Protein estimation in perfusate samples

Protein estimation was initially carried out on perfusate samples using a copper-based assay. The protein content of perfusate measured this way was notably high, despite the relatively dilute nature of perfusate. A compatibility issue between perfusate and the copper-based assay was suspected, and a Coomassie-based protein estimation assay was tested as an alternative.

An experiment was conducted using solutions of water and UW-MPS spiked with known quantities of BSA protein, in order to compare expected and measured protein recovery by each method and identify the most accurate protein estimation method for perfusate. Predicted and measured protein concentrations were normalised and the standard scores used to calculate percentage error, indicating the accuracy of each measurement method (Fig. 1B).

While both assays measured the concentration of BSA diluted in water reasonably accurately, with low and similar degrees of error (copper: 17%; Coomassie: 12%), the copper-based assay consistently overestimated the protein concentration of solutions containing UW-MPS. The percentage error between predicted and expected protein concentration was 121.8% and 129% for BSA diluted in perfusate and pure perfusate (with no protein spike) respectively. In comparison, the Coomassie-based assay continued to measure protein concentration accurately, with 6% and 4% error recorded for BSA in perfusate and pure perfusate respectively. From these findings it was established that copper-based assays were inappropriate for measuring perfusate concentration, and the Coomassie-based assay was therefore used for subsequent experiments.

Identification of proteins in kidney perfusate

A total of 24 human kidney perfusate samples were selected from the COPE COMPARE kidney study cohort and underwent protein estimation by Coomassie-based assay. Concentrations of samples ranged from 0.19 to 1.85 µg/µl (Fig. 1C). From these, 6 samples with enough material were selected and prepared for proteome profiling using both in-gel and in-solution digestion.

MS/MS analysis following gel digestion identified a total of 19,022 peptide-spectrum matches (PSMs), of which 1,369 represented unique peptides. This resulted in the identification of 478 proteins across all 6 samples with > 95% confidence. An average of 332 proteins were identified per perfusate sample.

The same 6 kidney perfusate samples underwent in-solution digestion, which identified a total of 468 proteins, 37,626 PSMs and 2084 unique peptides. An average of 346.2 proteins were identified per sample, ranging from 166 to 397 proteins; 138 (29.4%) were detected in all six samples. A complete overview of all identified proteins is provided by Supplementary Files 1 A and 1B.

Comparison of in-gel and in-solution digestion of kidney perfusate

To assess the efficiency of the protein digestion in in-solution against the standard gel digestion method, six kidney perfusate samples were selected and 2 aliquots from the same sample underwent both in-gel and in-solution digestion for direct comparison. We found that proteins identified following the latter digestion were associated with a greater number of unique peptides and peptide-spectrum matches (PSMs) per protein on average, compared to those identified following in-gel digestion, and a higher average sequence coverage of identified proteins (Table 1B).

Similar numbers of proteins were identified by both digestion methods but the number of peptides identified by in-solution digestion were almost twice that identified by the gel method (3,109 and 1,721 respectively). The percentage sequence coverage, indicating the proportion of each protein matched to identified peptides sequence by LC-MS/MS, ranged from 1 to 89% for proteins identified following in-gel digestion (average coverage 19.1% per protein), and from 1 to 100% for in-solution digestion (average 27.2% per protein). Of the in-gel-digested proteins, 94% had sequence coverage below 40% and 33% below 10%, whereas for in-solution digested proteins, 78% had < 40% and 22% had < 10% sequence coverage. Only 29 proteins were identified with more than 40% coverage following in-gel digestion, compared to 103 proteins following in-solution digestion. When we looked at proteins identified by one method and not the other, we found no notable difference: 257 proteins were identified by gel digestion and not by in-solution; 247 were identified only by in-solution digestion and not by gel. To check the efficiency of trypsin digestion by both methods, we compared the numbers of proteins that fell within various molecular weight ranges (10-20 kDa, 21-30 kDa, 31-40 kDa etc. up to > 100 kDa). Similar numbers of proteins were found within each mass range for both digestion groups, indicating equally successful digestion by both methods. Average protein and peptide characteristics per sample are presented in Table 1B.

