Utilization of HPASubC for the Identification of ... - ACS Publications

Mar 22, 2016 - Laboratory of Pathology, National Cancer Institute, Building 10, Room ... Department of Pathology, University of Colorado School of Med...
0 downloads 0 Views 8MB Size
Download PDF
Article pubs.acs.org/jpr

Utilization of HPASubC for the Identification of Sinusoid-Specific Proteins in the Liver Divine-Favour Anene,† Avi Z. Rosenberg,‡ David E. Kleiner,§ Toby C. Cornish,†,⊥ and Marc K. Halushka*,† †

Department of Pathology, Johns Hopkins University School of Medicine, 720 Rutland Avenue, Baltimore, Maryland 21205, United States ‡ Department of Pathology, Children’s National Medical Center, 111 Michigan Avenue Northwest, Washington, D.C. 20010, United States § Laboratory of Pathology, National Cancer Institute, Building 10, Room 2S235, MSC 1500, 10 Center Drive, Bethesda, Maryland 20892, United States ⊥ Department of Pathology, University of Colorado School of Medicine, Academic Office 1, Room L15-2109, 12631 East 17th Avenue, Aurora, Colorado 80045, United States S Supporting Information *

ABSTRACT: Mass spectrometry-based proteomes of human organs and tissues are powerful tools but fail to capture protein localization and expression at the cellular level. For example, the proteome signal in liver represents the combined protein expression across diverse cellular constituents that include hepatocytes, Kupffer cells, endothelial cells, and others. We utilized HPASubC and the Human Protein Atlas (HPA) to identify the sinusoidal component of protein liver expression to further subset and organize this homogeneous signal. We evaluated 51 109 liver images covering 13 197 proteins from the HPA and discovered 1054 proteins that were exclusive to sinusoidal cells. Sinusoidal staining patterns were identified in a Kupffer cell (n = 247), endothelial cell (n = 358), or lymphocyte (n = 86) specific pattern. Two-hundred and thirty-nine of these proteins were not present in the NextProt or Human Proteome Map liver data sets, potentially expanding our knowledge of the liver proteome. We additionally demonstrate unique endothelial cell expression patterns that distinguish between portal vein, hepatic artery, capillary sinusoids, and central vein regions. These findings significantly improve our understanding of the liver proteome with insight into the endothelial complexity across the hepatic vascular network. KEYWORDS: liver, human protein atlas, Kupffer cell, endothelial cell, lymphocyte, HPASubC, proteome, tissue microarray

■

INTRODUCTION Deep proteomic discovery in human tissues has fundamentally changed our understanding of protein expression.1,2 In recent years, mass spectrometry (MS)-based methods have identified thousands of proteins in most human organs including the liver. Although robust at a tissue-level, these data have not yet distinguished expression differences in cells within the same tissue. This can lead to confusion over the origin of a protein. For example, the liver is composed of hepatocytes, Kupffer cells, bile duct epithelial cells, sinusoidal endothelial cells, lymphocytes, arteriole and venule endothelial cells, stellate cells, and others.3−5 Some proteins are ubiquitous and expressed across all of these cells, while others have more limited expression profiles. It is a limitation of whole tissue-level MS data that the cellular origin of each of the proteins cannot be determined. The Human Protein Atlas (HPA) has undertaken a complementary approach to high-throughput proteomics.6,7 HPA has generated hundreds of tissue microarrays (TMAs) © 2016 American Chemical Society

containing representative material from 44 tissues and organs across numerous subjects. These TMAs have been stained by immunohistochemistry for over 16 900 proteins using >24 000 antibodies toward a goal to provide validated staining reagents for every known protein. The advantage of the HPA is the ability to identify the cell type expressing each protein. The disadvantages to this approach are the reliability of IHC staining and the need to create a unique, sensitive, and specific antibody for every protein of study. The HPA reports staining characteristics (including staining intensity, quantity, and localization) for most of the major cell types in each organ; however, no such data are available for some minority/rarer cell types. This is an important hole in the data that must be filled. To identify the expression in either rare cell types or to identify unusual subcellular staining patterns, we developed HPASubC.8 Received: January 28, 2016 Published: March 22, 2016 1623

