Proteomic profile in Pseudomyxoma peritonei

As we mentioned above, molecular characterisation of PMP is scarce and there are only a few papers showing some mucin isoforms (mainly MUC2 and MUC5AC) analysed by Western blot [15, 16]. Importantly, the results of these experiments were highly smeared band patterns due to the high levels of mucin proteins and their interactions. Consequently, these protein extracts are unsuitable for high throughput proteomic analyses.

To overcome this problem, we developed the first method specifically adapted to mucin samples to extract and isolate total protein with a high quality and purity for their use in proteomic analyses (Fig. 1). Thus, soft and hard mucin samples from low and high-grade PMP tumours processed by this method, called AMIPROM, were used to determine the proteomic profile of PMP in order to identify intracellular pathways altered in PMP as well as potential tumour cell markers through the application of quantitative proteomics. Healthy samples from an appendectomy performed for an unrelated medical condition or normal colon tissue were used as controls due to the lack of appropriate non-tumour mucin samples (Table 1).

Fig. 1

Schematic of the adapted method to isolate proteins from mucin. Adapted method to isolate proteins from mucin (AMIPROM): samples of soft and hard mucin obtained from LG-PMP and HG-PMP are cut into small pieces and homogenised using ultrasound. The homogenate is then centrifuged and the supernatant is collected and filtered. The homogenate is then subjected to liquid chromatography (LC) using two different columns to reduce the most abundant glycoproteins, IgGs, and albumin from the sample. The LC-derived protein extract is subjected to mass spectrometry analysis using an unbiased targeted proteomic approach with SWATH-MS

Nano-LC/MS–MS equipped with SWATH for label-free quantitative proteomics allowed us to generate the first proteome profile described in PMP. Considering all these proteins, a partial least squares discriminant analysis (PLS-DA) revealed a clear discrimination pattern between the proteomic profile of the soft mucin, hard mucin and control tissue samples (Fig. 2A). Furthermore, using a log2-fold change difference > 1 and a p-value < 0.05 to determine differentially expressed proteins compared to the control tissues, we identified 93 up-regulated and 243 down-regulated proteins in the soft mucin samples and 86 up-regulated and 27 down-regulated proteins in the hard mucin samples (Fig. 2B). Considering that mucins are the main proteins that characterise this entity, we then proceeded to identify the different mucin isoforms detected in the analysed PMP subtypes and detected a differential pattern of mucin isoforms between the soft and hard mucin samples (Fig. 2C). Specifically, we detected MUC2, MUC5AC, MUC5B, MUC6, and MUC13 in the soft mucin samples compared to the control tissues, with MUC2, MUC5AC, MUC5B, and MUC6 being significantly upregulated in these samples (Fig. 2C; left graph) and MUC1, MUC2, MUC4, MUC5AC, MUC5B, and MUC13 in the hard mucin samples, with MUC2, MUC5AC, and MUC13 being significantly upregulated in these samples (Fig. 2C; right graph). MUC2, although reduced by the glycoprotein affinity column, was the most highly expressed mucin isoform in all cases.

Fig. 2

Proteomic analysis of soft and hard mucin obtained from LG-PMP and HG-PMP samples compared to control samples. A Partial least squares discriminant analysis (PLS-DA) of the proteome profile between soft (left; n = 14) and hard (right; n = 15) mucin samples and control tissues (n = 10). B Volcano plots showing Log2 Fold Change expression vs –log10 (p-value) of differentially expressed proteins with a p-value < 0.05 and an absolute Log2 Fold Change > 1 in the same sample set. Green colour indicates up-regulated proteins and red colour indicates down-regulated proteins. C Protein expression levels of mucin isoforms identified in soft mucin (green bars) and hard mucin (blue bars) mucin samples from PMP compared to control tissue (set to 100%; dashed line). D, G PLS-DA analysis of the proteome profile of low and high-grade soft mucin (SM) (D) and hard mucin (HM) (G) samples compared to control tissues. E, H Volcano plots showing Log2 Fold Change expression vs –log10 (p-value) of differentially expressed proteins with a p-value < 0.05 and an absolute Log2 Fold Change > 1 in low (left panel) and high-grade (right panel) soft mucin (E) and hard mucin (H) samples compared to control tissues. F, I Protein expression levels of identified mucin isoforms in low (light bars) and high-grade (dark bars) soft mucin (F) and hard mucin (I) samples compared to control tissue (set to 100%; dashed line). * p < 0.05 and, ** p < 0.01, *** p < 0.001

