Establishment and validation of highly accurate formalin-fixed paraffin-embedded quantitative proteomics by heat-compatible pressure cycling technology using phase-transfer surfactant and SWATH-MS

The purpose of this study was to establish a quantitative proteomic method able to accurately quantify pathological changes in the protein expression levels of not only non-membrane proteins, but also membrane proteins, using formalin-fixed paraffin-embedded (FFPE) samples. Protein extraction from FFPE sections of mouse liver was increased 3.33-fold by pressure cycling technology (PCT) and reached the same level as protein extraction from frozen sections. After PCT-assisted processing of FFPE liver samples followed by SWATH-MS-based comprehensive quantification, the peak areas of 88.4% of peptides agreed with those from matched fresh samples within a 1.5-fold range. For membrane proteins, this percentage was remarkably increased from 49.1 to 93.8% by PCT. Compared to the conventional method using urea buffer, the present method using phase-transfer surfactant (PTS) buffer at 95 °C showed better agreement of peptide peak areas between FFPE and fresh samples. When our method using PCT and PTS buffer at 95 °C was applied to a bile duct ligation (BDL) disease model, the BDL/control expression ratios for 80.0% of peptides agreed within a 1.2-fold range between FFPE and fresh samples. This heat-compatible FFPE-PCT-SWATH proteomics technology using PTS is suitable for quantitative studies of pathological molecular mechanisms and biomarker discovery utilizing widely available FFPE samples.

broken lines represent 1.5-fold difference. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from fresh samples within a 1.5-fold range.
[f] This graph was generated using the data of the reported PCT-SWATH study 3 . Cytosolic proteins were selected according to the subcellular location information in the Uniprot human proteome database. Peptide peak areas were compared between FFPE and frozen human benign prostatic tissues obtained from the same resected tissue of patient 15, who showed the best agreement (as an inaccuracy value) between FFPE and frozen peptide peak areas among the 24 patients. The peptide samples of FFPE and frozen prostatic tissues were prepared with PCT treatment using urea buffer without heating, and then measured in the SWATH mode. The in silico peptide selection criteria were not applied. The broken lines represent 1.5-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE samples lie within a 1.5-fold range of those from frozen samples. [g, and h] Inaccuracy (g) and CV values (h) corresponding to panels a, b, c, and d were calculated as described in Materials & Methods, and the inaccuracy (g) was also calculated from panels e (gray column) and f (hatched column). Each column represents the mean ± SEM (n = 716-2329 peptides; number commonly detected in FFPE and fresh samples under each experimental condition). *p < 0.001, significant difference between two groups (Bonferroni-corrected Student's t-test). The presence or absence of a transmembrane region in each membrane protein was established from the Uniprot database.

Supplementary figure 7 Effect of PCT treatment and SWATH analysis on comprehensive quantification of BDL-induced changes in expression levels of cytosolic proteins in FFPE sections [a, b, c, and d]
Cytosolic proteins were selected according to the subcellular location information in the Uniprot mouse proteome database. The BDL-induced changes in expression level (BDL/control ratio) were compared between FFPE and fresh samples of mouse liver at the peptide level. Peptide samples from FFPE and fresh livers were prepared with PCT treatment and measured in the SWATH mode (a, PCT(+)-SWATH), without PCT treatment and measured in the SWATH mode (b, PCT(-)-SWATH), with PCT treatment and measured in the shotgun mode (c, PCT(+)-Shotgun), or without PCT treatment and measured in the shotgun mode (d, PCT(-)-Shotgun). Data analysis was carried out as described in Supplementary figure 3. The data were taken from Supplementary tables 7, 8, 9 and 10. Each point represents the mean (n=4). The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from fresh samples within a 1.2-fold range. [e, and f] Inaccuracy (e) and CV value (f) corresponding to panels a, b, c, and d were calculated as described in Materials & Methods. Each column represents the mean ± SEM (n=308-368 peptides; number commonly detected in FFPE and fresh samples under each experimental condition). **p < 0.001, *p < 0.01, significant difference between two groups (Bonferroni-corrected Student's t-test).