DOI: 10.1021/acs.jproteome.6b00073 J. Proteome Res. 2016, 15, 1623−1629

Article

Journal of Proteome Research HPASubC is a suite of tools that allows a user to download all images of a particular organ and review them sequentially for staining patterns of interest. HPASubC uses a Playstation-style gamepad controller to rapidly evaluate images at a user dependent rate of up to ∼1 image/second.8 We have previously demonstrated its utility by identifying endothelial cell and smooth muscle cell specific proteins in the heart. We became interested in the expression of nonhepatocyte proteins as a result of perusing proteomic data sets that globally assign proteins to the liver and identify some proteins that were likely not of hepatocyte origin. Our concern was that this data could mislead investigators into exploring the protein in an inappropriate cell type in the liver. Therefore, using HPA and HPASubC, we set out to identify all proteins expressed exclusively in the liver sinusoid component cells such as Kupffer cells (a phagocytic, macrophage-like cell), lymphocytes, stellate cells, and endothelial cells.

■

EXPERIMENTAL PROCEDURES

Utilization of HPASubC

We used HPASubC to download 51 109 liver core images covering 13 197 unique proteins from the Human Protein Atlas Web site. HPASubC was then used to individually review these images and classify the sinusoidal staining pattern present by two investigators (D.-F.A. and M.K.H.). All identifications made by the trainee (D.-F.A.) were reviewed, and assignments were altered as necessary by a board-certified pathologist (M.K.H.). Staining patterns were based upon a prestudy training set utilizing known cell-specific proteins including CD68, MSR, and CD163 for Kupffer cells and ICAM1, CD34, ENG, B2M, and VCAM1 for endothelial cells. An additional prestudy review of ∼1000 images indicated that many proteins had staining patterns consistent with lymphocytes. Other sinusoidal staining did not fit these distinct patterns and were aggregated in a nonspecific category. No specific stellate cell pattern that was clearly distinct from endothelial cells and Kupffer cells was recognized. Specific stellate cell staining was based on a comparison with the mouse data from Azimifar et al.9 The classification categories for the images were arbitrarily assigned the numbers 0−5: 0 indicated no sinusoidal-specific staining; 1 indicated a Kupffer cell pattern; 2 indicated an endothelial cell pattern; 3 indicated lymphocyte staining; 4 indicated stellate cell (only used for post analysis of sinusoidal staining proteins); and 5 indicated sinusoidal staining that was present but could not be further classified. Examples of some staining patterns of the different categories are shown in Figure 1. We excluded all staining that also demonstrated moderate to strong hepatocyte staining but included proteins with strong sinusoidal staining and a hepatocyte “blush” or granular hepatocyte staining. We did not exclude images that stained in the sinusoids but also had bile duct epithelial staining. Cases that were clearly endothelial, as noted by portal triad or central vein staining, but had additional staining, that is, bile duct epithelial staining, were specifically identified as “endothelial cell + other pattern”. Where staining differed between antibodies for the same protein, we favored calling a particular cell type (Kupffer/endothelial) if the second antibody had negative staining. If two different staining patterns emerged between antibodies, or even within a single antibody, we generally placed these in the undistinguished sinusoidal category. We removed proteins in which one antibody strongly

Figure 1. The diversity of staining patterns of the liver sinusoid. (A) Kupffer cell staining pattern of BTK. (B) Endothelial cell staining pattern of BST2. (C) Endothelial cell staining pattern of DYSF. (D) Lymphocyte staining pattern of PSMB9. Images from Human Protein Atlas.

stained hepatocytes. After assigning scores to each sample, we curated the assignments by mining additional resources such as GeneCards and Google Scholar.10 Multiple assignments were changed, usually from category 5 to a specific category, when clear evidence of cell specificity/function could be determined from these resources. HPASubC was used on a Dell Optiplex 9010 PC running Windows 7 SP1 with 16 GB RAM and a 3.4 GHz CPU and on a Lenovo X230 Tablet PC running Windows 7 SP1 with 4 GB RAM and a 2.6 GHz CPU. Gene Ontology (GO) Validation

