier “PXD011627” and the second dataset with identifier “PXD010092” were both submitted by Garcia-Puig et al. (2019) in ProteomeXchange. In this study, Garcia-Puig et al. established the zebrafish myocardium ECM enrichment protocol by decellularizing zebrafish ventricular tissue samples (Garcia-Puig et al., 2019). Moreover, this study assessed the dynamic changes in the ECM proteome of regenerating zebrafish hearts using an amputation model. The samples of regenerating heart ventricles were taken at 7-, 14- and 30-days post-amputation (DPA). Ventricular decellularization was done using SDS and Triton-X. The sham model was used as a control in their study.

In their study, MS data acquisition was done differently for the two datasets. Easy-nLC 1,000 (Thermo) was used for liquid chromatography separation. Nanoflex (Thermo) was used as an ESI source and LC-MS/MS was done on Q-exactive HF mass spectrometer (Thermo) with HCD fragmentation and the data-dependent acquisition opted for dataset “PXD011627”. For dataset “PXD010092” a nano HPLC system (Proxeon) coupled with Maxis Impact (Bruker) Q-TOF mass spectrometer was used and each sample of regenerating and sham model heart ventricle was analyzed using LC-MS in duplicates. Furthermore, healthy human heart sample (ECM) mass-spectrometry data from Barallobre-Barreiro et al. (PXD028908) (Barallobre-Barreiro et al., 2021), and mice heart sample (ECM) mass-spectrometry data from Padmanabhan et al (PXD002488) (Padmanabhan Iyer et al., 2016) was reanalyzed for the site-specific collagen PTM identification.

The publicly available datasets “PXD011627” and “PXD010092” were downloaded from ProteomeXchange. The complete *.d folders for “PXD010092” were shared by Prof. Angel Raya’s research group. Dataset “PXD011627” contains 9 (.raw) files and Dataset “PXD010092” contains 8 (.d) files, corresponding to two files (biological replicate) each for sham, 7-, 14- and 30-days post-amputation samples. Dataset “PXD010092” was not compatible with search engines because (.d) files are not readily accepted by many search engines. To overcome this, (.d) files were converted into. mgf and. mzML format using MSConvert with turbocharger filter. These converted files were further used as inputs by search engines MyriMatch (Tabb et al., 2007) and MSFragger (Kong et al., 2017), respectively. We also downloaded 60 (.raw) files from Barallobre-Barreiro et al (PX028908) and 6 (.raw) files from Padmanabhan et al. (PXD002488).

Two different search engines were employed to reanalyze these datasets to identify collagen PTMs from the zebrafish myocardium ECM. All the database searches were performed on the high-performance computing (HPC) facility at the Indian Institute of Technology (IIT)- Mandi. First, a general database search was performed on raw MS data (*.raw files) of dataset “PXD011627” with a complete Danio rerio uniprot database having 62,593 entries, downloaded on 20 October 2020. For the PXD028908 dataset, a general database search was performed on raw MS data (*.raw files) with a complete Homo sapiens uniprot database having 20,396 entries, downloaded on 16 February 2021. For the PXD002488 dataset, a general database search was performed on raw MS data (*.raw files) with a complete Mus musculus uniprot database having 17,090 entries, downloaded on 04 December 2021. Precursor ion tolerance was set at 10 ppm and fragment ion m/z tolerance was allowed up to 20 ppm for zebrafish and human data. However, for the mice dataset, a fragment ion tolerance of 50 ppm was used. Carbamidomethylation (+57.0236) on cysteine was used as static modification and methionine oxidation (+15.9949) and hydroxyproline (+15.9949) were used as dynamic modifications with up to 4 maximum dynamic modifications per peptide. A maximum of 2 missed cleavages were allowed for fully tryptic digestion. MyriMatch searches were performed to generate *.pepXML files for respective datasets. These *.pepXML files were further grouped for PSM matches, peptide, and protein group identification using IDPicker with <1% FDR. After a general database search, the identified list of ECM proteins from zebrafish and humans was exported from IDPicker as