We then performed an analysis to compare LG and HG PMP soft mucin samples, which revealed that PLS-DA could perfectly separate both groups and clearly distinguish them from the control samples (Fig. 2D). Considering only differentially expressed proteins compared to the control tissue samples, volcano plots showed 72 up-regulated and 220 down-regulated proteins in the LG-PMP soft mucin samples (Fig. 2E-left panel) and 19 up-regulated and 380 down-regulated proteins in the high-grade soft mucin samples (Fig. 2E-right panel). As we mentioned above, although we found a different pattern of mucin isoforms between soft and hard mucin samples, no significant differences were found between the mucin isoforms identified in the LG and HG soft mucin samples (Fig. 2F). In the same line, PLS-DA performed on LG and HG hard mucin samples revealed a clear discrimination pattern between the overall proteomic profile of these samples and that of the control tissues (Fig. 2G). Notably, as shown by the volcano plots (Fig. 2H), the number of differentially expressed proteins was significantly higher for soft mucin than for hard mucin compared to the control tissues. Furthermore, the expression levels of MUC5AC and MUC13 were significantly upregulated in LG but not in HG hard mucin samples compared to control tissues. However, there were no statistical differences in their expression between LG and HG hard mucin samples (Fig. 2I).

Finally, we analysed the extracts captured in the two columns used during the protocol to evaluate how many proteins were depleted/reduced along with glycoproteins, albumins, and immunoglobulins. To do this, we performed an electrophoresis on the extracts from the columns and the major bands were excised and analysed by mass spectrometry (see Supplemental Methods). The results are shown in Figure S1, Table S1 and Supplemental Excel 1. The results were analysed using a cut-off of the 1% spectral count value of the most abundant protein identified for each column. Briefly, a total of 85 proteins were identified in the extracts captured on the HiTrap Con A 4B column, of which 82 were secreted proteins (including glycoproteins), 15 were cell membrane proteins and 7 were cytosolic proteins. It is important to note that most of the cell membrane and cytosolic proteins are also considered secreted proteins. For the HiTrap Albumin and IgG depletion column, a total of 16 proteins were identified, all of which were secreted proteins (including albumin and immunoglobulins) and 7 of which were cell membrane proteins. In addition, 10 out of 15 (66.7%) of the identified cell membrane proteins captured on the HiTrap Con A 4B column and all (100%) of the identified cell membrane proteins captured on the HiTrap Albumin and IgG depletion column were immunoglobulins, which were one of the targets to be depleted.

Soft and hard mucin tissues are highly similar at functional level

To explore the functional relevance of soft and hard mucin PMP tissues, all differentially expressed proteins found in soft and hard mucin compared to control tissues were analysed using the Metascape database. In this sense, we found an 84.2% of unique proteins in soft mucin compared to controls and 53.1% in hard mucin samples compared to controls, with only a small fraction of proteins shared between both comparisons, as illustrated by the Circos plot (Fig. 3A). Interestingly, although most of the differentially expressed proteins were different between the two comparisons, they shared a high number of enriched pathways and processes (Fig. 3B, C). Indeed, from the top 100 enriched terms included in the heatmap (where the colour scale represents statistical significance), most of them were significantly altered in both comparisons (pattern 1), with the terms “Golgi lumen” and “Extracellular vesicles in the crosstalk of cardiac cells” enriched exclusively in the comparison hard mucin vs. control (pattern 3) and the terms “Cellular aldehyde metabolic process”, “Biological oxidations”, “Negative regulation of cell migration”, among others, enriched exclusively in the comparison soft mucin vs. control (pattern 2) (Fig. 3C). In general, an elevated number of enriched processes were related with extracellular matrix (including “extracellular matrix organisation”, “Naba matrisome associated”, “Naba core associated”, “collagen binding”, “focal adhesion”, etc.), regulation of cytoskeleton (including “actomyosin structure organisation”, cortical cytoskeleton organisation”, “structural constituent of cytoskeleton”, etc.), metabolism (including “Glycolysis/Gluconeogenesis”, “Carbon metabolism”, “metabolism of carbohydrates”, “pyruvate metabolism and Citric Acid (TCA) cycle”, etc.), and signalling pathways highly related with cancer (including “VEGFA VEGFR2 signalling”, “EPH-Ephrin signalling”, “signalling by Rho GTPases”, “regulation of MAPK cascade”, etc.).