GO was determined using the webtool at http://geneontology. org/.11 A list of all proteins from each category (Kupffer cell, endothelial cell, lymphocyte, stellate cell, nonspecific sinusoidal) was queried for “biological process”. Comparison to Other Data Sets

RNA sequencing (RNA-seq) data were obtained from HPA for all 32 tissues, and the subset of data for the liver was evaluated using their Fragments Per Kilobase of transcript per Million mapped read (FPKM) values. The entire list of sinusoidal proteins was compared to the liver and heart Human Proteome Map (http://www.humanproteomemap.org/)1 and Nextprot data set (http://www.nextprot.org/).12 The Human Proteome Map contains 9596 and 4889 proteins, respectively, for liver and heart, and Nextprot contains 10 623 and 10 472 proteins, respectively, for liver and heart. Evaluation of Endothelial Staining Patterns

Each of the proteins that stained for an endothelial cell pattern was re-evaluated for the segments of the vasculature they marked. All proteins were assigned into five categories: all vasculature, sinusoids only, nonsinusoidal only, portal triad only, central vein only, or other pattern. 1624

DOI: 10.1021/acs.jproteome.6b00073 J. Proteome Res. 2016, 15, 1623−1629

Article

Journal of Proteome Research

■

GO terms for endothelial cells were “biological adhesion” and “response to wounding” (Table 2). For the lymphocyte group,

RESULTS

Nonhepatocyte Protein Discovery

We identified 1054 proteins (∼8% of all proteins in HPA) that were localized to the sinusoids or small blood vessels and did not stain hepatocytes. Of these, we assigned 247 proteins as having a Kupffer cell pattern, 358 proteins an endothelial cell pattern, 86 proteins a lymphocyte pattern, 5 proteins a stellate cell pattern, and 358 remained undistinguished (shown in Supporting Information Tables S-1−S-5). The stellate cell pattern was generated by mining the stellate cell protein list of Azimifar et al. and cross-referencing it to our sinusoidal proteins and not by pattern recognition.9 Of the 358 endothelial cell proteins, 55 had additional nonhepatocyte staining (i.e., bile duct epithelium staining). The unclassified group was predominately composed of unusual and nondistinct staining patterns in cells located in the sinusoids (Figure 2).

Table 2. Biological Processes Associated with Proposed Endothelial Cell Proteins GO biological process

fold enrichment

biological adhesion response to wounding cell adhesion vesicle-mediated transport blood coagulation

3.46 4.02 3.4 3.07 4.12

P value 1.01 2.34 3.49 1.57 8.02

× × × × ×

10−10 10−10 10−10 10−08 10−07

the top GO terms were “antigen receptor-mediated signaling pathway” and “lymphocyte activation” (Table 3). The top GO Table 3. Biological Processes Associated with Proposed Lymphocyte Proteins GO biological process antigen receptor-mediated signaling pathway lymphocyte activation T-cell aggregation leukocyte cell−cell adhesion

fold enrichment

P value

>5

1.98 × 10−09

>5 >5 >5

7.49 × 10−08 7.66 × 10−08 5.65 × 10−07

terms for the unassigned sinusoidal proteins were “immune system process” and “regulation of cellular component organization” (Table 4). This suggested appropriate but not Table 4. Biological Processes Associated with Unassigned Sinusoidal Proteins GO biological process immune system process regulation of cellular component organization defense response blood coagulation

Figure 2. Nonspecific sinusoidal staining. The unclassified category included numerous proteins with nuclear staining, mixed staining, or unusual distributions of staining. (A) LCLAT1, (B) MEIS2, (C) PARP4, (D) S100A6. Images from Human Protein Atlas.