a *.FASTA database. This *.FASTA database (separately for zebrafish, humans, and mice) was used as a subset database to perform an in-depth collagen PTM search. In the subset database search, up to 4 missed cleavages for fully tryptic digestion were allowed with a similar precursor (10 ppm) and fragment ion tolerance (20 ppm) tolerance. For mice MS data, a fragment ion tolerance of 50 ppm was used. As mentioned previously, (Basak et al., 2016; Merl-Pham et al., 2019), a “Gly-Xaa-Yaa” based motif-specific PTM search strategy was employed using the MyriMatch motif-specific module for the identification of collagen PTMs. Carbamidomethylation on cysteine was used as static modification and methionine oxidation, hydroxyproline (P! +15.9949), hydroxylysine (GXK! +15.994,916), galactosyl-hydroxylysine (GXK! 178.047738) and glucosylgalactosyl-hydroxylysine (GXK! 340.100,562) were used as dynamic modifications with up to 10 maximum dynamic modifications per peptide. IDPicker was used for grouping the *.pepXML output file by controlling FDR at a 1% level for PSMs, peptides, and proteins. 3-hydroxyproline modifications were only considered in case a proline residue was found to be hydroxylated at the “X” position of a “G-Xaa-HyP” motif in the collagen chains. Furthermore, pLabel was used for manual inspection, analysis, and validation of subset database search PSMs for assigning a specific collagen PTM containing peptide.

mzML files were used for MSFragger mediated database search. In MSFragger based pipeline, the decoy database was generated by a philosopher from the same uniprot database (Danio rerio, 62,593 entries, downloaded on 20 October 2020) as in the case of the MyriMatch general search. Precursor tolerance and fragment ion tolerance were kept at 50 and 25 ppm, respectively. PSM, peptide, and protein level FDR were kept at < 1%, with a maximum of three missed cleavages of fully tryptic peptides. Carbamidomethylation (+57.0236) on cysteine was used as static modification and methionine oxidation (+15.9949), hydroxyproline (+15.9949), N-terminal acetylation (+42.010565), hydroxylysine (+15.994,916), galactosyl-hydroxylysine (+178.047738) and glucosylgalactosyl-hydroxylysine (+340.100,562) were used as dynamic modifications with up to 5 maximum dynamic modification per peptide. IDPicker was used for grouping the *.pepXML output file generated from MSFragger searches by controlling FDR at <1% level for PSMs, peptides, and proteins. Similar to MyriMatch-based workflow, pLabel was used for manual inspection and validation for assigning a specific PSM to identify site-specific collagen PTM containing peptides.

Spectral counts (referred to the total number of peptides detected at ∼1% FDR per protein group) from the results of the MyriMatch database search were exported from IDPicker. These spectral counts were normalized with the total spectral counts of each raw MS file ID. The normalized spectral counts corresponding to unique protein groups were further used to quantitate the relative abundance of collagens in wild-type and regenerating zebrafish hearts. Additionally, the same normalized spectral counts data for collagens from MyriMatch database search results were used to generate a heatmap of relative collagen expression during regeneration. Collagen chains identified with ≥3 normalized spectral counts were only considered for the heatmap representation.

Database search results (*.pepXML files) generated by either MyriMatch and/or MSFragger were parsed through PeptideProphet (TPP pipeline module) for importing the probability scores (0-1) (Pedrioli, 2010). After parsing the *.pepXML files through PeptideProphet, these *.pepXML files were utilized to build a spectral library (.blib) using open-source Skyline (MacLean et al., 2010). A spectral library was used in Skyline for extracting the raw abundance intensity of unmodified and different forms of modified collagen peptides. Occupancy calculation of prolyl hydroxylation sites and microheterogeneity (unmodified, hydroxylated, O-glycosylation of lysines) of lysine sites were performed using MS¹ based targeted extraction pipeline as described previously (Merl-Pham et al., 2019).