Fig. 3

Visualisation of the functional enrichment meta-analysis based on two protein lists [soft mucin (SM) vs. control (CTRL) and hard mucin (HM) vs. control (CTRL)]. A Circos plot visualisation of the overlap between the protein lists (SM vs. CTRL and HM vs. CTRL). Each candidate protein is assigned to a point on the arc of the corresponding protein list(s). Proteins common to both lists are connected by purple curves. B Circos plot visualisation with blue curves connecting those candidate proteins that have different identities but share an enriched pathway/process, i.e. they represent the functional overlap between the protein lists. C Heatmap showing the top 100 enrichment clusters, one row per cluster, using a discrete colour scale to represent statistical significance. Grey colour indicates the lack of enrichment for that term in the corresponding gene list, light yellow colour indicates the boundary between significance and insignificance, deep yellow colour indicates a high degree of significance. D Enrichment network visualisation for results from the two protein lists, where nodes are represented by pie charts indicating their associations with each input list. Cluster labels have been added manually. The colour code represents the identities of the protein lists, where blue indicates SM vs. CTRL and red indicates HM vs. CTRL. E Visualisation of the PPI network and MCODE components identified from the combined protein list, where each node represents a protein with a pie chart encoding its origin. Colour codes for pie sectors represent a protein list

To facilitate the understanding of pathways/processes that are shared between the two comparisons, additional visualisations were developed. For example, an enrichment network visualisation including the results from both comparisons confirmed the same results as the heatmap, showing an overlap of biological processes by both soft and hard mucin tissues, as these proteins probably are likely to capture different parts of the same biological processes (Fig. 3D). Furthermore, a protein–protein interaction (PPI) network was also generated to elucidate common/selective functional clusters. In this sense, 9 different clusters were identified based on the MCODE algorithm, most of which were shared between both comparisons, and only cluster 9 (related to regulation of insulin-like growth factor transport and uptake) was specifically enriched in soft mucin vs. control. Cluster 6 (related to signalling by ROBO receptors and metabolism of amino acids) was also mainly enriched in soft mucin vs. control (Fig. 3E and Table S2).

Low and high-grade soft mucin tissues share a high number of biological processes/pathways

Next, we wanted to explore the functional relevance of LG and HG soft mucin PMP tissues. Therefore, we compared all differentially expressed proteins found in LG and HG soft mucin compared to control tissues and in LG compared to HG using the Metascape database. In this case, the Circos plot showed an elevated number of differentially altered proteins shared between the three comparisons, with only 28.4% of unique proteins in LG vs. control, 40.6% of unique proteins in HG vs. control and 20.6% of unique proteins in LG vs. HG (Fig. 4A). Interestingly, 94% of the proteins detected in the LG vs. HG comparison were common to the HG vs. control protein list, suggesting that most of these proteins are specifically altered in HG. In terms of functional enrichment, an increased number of biological processes/pathways were found to be altered in both LG vs. control and HG vs. control, which was corroborated by comparing all differentially altered proteins in LG vs. HG, showing a high number of common enriched terms between LG and HG (grey colour; pattern 2) (Fig. 4B-C). Among these common terms, we found processes and pathways related to protein regulation (e.g. “unfolded protein binding”, “protein homodimerisation activity”, “positive regulation of protein localisation”, regulation of protein stability”, “negative regulation of protein polymerisation”, etc.), metabolism (e.g. “biological oxidations”, “cellular aldehyde metabolic process”, “generation of precursor metabolites and energy”, “pyruvate metabolism and citric acid (TCA) cycle”, etc.), and extracellular matrix (e.g. “glycosaminoglycan binding”, “extracellular matrix structural constituent”, “cell–cell junction”, “collagen-containing extracellular matrix”, etc.). In addition, we found some enriched terms that were differentially altered between LG and HG (pattern 1), which could be used to understand the differences between these two disease grades. Some of these terms were related to the regulation of the cytoskeleton (e.g. “establishment or maintenance of cell polarity”, “cortical cytoskeleton organisation”, “cytoskeleton-dependent cytokinesis”, “endocytosis”, “regulation of vesicle-mediated transport”, “regulation of actin-filament organisation”, etc.), signalling pathways highly related with cancer (e.g. “G13 signalling pathway”, “nuclear receptors meta pathway”, VEGFA VEGFR2 signalling”, and “gene and protein expression by JAK-STAT signalling after Interleukin-12 stimulation”), metabolism (e.g. “glycolysis/gluconeogenesis”, “monocarboxylic acid metabolic process”, “small molecule catabolic process”, “isomerase activity”, “peptidase activity”, etc.), and other important extracellular matrix-related pathways such as “Proteoglycans in cancer”. The pathway “amino sugar and nucleotide sugar metabolism” was also found to be altered in LG vs. HG and in HG vs. control, but not in LG vs. control, suggesting that this pathway might be specifically altered in HG-PMP samples (Pattern 3; Fig. 4C). Consistent with these results, the enrichment network visualisation showed the same overlapping pattern, with most of the enriched terms shared between HG vs. control (red) and LG vs. control (blue), and only a few of them also shared with the LG vs. HG comparison (green) (Fig. 4D). Furthermore, the PPI network revealed 7 different functional clusters, all of them related to the cytoskeleton, signalling pathways and metabolism, and all shared between the three protein lists (Fig. 4E; Table S3).