Because of the variety of staining patterns observed in the sinusoids and our concern about how consistent each pattern was specific to a particular cell type, we performed a GO query on each group to determine if the assigned protein types matched a reasonable function for the presumed cell type. For proteins assigned to Kupffer cells, the top GO terms were “defense response” and “immune response” (Table 1). The top

defense response immune response innate immune response positive regulation of response to stimulus

3.28 3.1 3.62 2.56

× × × ×

2.1 3.03

2.11 × 10−03 6.52 × 10−03

HPA provides RNA expression levels for 19 692 ENSGID genes. Of these, 14 407 have a FPKM ≥ 0.1 in liver. We assessed how well the sinusoidal protein expression data correlated with this gene expression RNA-seq data. Of 1054 sinusoidal proteins, 908 (86%) also had gene expression ≥ 0.1 FPKM with an overall median value of 3.3 FPKM. Exclusive of proteins with no RNA expression, the median FPKM was 4.9. We compared this to the relationship between bile duct epithelium protein expression and RNA-seq data. HPA has curated bile duct staining as high, medium, or low for 5660 proteins. We removed the 4911 proteins that also had concurrent hepatocyte staining. For the remaining 749 “exclusive” bile duct proteins, 688 (92%) had ≥ 0.1 FPKM RNA expression with a median FPKM value of 3.6. Exclusive of genes with no expression values, the median FPKM was 4.1.

P value 5.01 2.01 2.26 4.20

1.68 × 10−05 4.83 × 10−04

Comparison to HPA RNA-seq Data

Table 1. Biological Processes Associated with Proposed Kupffer Cell Proteins fold enrichment

P value

2.05 1.94

perfect enrichment based on perceived staining patterns for Kupffer cells, endothelial cells, and lymphocytes. The stellate cells had an insufficient number of proteins to identify any biological processes.

GO for Each Cell Class

GO biological process

fold enrichment

10−12 10−09 10−09 10−08 1625

DOI: 10.1021/acs.jproteome.6b00073 J. Proteome Res. 2016, 15, 1623−1629

Article

Journal of Proteome Research Comparison to Other Proteomic Liver Data Sets

Table 5. Distribution of HPASubC Discovered Liver Sinusoidal Proteins in Two Large MS Based Proteomic Datasets

We compared our data set of 1054 proteins (searched as one “sinusoidal protein” set) against two established liver MSderived tissue proteomic data sets from the Human Proteome Map and NextProt (Figure 3).1,12 These two large data sets

database Human Proteome Map NextProt a

liver only

heart only

neither organ

both heart and liver

293

35

394

332

a

b

303

566

95

90

Enriched for leukocyte functions. functions.

b

Enriched for endothelial cell

data set, there was enrichment of endothelial cell functions explaining the relatively high match to the heart data sets. Identification of Generic Proteins

Because of the agnostic approach to protein identification in HPASubC, we were able to find a number of poorly characterized proteins that stained sinusoidal cells. We developed a list of 113 proteins that have generic protein names such as chromosome ORFs (i.e., C12orf73), zinc finger proteins (i.e., ZNF583), KIAA genes (i.e., KIAA1239), transmembrane proteins (i.e., TMEM247), and motif-containing proteins (i.e., WDR87, TRIM56, CCDC81) (shown in Supporting Information Table S-7). Twenty had Kupffer cell staining patterns, and 31 had endothelial cell staining patterns. Some of these proteins had interesting staining specificities. JADE2 (Jade family PHD finger 2) gave an endothelial cell pattern in the liver sinusoids. In other organs, there was strong microvilli staining and endothelial cell staining in certain organs (lung, colon, brain). Jade-2 has only been evaluated in the literature in a neurogenesis model.13 C11orf1 has not been published on since being identified in 2000.14 C11orf1 was expressed in Kupffer cells and macrophage cells throughout many organs. It also had a wispy pattern of axonal staining in brain tissues.