GraphPad prism was used for statistical analysis for calculating the dynamics of different collagen site-specific PTM level quantitation. ANOVA was performed and a p value < 0.05 was considered statistically significant.

To identify and quantitate the abundance of collagen chains present in wild-type zebrafish heart ECM, a dual database search engine-based (MyriMatch and MSFragger) comprehensive strategy was implemented. MyriMatch (Tabb et al., 2007) and MSFragger (Kong et al., 2017) were utilized in this study for in-depth identification of collagen chains present in the myocardium ECM of zebrafish from the publicly available datasets (PXD011627, PXD010092) as mentioned earlier. This analytical pipeline identified a total of 36 collagen chains present in the ECM of the zebrafish heart (Figure 1A; Supplementary Table S1). Out of these 36 collagen chains, 17 are commonly identified and 19 collagen chains are newly identified in the ECM of zebrafish heart (Figure 1A, Supplementary Table S1). Previously, Garcia-Puig et al. identified a total of 21 collagen chains from the enriched ECM of zebrafish hearts using this dataset (Garcia-Puig et al., 2019). Since collagens are heavily modified with prolyl hydroxylation, a standard proteomic data analysis pipeline will not be able to identify hydroxylated peptides of collagen chains present in the raw data (Merl-Pham et al., 2019). However, in our analytical pipeline, we included prolyl-hydroxylation as a dynamic modification in the database search yielding a higher number of collagen chain identification. Database searches yielded almost 19,456 and 18,274 total peptide ids from MyriMatch and MSFragger respectively. As expected, almost 61 and 40% of these identified peptides contained prolyl-hydroxylation (Figure 1B, Supplementary Table S2). Thus, it substantiates the use of prolyl-hydroxylation as an important “dynamic modification” in order to explore ECM mass-spectrometry data to increase the identification of collagen chains along with their increased sequence coverage. Normalized spectral counts were used to calculate the abundance of the top 10 collagen chains present in the ECM of wild-type zebrafish hearts (Figure 1C). Both MyriMatch and MSFragger unambiguously yielded COL1A1a, COL1A1b, and COL1A2 chains of collagen 1 to be the highest abundant collagen chains present in the cardiac ECM of zebrafish. These three chains form the triple-helical collagen 1 protomers (Gistelinck et al., 2016), the most abundant protein present in the extracellular matrix of wild-type zebrafish hearts. Apart from Collagen 1, COL5A1, COL5A2a, COL6A3, COL4A1, COL6A1, and COL4A2 chains were found to be in the 10 most abundant collagen chains present in the ECM of WT zebrafish heart. These abundance plots yielded from two different search engines were in agreement highlighting the robustness of the analysis.

The cardiac regeneration process involves ECM remodeling. Being the most abundant component of ECM (Di Lullo et al., 2002), levels of collagens are also altered during regeneration (Garcia-Puig et al., 2019) (Sánchez-Iranzo et al., 2018). As expected, normalized spectral counts of both the search engines show that Collagen 1 (COL1A1a, COL1A1b, and COL1A2) is most abundant in zebrafish heart ECM during 7, 14, and 30 days post-amputation (Supplementary Figure S1). Here, we specifically focused to showcase the relative changes of different collagen chains present in the ECM during regeneration (Figure 1D, Supplementary Table S3). Out of the total 36 collagen chains identified, relative abundances were calculated for 24 collagen chains summarized in the heatmap presented in Figure 1D (raw data for individual chains are provided in Supplementary Table S3). Interestingly, the level of total collagen 1 (summed abundances of all three chains of the triple-helix) decreased (to 0.68 fold, 32% decrease in the level, not significant) at 7 days post-amputation, was slightly increased at 14 days post-amputation (0.86 fold compared to control, 14% decrease in level compared to sham control model) and it also further increased (0.93 fold compared to sham model (control), 6.2% decrease in level compared to sham model) at 30 days post-amputation. Although these changes were not found to be statistically significant, the expression trend corroborated the previous analysis (Garcia-Puig et al., 2019). A similar expression was found for COL5A1 and COL5A2a during the regeneration process. Levels of these 2 collagen chains were first decreased at 7 days post-amputation and then increased at 14 and 30 days post-amputation (Garcia-Puig et al., 2019). The expression of COL4A2, COL5A1, and COL5A2a in our analysis is also similar to the analysis done by Garcia-Puig et al. (2019) in which they found a decrease in the level of COL4A2, COL5A1, and COL5A2a on 7 days post-amputation indicating a feedback post-transcriptional regulation. Furthermore in this analysis, we quantitated the relative abundances of 8 new collagen (COL6A4, COL11A1, COL2A1b, COL28A2a, COL4A5, COL5A3b, COL2A1a, COL1A1) (Figure 1D, marked with red dots) chains present in the myocardium ECM during heart regeneration. The oscillatory dynamics of abundances of these new collagen chains are depicted in the heatmap (Figure 1D, Supplementary Table S3). These expressions of collagen chains highlight the role of collagen in the extracellular remodeling during regeneration.