Fig. 4

Visualisation of the functional enrichment meta-analysis based on three protein lists [LG vs. control (CTRL), HG vs. control (CTRL) and LG vs. HG)] in soft mucin samples compared to control tissues. A Circos plot visualising the overlap between the protein lists. Each candidate protein is assigned to a point on the arc of the corresponding protein list(s). Proteins common to both lists are connected by purple curves. B Circos plot visualisation with blue curves connecting those candidate proteins that have different identities but share an enriched pathway/process, i.e. they represent the functional overlap between protein lists. C Heatmap showing the top 100 enrichment clusters, one row per cluster, using a discrete colour scale to represent statistical significance. Grey colour indicates the lack of enrichment for that term in the corresponding gene list, light yellow colour indicates the boundary between significance and insignificance, deep yellow colour indicates a high degree of significance. D Enrichment network visualisation for results from the three protein lists, where nodes are represented by pie charts indicating their associations with each input list. Cluster labels have been added manually. Colour code represents the identities of protein lists, where blue indicates LG vs. CTRL, red indicates HG vs. CTRL and green indicates LG vs. HG. E Visualisation of the PPI network and MCODE components identified from the combined protein list, where each node represents a protein with a pie chart encoding its origin. Colour codes for pie sectors represent a protein list

Identification of functional enrichment terms specifically associated with LG or HG hard mucin tissues

We then examined the functional relevance in LG and HG hard mucin tissues as we did above for soft mucin tissues. Circos plot analysis revealed a higher number of differentially altered proteins in the LG vs. HG comparison in hard mucin compared to the number of differentially altered proteins found in soft mucin (Figs. 4A and 5A). In general, the number of shared proteins between the three protein lists was lower in hard mucin tissues than in soft mucin tissues, with a higher number of unique proteins (48% of unique proteins in LG vs. control, 61.3% in HG vs. control and 36.9% in LG vs. HG) (Fig. 5A). The functional enrichment derived from these protein lists generated a heatmap with five different patterns according to the distribution of the enriched terms. Patterns 1 and 4 (grey colour) show all common enriched terms found between LG and HG, but differentially altered compared to control tissues. This is also illustrated in the Circos plot (Fig. 5B, C). Pattern 2 includes all enriched terms that are able to discriminate LG from HG hard mucin tissues. Among these terms we found processes/pathways related to metabolism (e.g. “amino sugar and nucleotide sugar metabolism”, “small molecule catabolic process”, “glycolysis/gluconeogenesis”, etc.), cellular homeostasis and detoxification (e.g. “detoxification of reactive oxygen species”, “cellular detoxification”, “programmed cell death” and immune response (e.g. “acute-phase response”, “innate immune response”, “complement and coagulation cascades”, etc.). In addition, patterns 3 and 5 revealed differentially enriched terms that are specifically associated with LG or HG hard mucin PMP tissues. For example, pattern 3 included important cellular activities such as “peptidase activity”, “lyase activity”, “hydrolase activity”, and also other processes such as “organic acid binding”, “insulin-like growth factor binding” and “protein-folding chaperone binding”, which may be specifically associated with LG hard mucin tissues, as they were altered in LG vs. HG and LG vs. control, but not in HG vs. control. In the same line, pattern 5 included important metabolic processes such as “pyruvate metabolism” and “cellular aldehyde metabolic process” and others such as “protein tetramerization” and “intramolecular phosphotransferase activity”, which could be associated with HG hard mucin tissues (Fig. 5C).