We used Mascot Server (version 2.5.0; Matrix Science Ltd., London, UK) for identification, characterisation and quantitation of proteins in perfusate samples. Mascot scores, indicating the statistical probability of an accurate protein identification from the sequence database, were consistently higher on average for proteins identified following in-solution digestion compared to those identified following in-gel digestion, with the average Mascot score per protein 656.4 for the latter and 1688 for the former (Table 1B). Thus, the results suggest that digestion using in-solution is more reliable compared to the in-gel method.

Application of the method to other perfusate types

Having demonstrated the efficiency of our in-solution digestion method with kidney perfusion solution, we wanted to show its applicability to other perfusion solutions and perfusion types.

We selected a perfusion solution from the Liver Defatting Study, which represents a significantly different type of perfusate material: this solution was supplemented packed red blood cells and numerous additional pharmacological agents (see methods), and the perfusion process itself took place at normothermic temperature (as opposed to hypothermic in the case of the kidney). The liver itself is also a substantially more metabolically active organ than the kidney, and it was hoped this would be apparent in the resulting protein identifications.

Patient demographics for the two livers included in this study are shown in Table 1 A. We analysed six liver perfusate samples by LC-MS/MS following in-solution digestion and a total of 244 proteins were identified with > 95% confidence from 1,067 distinct peptides derived from 7,208 PSMs (Supplementary Table 1 C). In-gel digestion was then performed on two of the six liver perfusate samples in order to compare the efficacy of the methods against each other in the context of liver perfusion solution. From these, 87 proteins/198 peptides were identified by in-gel digestion, versus 89 proteins/344 peptides identified in the same two samples by in-solution digestion (Supplementary Table 1D). As with the kidney, protein identifications following in-solution digestion were associated with a greater number of peptides, unique peptides and peptide-spectrum matches, suggesting more reliable protein identifications. Sequence coverage of identified proteins ranged from 1 to 55% for in-gel digestion and 2–80% for in-solution. Average sequence coverage was higher for in-solution digestion, at 19.7% per protein, compared to 16% for in-gel digestion (Table 1B). Interestingly, we observed that only 68 proteins were identified by gel digestion and not by in-solution digestion, whereas a much greater 225 were identified following in-solution digestion and not by gel digestion.

Protein characteristics

The distribution of proteins based on their molecular weight (MW) and isoelectric point (pI) was evaluated for both sample preparation methods. We observed that the in-solution method allowed us to recover a higher number of larger proteins. The molecular weights of gel-digested proteins ranged from 11.1 kDa to 192.8 kDa, while in-solution digestion identified a broader range of protein weights, from 4.5 kDa to 515.2 kDa, indicating a greater diversity of recovered proteins. Among the proteins identified through gel digestion, 70% fell within the 10–30 kDa range, indicating an overrepresentation of lower molecular weight proteins. In contrast, proteins identified following in-solution digestion were more evenly distributed across the entire weight range (see Fig. 2).

Fig. 2

Ridgeline plots showing the distributions of the Molecular weight (kDa) and isoelectric point (pI) of proteins identified in Kidney (A and C) and liver (B and D) perfusate samples by in-gel and in-solution digestion methods. Each raincloud dot is the individual protein identified in perfusate samples and proteins present in the human reference proteome

In comparison with the expected distributions based on all proteins present in the human reference proteome (UniProtKB Homo sapiens UP000005640, canonical with 92,158 entries), relatively fewer small and basic proteins were detected by the different methods (see Fig. 2). Mascot scores were once again higher overall in the in-solution digested group compared to gel (see Table 1B).

Interpretation of proteome changes in kidney and liver perfusate

Having demonstrated the efficiency of the in-solution digestion method in both types of perfusate, we used the data to assess the proteomes of the kidney and liver perfusate in more detail.

Data from 8 kidney perfusate samples that had undergone hypothermic machine perfusion in the presence and absence of oxygen (HMPO2 and HMP), for three timepoints: P1 (15 min after the start of perfusion, n = 2), P2 (during perfusion, before leaving the donor centre, n = 2) and P3 (end of perfusion, n = 4) was used. There was an overall increase in the number of proteins identified at timepoints P2 and P3 compared to P1 (Fig. 1D).

Proteomic profiles were compared between perfusion timepoints. The majority of proteins were identified at all timepoints, with a high degree of similarity also seen between P2 and P3 timepoints. P3 samples had the greatest number of unique proteins not found at any other timepoints (56) (Fig. 3).