Collagen PTMs such as O-glycosylation (on hydroxylysine residues) and hydroxylation (on lysine and proline residues) have been shown to be crucial for embryonic development, assembly of ECM fibrils, and cell-matrix interactions (Pokidysheva et al., 2014; Hudson et al., 2015). Collagen molecules get heavily post-translationally modified during the biosynthesis of new chains in the endoplasmic reticulum (Hennet, 2019). Recently, the use of a hydroxyproline-based search strategy using mass-spectrometry data yielded increased coverage for collagen chains (Merl-Pham et al., 2019; Basak et al., 2016; Shao et al., 2020). However, it is challenging to identify the site-specific O-glycosylation sites present in the big collagen chains (Basak et al., 2016). The role of collagen PTMs during tissue regeneration is completely unexplored. Classical amino acid analysis-based approaches could throw insights regarding the composition (including modified amino acids) of these collagen triple helices. However, site-specific identification of these collagen PTMs (hydroxylation, O-glycosylation)–can only be achieved by high-resolution mass spectrometry. So, this study primarily focused on the identification of site-specific collagen PTMs present in the ECM of the zebrafish heart. Furthermore, the dynamics of these PTMs in collagen I (most abundant collagen in heart ECM, a heteromeric trimer constituting three different chains COL1A1a, COL1A1b, and COL1A2) is being showcased during the regeneration of zebrafish heart. A total of 36 collagen chains were identified from the zebrafish heart ECM in this analysis as mentioned previously. Identification of collagen PTMs from a global MS analysis, without biochemical purification, has remained challenging. We had previously developed a MyriMatch-based workflow to identify site-specific collagen PTMs from crude ECM proteome preparations using high-resolution mass spectrometry (Basak et al., 2016). Here, our inhouse collagen PTM identification pipeline (Figure 2) with a dual search engine-based strategy maximized the identification of prolyl-hydroxylation, lysyl-hydroxylation, and O-glycosylation sites present in 23 collagen chains from zebrafish heart ECM (Table 1). We detected a total of 95 3-hydroxyproline, 108 hydroxylysine, 29 galactosyl-hydroxylysine, and 128 glucosylgalactosyl-hydroxylysine sites in 23 collagen chains (Table 1) present in the zebrafish heart ECM. This is the first catalog of site-specific collagen PTMs identified in the ECM of the zebrafish heart. Notably, site-specific collagen PTMs were identified in the highest abundant COL1A1a, along with COL11A1a and COL22A1 spanning across three orders of magnitude (log scale) (see Table 1). This highlights the sensitivity and the depth of dynamic range of our analytical pipeline to explore the PTMs of collagens present in the ECM of zebrafish heart. Collagen I seemed to have a higher number of prolyl-3-hydroxylations and lysyl-hydroxylation sites. Collagen 5 is a minor constituent of fibril assembly (Wenstrup et al., 2004). However, this analysis revealed 12 (P⁶⁸⁶, P⁷⁴⁶, P⁸²⁴, P¹¹¹², P¹¹¹⁵, P¹¹⁶³, P¹¹⁶⁶, P¹¹⁹⁰, P¹¹⁹³, P¹²⁵³, P¹⁴²¹, P¹⁴²⁴) and nine sites (P²⁴¹, P²⁷⁷, P⁵³², P⁶⁶¹, P⁸⁶⁸, P⁹⁰¹, P⁹³⁷, P⁹⁶⁷, P¹¹⁰⁵) of 3-HyP in COL5A1 and COL5A2a, respectively. COL5A2a was