Supplementary figure 8 Effect of PCT treatment and SWATH analysis on comprehensive quantification of BDL-induced change in expression levels of membrane proteins in FFPE sections [a, b, c, and d]
Membrane proteins were selected according to the subcellular location information in the Uniprot mouse proteome database. The BDL-induced changes in expression level (BDL/control ratio) were compared between FFPE and fresh samples of mouse liver at the peptide level. Peptide samples of FFPE and fresh livers were prepared with PCT treatment and measured in the SWATH mode (a, PCT(+)-SWATH), without PCT treatment and measured in the SWATH mode (b, PCT(-)-SWATH), with PCT treatment and measured in the shotgun mode (c, PCT(+)-Shotgun), or without PCT treatment and measured in the shotgun mode (d, PCT(-)-Shotgun). Data analysis was carried out as described in Supplementary figure 3. The data were taken from Supplementary tables 7, 8, 9 and 10. Each point represents the mean (n=4). The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from fresh samples within a 1.2-fold range. [e, and f] Inaccuracy (e) and CV value (f) corresponding to panels a, b, c, and d were calculated as described in Materials & Methods. Each column represents the mean ± SEM (n=119-190 peptides; number commonly detected in FFPE and fresh samples under each experimental condition). *p < 0.001, significant difference between two groups (Bonferroni-corrected Student's t-test).

Supplementary figure 9 Effect of heat-compatible PTS buffer and peptide/data selection criteria on the comprehensive quantification of pathological changes in protein expression level in FFPE tissues [a, and b]
The BDL-induced changes in expression level (BDL/control ratio) were compared between FFPE and fresh samples of mouse liver at the peptide level. Peptide samples of FFPE and fresh livers were prepared with PCT treatment using PTS buffer (95 °C heating for lysis) and measured in the SWATH mode [PCT(+)-SWATH]. Data analysis was conducted with in silico peptide selection criteria and the data selection illustrated in Supplementary figure 3 (a) or without the peptide and data selections (b). Each point represents the mean (n=4). The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from fresh samples within a 1.2-fold range.
[c] This graph was generated using the data from the reported PCT-SWATH study 3 for FFPE and frozen tumorous and benign prostatic tissues. Tumor/benign ratios were compared between FFPE and frozen tissues obtained from the same resected tissues of patient 15, who showed the best agreement (as an inaccuracy value) between FFPE and frozen tumor/benign ratios among the 24 patients. The peptide samples of FFPE and frozen prostatic tissues were prepared with PCT treatment using urea buffer without heating, and then measured in the SWATH mode. The in silico peptide selection criteria and the data selection illustrated in Supplementary figure 3 were not applied. The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from frozen samples within a 1.2-fold range.
[d] Inaccuracy values corresponding to panels a, b and c were calculated as described in Materials & Methods. Each column represents the mean ± SEM (n=1112-6976 peptides; number commonly detected in FFPE and fresh (frozen) samples under each experimental condition). *p < 0.001, significant difference between two groups (Bonferroni-corrected Student's t-test).

Supplementary figure 10 Effect of heat-compatible PTS buffer and peptide/data selection criteria on the comprehensive quantification of pathological changes in expression levels of cytosolic proteins in FFPE tissues [a, and b]
Cytosolic proteins were selected according to the subcellular location information in the Uniprot mouse proteome database. The BDL-induced changes in expression level (BDL/control ratio) were compared between FFPE and fresh samples of mouse liver at the peptide level. Peptide samples of FFPE and fresh livers were prepared with PCT treatment using PTS buffer (95 °C heating for lysis) and measured in the SWATH mode [PCT(+)-SWATH]. Data analysis was conducted with in silico peptide selection criteria and the data selection illustrated in Supplementary figure 3 (a) or without the peptide and data selections (b). Each point represents the mean (n=4). The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from fresh samples within a 1.2-fold range.
[c] This graph was generated using the data from the reported PCT-SWATH study 3 for FFPE and frozen tumorous and benign prostatic tissues. Cytosolic proteins were selected according to the subcellular location information in the Uniprot human proteome database. Tumor/benign ratios were compared between FFPE and frozen tissues obtained from the same resected tissues of patient 15, who showed the best agreement (as an inaccuracy value) between FFPE and frozen tumor/benign ratios among the 24 patients. The peptide samples of FFPE and frozen prostatic tissues were prepared with PCT treatment using urea buffer without heating, and then measured in the SWATH mode. The in silico peptide selection criteria and the data selection illustrated in Supplementary figure 3 were not applied. The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from frozen samples within a 1.2-fold range.
[d] Inaccuracy values corresponding to panels a, b and c were calculated as described in Materials & Methods. Each column represents the mean ± SEM (n=368-2329 peptides; number commonly detected in FFPE and fresh (frozen) samples under each experimental condition). *p < 0.001, significant difference between two groups (Bonferroni-corrected Student's t-test).