Fig. 5

Visualisation of the functional enrichment meta-analysis based on three protein lists [LG vs. control (CTRL), HG vs. control (CTRL) and LG vs. HG] in hard mucin samples compared to control tissues. A Circos plot visualising the overlap between protein lists. Each candidate protein is assigned to a point on the arc of the corresponding protein list(s). Proteins common to both lists are connected by purple curves. B Circos plot visualisation with blue curves connecting those candidate proteins that have different identities but share an enriched pathway/process, i.e. they represent the functional overlap between protein lists. C Heatmap showing the top 100 enrichment clusters, one row per cluster, using a discrete colour scale to represent statistical significance. Grey colour indicates the lack of enrichment for that term in the corresponding gene list, light yellow colour indicates the boundary between significance and insignificance, deep yellow colour indicates a high degree of significance. D Enrichment network visualisation for results from the three protein lists, where nodes are represented by pie charts indicating their associations with each input list. Cluster labels have been added manually. Colour code represents the identities of protein lists, where blue indicates LG vs. CTRL, red indicates HG vs. CTRL and green indicates LG vs. HG. E Visualization of the PPI network and MCODE components identified from the combined protein list, where each node represents a protein with a pie chart encoding its origin. Colour codes for pie sectors represent a protein list

As before, to better understand the pathways/processes that are shared between the three comparisons, we generated the enrichment network visualisation and the PPI network. In the enrichment network, we found that most of the enriched terms were shared by the three protein lists, as they mainly represented patterns 1 and 2 from the heatmap. Nevertheless, processes such as “vesicle-mediated transport” and “programmed cell death” were mainly associated with LG vs. control and LG vs. HG, suggesting that these processes were mainly altered in LG-PMP tissues (Fig. 5D). Furthermore, the PPI network revealed 12 different functional clusters, all of them related to the proteasome, collagens, mRNA processing, complement activation and secretory granule lumen. Although no cluster was found to be specific to any protein list, clusters 2, 8 and 10 were mainly enriched in HG vs. control and LG vs. HG protein lists (Fig. 5E; Table S4).

Validation of MUC13 alteration in soft and hard mucin samples of Pseudomyxoma peritonei

As we mentioned above, mucin isoforms are the main entity characterising this pathology. Interestingly, we observed a different mucin isoform pattern between soft and hard mucin tissues, with MUC13 being the only membrane-associated mucin found to be altered (Fig. 2), making it a potential candidate to be considered as a cellular tumour marker or cellular therapeutic target. For these reasons, MUC13 was quantified by Western blot and an enzyme-linked immunosorbent assay (ELISA) in soft and hard mucin samples obtained from patients with PMP (Fig. 6). In the Western blot analysis (Fig. 6A), MUC13 showed significantly higher expression in LG and HG soft mucin and in LG hard mucin tissues compared to control tissues, and was also found to be overexpressed in LG and HG soft mucin compared to hard mucin. In addition, MUC13 was quantified by ELISA in a larger cohort of PMP samples and was found to be overexpressed in LG soft mucin samples compared to control tissues (Fig. 6B). Importantly, MUC13 was not found in the depleted extracts (Supplemental Excel 1).

Fig. 6

Validation of MUC13 expression levels in PMP. A Protein expression levels of MUC13 in soft mucin (SM – green bars) and hard mucin (HM – blue bars) [low (LG-PMP; n = 4) and high-grade (HG-PMP; n = 4)] compared to control tissues (n = 4; no tumoral appendix) evaluated by Western Blot. The arbitrary densitometric unit (ADU) for each protein was normalised by the Total Protein Normalisation (TPN) value. B Cohort validation of MUC13 by ELISA quantification SM and HM samples [low (LG-PMP) and high-grade (HG-PMP)] compared to control tissues (n = 16) (number of PMP samples analysed is indicated in the bars of the graph). One-way ANOVA analysis was performed for multiple comparisons (LG and HG-PMP vs Control). * p < 0.05 and *** p < 0.001