then further mapped with 11 sites (K³²⁹, K³³⁸, K⁵⁰⁰, K⁵⁵⁷, K⁵⁷², K⁷⁶¹, K⁷⁶¹, K⁷⁹⁴, K⁶³⁸, K⁸⁰³, K⁸⁸⁷) of glucosyl-galactosyl hydroxylysines. Only one galactosyl-hydroxylysine site (K¹³⁴⁴) was identified in COL5A1. As previously shown in the zebrafish genome (from fin tissue) analysis (Duran et al., 2015), no type III collagen chain is detected in our analysis from the zebrafish cardiac ECM. The absence of type III collagen is a unique feature of zebrafish heart ECM and warrants further scope to investigate the role of collagen III during regeneration. Four different chains of collagen 6 (COL6A1, COL6A2, COL6A3, COL6A4a) were found to harbor O-glycosylation but only one prolyl-3-hydroxylation site (3-HyP⁴⁷⁰ in COL6A2) (Table 1). Although basement membrane collagen IV is not a major constituent of heart ECM; in the case of collagen IV alpha 1 only seven sites (K⁴⁶³, K⁴⁶⁶, K⁸⁸², K⁹⁰⁹, K¹¹⁴⁹, K¹¹⁷⁹, K¹¹⁸²) of glucosyl-galactosyl-hydroxylation were identified. Surprisingly, 24 glucosyl-galactosyl-hydroxylation sites were identified in COL4A2. A total of 4 (P²⁰¹, P²⁰⁴, P²⁹⁴, P²⁹⁷, p¹³³⁸) and 3 (P³³⁹, P⁵⁵⁵, P⁶¹³) prolyl-3-hydroxylation sites were identified in COL4A1 and COL4A2, respectively. It appeared that COL4A2 is way more glycosylated than COL4A1 in the zebrafish cardiac ECM compared to mice (Basak et al., 2016). Interestingly, we have also identified the COL4A5 chain with a common hydroxylysine site harboring microheterogeneity of galactosyl-hydroxylysine and glucosyl-galactosyl-hydroxylysine (K²²⁵). Additionally, many other sites of prolyl-3-hydroxylations, lysyl-hydroxylation and O-glycosylation sites of lysine have been identified in various other collagen types (COL11A1b, COL7A1, COL22A1, etc.; see Table 1) highlighting the advantageous nature of these analyses increasing the depth of site-specific PTM identification coverage in the ECM. Taken together, these results showcased the comprehensive mapping and site-specific identification of many collagens present in the zebrafish heart ECM.

Collagen I is the highest abundant ECM protein present in the zebrafish hearts (Gistelinck et al., 2016). The composition of the heterotrimer of the collagen I chain forming a protomer (triple helices) is different compared to mammals (Gistelinck et al., 2016). In mammals, two alpha 1 chain and one alpha 2 chain form the protomer. However, in zebrafish, three different genes (COL1A1a, COL1A1b, and COL1A2) code for three different collagen I chains (collagen alpha 1a, collagen 1 alpha 1b (alpha 3), and collagen 1 alpha 2), respectively. High expression of these three genes has been documented in adult zebrafish tissues as well as during embryonic and larval development indicating equimolar stoichiometry in forming the protomers (Gistelinck et al., 2016). However, different glycosylated forms of COL1A1a have been found in zebrafish embryos indicating altered PTM levels during development (Gistelinck et al., 2016). Thus, tissue-specific mapping of collagen I site-specific PTMs are of the highest importance in order to further delineate their function. Here, we have comprehensively mapped the site-specific prolyl-hydroxylation, lysyl-hydroxylation, and lysine O-glycosylation in three different chains (COL1A1a, COL1A1b, and COL1A2) of collagen I in zebrafish heart ECM.