Figure 3. Comparisons of sinusoidal HPA protein expression against MS based methods. A Venn diagram indicates the distribution of sinusoidal proteins found by HPASubC against the MS-generated NextProt and the Human Proteome Map liver protein data sets. There were 239 novel liver proteins identified by using this complementary method.

identified a total of 12 884 proteins with roughly half (6047) being identified by both studies. We found 77% (854) of the sinusoidal liver proteins had previously been reported in either of the two larger data sets. These sinusoidal-only proteins represent ∼6% of all proteins identified in the liver. When correlated with HPA RNA-seq liver gene expression data, 95% of proteins observed in both data sets had a FPKM ≥ 0.1 and a median FPKM of 5.65. We also found 239 sinusoidal proteins identified by HPASubC were not reported in either of these two robust MS data sets. These proteins had a similar distribution of cell types as the larger data set (shown in Supporting Information Table S-6). Of this set of proteins, 58% had a FPKM ≥ 0.1, and the median FPKM was 0.1. Exclusive of genes with no expression, the median FPKM was 0.8. We used the Human Proteome Map and NextProt heart proteomic databases as a comparison group. We found 68% (721) of our sinusoidal proteins were also present in either or both of the heart Human Proteome Map and NextProt data sets. This smaller percentage than liver matching suggested a nonrandomness to the HPASubC sinusoidal calls but perhaps a high level of endothelial signal in the sinusoidal data set, which would correspond with the highly vascular heart tissue. We then investigated how the distribution of certain sinusoidal cell types may differ between the liver and heart. The heart natively has an endothelial cell-rich vasculature supply but very few inflammatory cells. There were 332 sinusoidal proteins that appeared in both the liver and heart Human Proteome Map data set (Table 5). For the NextProt data, 566 sinusoidal proteins were also found on both organs’ MS-generated lists. We further investigated HPASubCidentified sinusoidal proteins that exclusively matched either the liver or heart NextProt data. Ninety-five sinusoidal proteins were found exclusively matching the liver NextProt data set, and a GO search indicated these proteins were enriched for leukocyte functions. Conversely, for the 90 sinusoidal proteins found to match the NextProt heart data set but not the liver

Subclassifying Endothelial Cell Staining in the Liver

The sinusoidal endothelium represents a special class of discontinuous endothelial cells.15 As such, it would be expected to have some unique protein expression relative to other arterial or venous endothelium. We therefore investigated what proteins may distinguish the different types of endothelial cells in the liver. From the 358 endothelial cell proteins, we found numerous examples of unique expression patterns in which proteins were found exclusive to the sinusoids or absent from the sinusoids but present in other liver vasculature (Figures 4 and 5). Ras homologue family member B (RHOB) was one protein that was found with exclusive sinusoidal staining. We examined its staining pattern throughout all of the other tissues in HPA. The only other positive endothelial cell staining was in a subset of small vessels within inflammatory cell collections in the appendix and lymph node. Beyond the endothelium, it stained gastric, respiratory, epididymal, and fallopian tube epithelium; placental syncytiotrophoblasts; and brain parenchyma and neurons (Figure 6).

■

DISCUSSION We utilized the power of the HPASubC tool to identify and subset the cells present in the liver sinusoids to further “thinslice” the sources of liver protein expression. We identified 1054 proteins that are present in the liver but were not expressed in hepatocytes. Of these proteins, 239 were not 1626

DOI: 10.1021/acs.jproteome.6b00073 J. Proteome Res. 2016, 15, 1623−1629

Article

Journal of Proteome Research

Figure 6. RHOB expression across multiple organs. RHOB had focal strong staining of endothelium within lymphoid areas in the lymph node and appendix. It also demonstrated epithelial cell staining of the stomach, bronchus, and fallopian tube. There was robust staining in the brain. Images from Human Protein Atlas.