Supplementary figure 11 Effect of heat-compatible PTS buffer and peptide/data selection criteria on the comprehensive quantification of pathological changes in expression levels of membrane proteins in FFPE tissues [a, and b]
Membrane proteins were selected according to the subcellular location information in the Uniprot mouse proteome database. The BDL-induced changes in expression level (BDL/control ratio) were compared between FFPE and fresh samples of mouse liver at the peptide level. Peptide samples of FFPE and fresh livers were prepared with PCT treatment using PTS buffer (95 °C heating for lysis) and measured in the SWATH mode [PCT(+)-SWATH]. Data analysis was conducted with in silico peptide selection criteria and the data selection illustrated in Supplementary figure 3 (a) or without the peptide and data selections (b). Each point represents the mean (n=4). The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from fresh samples within a 1.2-fold range.
[c] This graph was generated using the data from the reported PCT-SWATH study 3 for FFPE and frozen tumorous and benign prostatic tissues. Membrane proteins were selected according to the subcellular location information in the Uniprot human proteome database. Tumor/benign ratios were compared between FFPE and frozen tissues obtained from the same resected tissues of patient 15, who showed the best agreement (as an inaccuracy value) between FFPE and frozen tumor/benign ratios among the 24 patients. The peptide samples of FFPE and frozen prostatic tissues were prepared with PCT treatment using urea buffer without heating, and then measured in the SWATH mode. The in silico peptide selection criteria and the data selection illustrated in Supplementary figure 3 were not applied. The broken lines indicate 1.2-fold differences. The % in each scatter plot is the proportion of peptides whose peak areas from FFPE are consistent with those from frozen samples within a 1.2-fold range.
[d] Inaccuracy values corresponding to panels a, b and c were calculated as described in Materials & Methods. Each column represents the mean ± SEM (n=190-1227 peptides; number commonly detected in FFPE and fresh (frozen) samples under each experimental condition). *p < 0.001, significant difference between two groups (Bonferroni-corrected Student's t-test).

Normal and bile duct ligation (BDL) mice
Male C57BL/6 mice (7-8 weeks of age) were purchased from Japan SLC (Hamamatsu, Japan). All experimental protocols were approved by the Institutional Animal Care and Use Committee in Tohoku University, and were performed in accordance with the guidelines of Tohoku University.
The bile duct ligation (BDL) mouse model was prepared as described previously. 1 Briefly, under anesthesia induced with isoflurane, the abdominal cavity was opened and the common bile duct was dissected, and doubly ligated with commercial surgical sutures, leaving the gall bladder intact. The abdominal muscle and wound were sutured with commercial surgical sutures. The liver was isolated at 7 days after BDL. We confirmed that the BDL model mice had been properly prepared by staining the liver with hematoxylin and eosin (H&E); pathological review revealed liver damage, as expected (Supplementary figure 1). In addition, the changes in expression levels of ntcp, bsep (bile acids transporters) and abcb4 (phospholipid transporter), measured in the SWATH mode, were consistent with those reported previously 1 (Supplementary figure 1).

Evaluation of PCT-assisted protein extraction from FFPE sections
The effect of PCT on protein extraction from FFPE sections was evaluated by comparison with the extraction of proteins from frozen sections into PTS buffer. The FFPE tissue suspension prepared as (Pressure BioSciences, South Easton, MA). All samples were incubated at 95°C for 60 min in block incubator (Eppendorf, Hamburg, Germany) with mixing at 1000 rpm. Thereafter, both FFPE and frozen samples were processed in two ways, with or without PCT treatment. For PCT treatment, samples were incubated in a Barocycler (NEP 2320 Enhanced; Pressure BioSciences, South Easton, MA) in two steps as follows: firstly 60 cycles of 95 seconds at 45000 psi and 5 seconds at atmospheric pressure at 95 °C, and secondly 50 cycles of 20 seconds at 45000 psi and 15 seconds at atmospheric pressure at 95 °C. Samples without PCT treatment were simply incubated in the block incubator (Eppendorf, Hamburg, Germany; without mixing) at 95 °C for the same time as the PCT treatment. Finally, all samples were cooled to room temperature, and centrifuged at 15000 rpm for 3 min. The protein concentration in the supernatant was determined by means of BCA assay.

Generation of the spectral library for SWATH analysis
The tryptic digests of FFPE tissue suspension of mouse liver were fractionated by isoelectric focusing and measured in the shotgun mode as described previously. 4 Shotgun data were analyzed using the Paragon algorithm of ProteinPilot Version 4.5 (SCIEX), and the UniProt Mouse proteome database (release2018_03, entries) was searched. The six user-defined options were: (i) cysteine alkylation, iodoacetamide; (ii) digestion, trypsin digestion; (iii) special factors, none; (iv) species, Mus musculus; (v) identification focus, biological modification; and (vi) search effort, thorough identification search. The identification confidence is >99%. The FDR values were all lower than 1%. Finally, excluding overlapping peptides or proteins, 30179 peptides and 3140 proteins were included in the spectral library, which was uploaded in the PeptideAtlas webpage with identifier PASS01457.