Identification of prolyl-3-hydroxylation sites in collagen chains of heart ECM provided new insights into the tissue-specific variation of collagen molecules providing structural support in ECM. Latest studies have revealed that the absence of prolyl-3-hydroxylation in one site (P¹¹⁶⁴) of COL1A1 is associated with osteogenesis imperfecta. Mutation in the prolyl-hydroxylases can cause the formation of dysfunctional collagen (without a 3-HyP site) as found in osteogenesis imperfecta disease conditions (Cabral et al., 2007). Furthermore, Merl-Pham et al. recently showed that the occupancy of 3-HyP⁷⁷¹ of human COL1A1 was increased during TGF-beta treated fibrotic conditions in primary human lung fibroblast (Merl-Pham et al., 2019). These indicate that changes in the occupancy of the 3-HyP level can have important consequences on the functionality of the collagen molecules present in the ECM. Therefore, we check the occupancy level of 3-HyP in collagens during the regeneration of the zebrafish heart. We utilized our proteomic pipeline and quantitated the occupancy levels of collagen PTMs during zebrafish heart regeneration in the sham model used as control, 7-, 14- and 30-day post-amputation. The critical prolyl-3-hydroxylation site (P¹¹⁶⁴) involved in osteogenesis imperfecta had been previously shown to be fully 3-hydroxylated in mouse skin fibroblast (Morello et al., 2006). Surprisingly, this conserved site present in zebrafish (3-HyP¹¹⁴⁸) control cardiac ECM is only about half (48%) 3-hydroxylated (Table 2; Figure 7). Furthermore, the occupancy of this site decreased to ∼7% in the 30 DPA regenerated zebrafish heart. The occupancy level of two other 3-HyP sites (P⁸⁶⁹, P⁸⁷⁸) of COL1A1a decreased at 30 DPA compared to control and 7 DPA occupancy (Table 2; Figure 7). Only 3-HyP⁷⁰⁷ occupancy level was increased (7.5–20%) at 30 DPA regenerated zebrafish heart compared to sham. This increased occupancy may favor the assembly of ECM by enhancing collagen fibril formation by forming a water bridge with the available free carbonyl group of the subsequent collagen chain. However, a decrease in occupancy of other 3-HyP sites may favor the reduced interaction of deposited collagen molecules in ECM thereby favoring the transition to regeneration from the fibrotic milieu. Although, the exact functional role of these site-specific 3-HyPs needs to be experimentally assessed. A total of five 3-hydroxylation sites (P^{404,554,914,1031,1109}) occupancy levels were quantitated in COL1A1b. Out of these, 4 sites (P^{404,914,1031,1109}) showed the trend of increased occupancy level at 30 DPA regenerated zebrafish heart. We quantitated the occupancy level of 4 sites in COL1A2. Most of these sites were having <10% prolyl-3-hydroxylation. Interestingly, we identified a cluster of 3-HyP sites (P^1195,1201) in COL5A2a to be highly 3-hydroxylated (24%) after 7 DPA samples compared to only 1.7% occupancy in control. The occupancy of this 3-HyP cluster showed dynamic changes at 14 DPA (55%) and further reverting to 30% at 30 DPA, in the regenerated zebrafish heart. The site-specific variation in the occupancy level of 3-hydroxyproline could be due to varied gene expression of specific enzymes expressed in the zebrafish heart during post-amputation regeneration. A recent report from Sanchez-Iranzo et al. showed the expression of prolyl-3-hydroxylase-3 (P3H3) and prolyl-3-hydroxylase-4 (P3H4) decreased at 7 days post cryoinjury (log fold change -0.4,-0.2, respectively), in the zebrafish heart. However, in the same RNA seq analysis, the expression level of prolyl-3-hydroxylase-1 (P3H1) and prolyl-3-hydroxylase-2 (P3H2) were found to be increased (log fold change1.41,1.75 respectively) in the zebrafish heart at 7 days post-cryoinjury model. Following the regeneration in the same model till 60 days post-cryoinjury, the expression level of P3H2 and P3H4 increased (log fold change 0.94, 1.73 respectively) and the expression of P3H1 and P3H3 decreased (log fold change −0.56, −1.13) respectively highlighting the dynamic changes in the expression of these isoforms of prolyl-3-hydroxylases. However, the enzyme-substrate specificity for these different isoforms of prolyl-3-hydroxylases is not known yet. Thus, ascertaining a direct correlation between the expression level of one enzyme isoform with the substrate (occupancy level of site-specific modifications) specificity in this dataset is difficult. But, the varied expression of these enzymes during the post-cryoinjury regeneration model in zebrafish heart further supports the specific regulation of site-specific dynamic prolyl-3-hydroxylation changes in collagens deposited in the zebrafish heart ECM during regeneration. The dynamics of prolyl-3-hydroxylation occupancy in collagen molecules deposited in ECM during zebrafish heart regeneration are reported here for the first time.

In collagens, lysine hydroxylation and glycosylation play important role in collagen assembly and fibril formation (Terajima et al., 2014; Terajima et al., 2016). Musculoskeletal defects and cerebral small vessel disease are caused due to lack of collagen glycosylation (Geister et al., 2019; Miyatake et al., 2018; Kivirikko et al., 1973). A total of 23 HyK, 9 G-HyK, and 11 GG-HyK sites in zebrafish COL1A1a were identified from wild-type (Sham) zebrafish heart ECM (Figure 3). Classical amino acid analysis has revealed that fibrillar collagens are less glycosylated compared to basement membrane collagen IV (Terajima et al., 2014; Perdivara et al., 2013; Yamauchi and Sricholpech, 2012). However, collagen glycosylation in fibrillar collagen is important for crosslinking mediated fibrillar assembly. Microheterogeneity of lysine sites has been determined in collagens chains (Yamauchi et al., 1982). A single lysine site of a collagen chain could be present in the form of unmodified, hydroxylysine (HyK), galactosyl-hydroxylysine (G-HyK), and glucosyl-galactosyl-hydroxylysine (GG-HyK). This microheterogeneity was found to be highly dynamic in the COL1A1a chain during zebrafish heart regeneration. We detected microheterogeneity on a total of eight lysine sites in the COL1A1a chain. Figure 5 and Figure 6 respectively show the site-specific mapping of lysine O-glycosylation sites in COL1A1b and COL1A2 chains from wild-type zebrafish heart ECM. In COL1A1b a total of 24 HyK, 4 G-HyK (K²⁶⁴, K²⁷³, K⁸⁴⁹, K¹⁰²⁰), and 4 GG-HyK (K²⁶⁴, K²⁷³, K³⁹⁹, K⁸⁴⁹) sites were identified. Three of these sites showed microheterogeneity in glycosylation patterns. Similarly, a total of 21 HyK, 6 G-HyK (K⁷⁴, K¹⁸⁸, K²⁵⁴, K³⁴⁴, K³⁵⁰, K⁶⁴⁴), and 9 GG-HyK (K⁷⁴, K¹⁶⁷, K¹⁸⁸, K²⁵⁴, K²⁹⁹, K⁵⁷⁸, K⁶⁴⁴, K⁶⁴⁷, K⁹³⁵) sites were identified in COL1A2 chain for the first time from zebrafish heart ECM. Four of these sites showed microheterogeneity in different glycosylation moieties (galactosyl or glucosyl-galactosyl). Overall, our analysis mapped the different lysine O-glycosylation sites in three different chains of collagen 1 chains. Collagen 1A1a was found to be less deposited during the regeneration of the zebrafish heart by our as well as Garcia-Puig et al. analysis (Garcia-Puig et al., 2019). Furthermore, we wanted to assess the changes in O-glycosylation of collagen I chains during zebrafish heart regeneration.