and lymphocyte-assigned proteins and a more blood vesselrelated set of processes for the endothelial cell proteins. The unclassified proteins had a range of biological processes that included immune cell activities. The use of GeneCards and Google Scholar searching of protein names with different cell types improved our assignment of proteins as being Kupffer, lymphocyte, or endothelial cell in origin. Our sinusoidal proteins represent ∼6% of proteins listed in the NextProt liver proteomic database. This should warrant caution to anyone who assumes that all liver proteins are expressed in hepatocytes. Additionally, we did not investigate exclusive bile duct epithelial cell staining in this project, but HPA annotations suggest ∼749 “liver” proteins would be specifically expressed in these cells, increasing the number of nonhepatocyte proteins for which investigators should be aware. There is some overlap of our sinusoidal proteins and bile duct epithelial cells as seen for FLNB in Figure S-1. Interestingly, ∼22% of our proteins were not present in either large MS based proteomic database (NextProt and Human Proteome Map). This can be interpreted in several ways. One is that the sinusoidal cells make up a small percentage of liver area and that modestly expressed proteins in these rarer cells will not be found by MS. Also, there are always gaps in discovery MS methods due to variability in peptide profiles from proteins. Another interpretation would be that immunohistochemistry is fraught with false positive staining and many of our identified proteins are a feature of poor specificity of the staining. Supporting this were the 101 proteins with no gene FPKM levels. It is likely that all of these factors contribute. Our in silico proteomic dissection of the endothelial cell staining patterns of numerous proteins begins to show a true diversity of protein expression by exact location within the liver (Figures 4 and 5, Supporting Information Figure S-1). Differential expression between sinusoidal and nonsinusoidal endothelial cells for a number of proteins was clear. These protein differences are more extensive than the miRNA homogeneity we described among endothelial cells taken from different locations.16 It is likely these variable expression patterns can inform on pathways regulating different physiologic/functional requirements of endothelium in different suborgan locations.15,17 This sinusoidal expression information has a variety of uses. Investigators who identify a specific gene or protein signal of interest from a whole liver preparation can use our HPASubCderived sinusoidal expression data to determine if the protein

Figure 4. Characterization of variable endothelial cell expression. This schematic identifies unique staining patterns of endothelium by location (red for hepatic artery and purple for portal vein) and intensity (line thickness). Representative figures demonstrating these unique patterns are shown in Supplemental Figure 1.

Figure 5. Unique endothelial cell staining patterns. POU2AF1 has only hepatic artery staining. CRIP2 has strong hepatic artery, peribiliary plexus, and portal vein staining. It also has weak central vein staining (not seen). RHOB and CLEC4M have strong sinusoidal staining but absent portal triad staining. CD36 has strong endothelial staining except in the hepatic artery. ZFYVE28 has robust endothelial cell staining in all areas. Images from Human Protein Atlas.

previously described in either of the two large human liver proteome data sets.1,12 Distinguishing between the cells of the sinusoid (Kupffer cell, endothelial cell, lymphocyte, stellate cell) by staining alone was a surprisingly daunting task as many staining patterns were equivocal. These challenging patterns included nuclear staining, variable staining across different tissue cores, and atypical patterns of positive cell localization. Therefore, ∼34% of all identified proteins remained classified as distinctly sinusoidal. Of the remaining 696 proteins, 51% were classified as endothelial cells. The presence of portal triad and central vein endothelial staining, when present, was useful in aiding protein assignment toward endothelial cells. We were able to leverage a number of online genomic tools to help classify and verify our protein assignments. Our GO analysis clearly showed immune patterns for our Kupffer cell 1627

DOI: 10.1021/acs.jproteome.6b00073 J. Proteome Res. 2016, 15, 1623−1629

Article

Journal of Proteome Research

nonhepatocyte, liver expressed proteins toward a goal of building extensive protein expression maps that distinguish protein expression at the cellular level.