In-solution digestion was carried out on six liver perfusate samples in total, taken from two livers (L1 and L2) at 3 timepoints each, from 0 to 30 h perfusion duration (T0-30). It was observed that the average number of proteins identified at each timepoint increased with perfusion duration (Fig. 1E). A large proportion of proteins (78) identified in liver perfuste samples were identified at all timepoints, with an even larger number (92) identified only at timepoints T0.5, T9 and T30, possibly suggesting an uptick in protein secretion triggered by the start of perfusion (Fig. 3).

Fig. 3

Similarities in proteomic profiles of kidney and liver perfusate samples. Numbers indicate the number of proteins in common between each combination of timepoints

Interpretation of biological changes in kidney and liver perfusate

Taking into account the numbers of perfusate proteins identified per sample and the confidence of identification data, we established that the in-solution digestion method shows significant advantages over in-gel digestion. Therefore, only proteins identified by this method were used for Gene Ontology (GO) enrichment analysis. A total of 261 genes from kidney perfusate and 138 genes from liver perfusate were analysed using FunRich analysis software (www.funrich.org).

Cellular component analysis of kidney and liver perfusate showed similar proportions of cytoplasm, extracellular and lysosome, membrane and Golgi apparatus related genes, (Fig. 4A (Supplementary Table 2 A)). Biological processes enriched within liver perfusate included energy pathways, metabolism, anti-apoptosis and aldehyde metabolism pathways, while kidney perfusate samples were enriched in cell growth and/or maintenance, signal transduction, cell communication, cell growth and regulation of cell cycle (Fig. 4C (Supplementary Table 2B)). The liver perfusate proteome had higher levels of proteins involved in catalytic, oxidoreductase and hydrolase activities, whereas kidney perfusate samples were enriched in proteins involved in cytoskeletal binding, molecular structural activity and transcriptional regulator activity (Fig. 4B (Supplementary Table 2 C)).

Fig. 4

(A-C) Gene Ontology analysis of perfusate proteins. Number of proteins identified within kidney and liver perfusate datasets expressed as a percentage of total number of available genes in background dataset/database. (D) STRING analysis showing interaction network of complement and coagulation cascade proteins in kidney and liver perfusate

A substantial number of the proteins identified in kidney and liver perfusate samples, almost 44% and 51% of their proteomes respectively, have been reported in human plasma (http://www.plasmaproteomedatabase.org/). This included members of the complement and coagulation cascade, of which 14 (A2M, C1R, C3, C4B, C4BPA, C5, C6, F2, FGB, FGG, PLG, SERPIND1, SERPINF2, VTN) were identified in perfusate collected from both kidney and liver perfusions (Fig. 4D), 6 proteins (C4A, C7, C8A, C8B, C8G, CPB2) were unique to kidney perfusate and 2 (CFH and CFHR1) were unique to liver perfusate. Also identified were members of the peroxiredoxin family of antioxidant enzymes (Prx-1, 2, 3, 5 and 6 identified in kidney and liver perfusate) and the high density lipoprotein family (APOA1, APOA2, APOB, APOC3, APOE present in perfusate collected from both organs and APOA1BP, APOA4, APOC4, APOC2 present only in kidney perfusate samples). Two haemoglobin isoforms: HBB and HBD were identified in both perfusate types, while HBA1 was detected in liver and HBA2 in kidney perfusate samples.

A total of 8 genes known to be specifically enriched in the kidney were identified in kidney perfusate samples (https://www.proteinatlas.org/) ATPase H + transporting V1 subunit B1 (ATP6V1B1), crystallin lambda 1 (CRYL1), dimethylarginine dimethylaminohydrolase 1 (DDAH1), fructose-bisphosphatase 1 (FBP1), glutathione peroxidase 3 (GPX3), lactate dehydrogenase B (LDHB), phosphotriesterase-related (PTER), and uromodulin (UMOD) while a total of 93 gene products identified in liver perfusate samples have high expression in liver compared to other organs. Among those, 19 genes are already FDA approved drug targets, such as coagulation factor II, thrombin (F2), acetyl-CoA acyltransferase 1 (ACAA1), alcohol dehydrogenase 1 A (ADH1A), aldehyde dehydrogenase 2 (ALDH2), aminolevulinate dehydratase (ALAD), catalase (CAT), fibrinogen beta chain (FGB), fibrinogen gamma chain (FGG), plasminogen (PLG), and guanidinoacetate N-methyltransferase (GAMT).