In order to assess the changes in the microheterogeneity of lysine O-glycosylation sites on collagen I during regeneration of zebrafish heart, the Skyline-based targeted MS¹ quantitative method was integrated into our pipeline as described previously (Merl-Pham et al., 2019). The relative microheterogeneity analysis of site-specific lysine in three different chains of collagen I was performed. A total of 12 sites of lysine microheterogeneity from three different chains of collagen I was assessed (Table 3). Figure 8 represents an example of a COL1A1a peptide ¹⁰¹¹DGAAGPKGDRGETGPSGTPGAPGPPGAAGPIGPAGK¹⁰⁴⁶ eluting at 42.3 min during the chromatographic separation as non-glycosylated form with three 4-HyP (P^{1029,1032,1035}) sites. Hydroxylation of K¹⁰¹⁷ resulted in an early elution of this peptide at ∼42 min in the C18 column-based separation. Further, the glucosyl-galactosyl-hydroxylysine form of K¹⁰¹⁷ was further eluted earlier (at 41.2 min) than the hydroxylation form. We determined the microheterogeneity of this site (K¹⁰¹⁷) across control, 7-day post-amputation (DPA), 14 DPA, and 30 DPA samples across zebrafish heart regeneration. Interestingly, the glucosyl-galactosyl-hydroxylation of K¹⁰¹⁷ increased significantly (p < 0.01) to 15.4% in the 30 DPA regenerated heart as compared to control (2.74%), (3.35%) 7 DPA. and 14 DPA (10.5%) samples (Figure 8; Table 3). The increase in the micro heterogenic distribution of O-glycosylation the in K¹⁰¹⁷ site was evident by a decrease in the HyK level during regeneration (Figure 8; Table 3). Using a similar strategy, we found K⁸⁴⁶ (COL1A1a) to be specifically present in a full glycosylated form (either G-HyK or GG-HyK). The GG-HyK⁸⁴⁶ was decreased at 30 DPA regenerated zebrafish heart. We also quantitated the microheterogeneity of three lysine sites (K^264,273,849) in COL1A1b and 2 sites in COL1A2 (K^254,644) (see Table 3). The micro-heterogenic distribution of G-HyK²⁶⁴ of COL1A1b was found to have an increasing trend at 30 DPA. The changes in the level of hydroxylation and O-glycosylation of lysine sites present in collagen I could be potentially due to the gene expression changes of lysyl-hydroxylases and glycosyltransferases during zebrafish heart regeneration. In a similar model of zebrafish heart regeneration, proteomic analysis performed by Ma et al. (2018) revealed a significant elevation of LH2 levels at 2 DPA (Supplementary Figure S4). However, at 14 DPA the LH2 protein level stabilizes to sham levels. Similar oscillatory gene expressional observations of the genes (PLOD1a, PLOD2, and PLOD3) coding for different isoforms of lysyl-hydroxylases has been documented in the zebrafish heart cryoinjury model (Sánchez-Iranzo et al., 2018). Global gene expression analysis revealed increased expression of these three genes in the zebrafish heart at 7 days post-injury. However, the level of gene expression of PLOD1a, PLOD2, and PLOD3 decreases at 60 days post injury in the regenerated heart (Sánchez-Iranzo et al., 2018). These micro-heterogenic site-specific glycosylation profiles point towards the differences present in three different chains of collagen I that may play an important role in the ECM remodeling mediated zebrafish heart regeneration.