has hepatocyte or nonhepatocyte localization. This could avoid improperly ascertaining the gene/protein function in a hepatocyte cell line, such as HepG2, when the gene/protein may not be expressed by that cell type. The data we uncovered on differential endothelial cell expression can be used by researchers seeking to better understand the biology of sinusoidal endothelial cells and how their protein expression differences relate to their unique discontinuous membrane. We thought it was interesting that RHOB was found in sinusoids and in occasional small blood vessels in lymphoid organs. Interestingly, RhoB, described above as a sinusoidal-only expressed liver protein, was shown to have relatively high levels of gene expression in HepG2 cells.18 This study has several limitations, particularly in the classification of proteins. There are false positive and false negative staining patterns in HPA. Even when there is appropriate staining, there is a fair amount of inconsistency. The separation between nonspecific “blush” staining of hepatocytes and real low-level hepatocyte staining was based on a qualitative assessment by the reviewers and is therefore associated with a degree of uncertainty. Also, pattern differences are frequent and most pronounced when multiple antibodies are used to assay the same protein. The HPA is continuing to evolve, and better antibodies are being added such that some staining patterns will change over time as new images replace inferior studies. In particular, they are developing an “annotated protein expression (APE) score”, which takes into account staining data from at least two antibodies, literature, and additional experimental data.7 At this time, only 31% of their proteins have “supportive” values suggesting the antibodies used will continue to increase and potentially change. As their data set becomes more robust, it is certain our protein list will require revision. Additionally, not all liver core images had an evaluable portal triad to assess endothelial cell staining, so this class may be underreported. We also identified 146 proteins without any reported liver gene expression, 69% of which were also not in either large MSbased proteome. These would have to be considered more suspect or lowly expressed protein candidates. The HPASubC method also has limitations. As described before, there is a 0.65% basal human error rate in classifying tissue type by HPASubC.8 Because of the difficulties described above in assigning specific staining patterns, we most certainly misclassified proteins. We were unable to resolve a distinct stellate cell staining pattern and relied on the data in Azimifar et al. post hoc, which characterized stellate cell proteins in a mouse model.9 Additionally, of their 29 stellate cell specific proteins that they claimed to be identifiable in HPA, we believe that only five showed a reasonable sinusoidal staining pattern. We are certain this class was significantly underrepresented in our findings. Additionally, it is certain that some serum proteins, extracellular matrix/basement membrane proteins, and transiting cells such as neutrophils were also included in our protein list. We purged those that, by name or known function, did not belong in our lists (e.g., collagens, neutrophil-named proteins), but some certainly remain represented in our data set. Nonetheless, this imperfect data set represents the first attempt to leverage the enormous HPA resource to tease out nonhepatocyte proteins that would otherwise be confused with hepatocyte signals, and it generally improves our knowledge of liver protein expression. In conclusion, we have utilized the HPASubC suite of tools on the HPA resource to identify over 1000 proteins that are

■

ASSOCIATED CONTENT

* Supporting Information S

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jproteome.6b00073. Additional patterns of endothelial cell staining in the liver (these proteins were reported in Figure 4; images from Human Protein Atlas) (PDF) Proteins with a Kupffer cell staining pattern; proteins with an endothelial cell staining pattern; proteins with a lymphocyte staining pattern; proteins with a stellate cell staining pattern; proteins with a sinusoidal staining pattern that could not be further classified (XLSX) List of putative proteins identified by HPASubC but not present in two MS based liver proteomic data sets (XLSX) List of generic proteins found in the sinusoids (XLSX)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 410-614-8138. Fax: 410502-5862. Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Kim, M. S.; Pinto, S. M.; Getnet, D.; Nirujogi, R. S.; Manda, S. S.; Chaerkady, R.; Madugundu, A. K.; Kelkar, D. S.; Isserlin, R.; Jain, S.; Thomas, J. K.; Muthusamy, B.; Leal-Rojas, P.; Kumar, P.; Sahasrabuddhe, N. A.; Balakrishnan, L.; Advani, J.; George, B.; Renuse, S.; Selvan, L. D.; Patil, A. H.; Nanjappa, V.; Radhakrishnan, A.; Prasad, S.; Subbannayya, T.; Raju, R.; Kumar, M.; Sreenivasamurthy, S. K.; Marimuthu, A.; Sathe, G. J.; Chavan, S.; Datta, K. K.; Subbannayya, Y.; Sahu, A.; Yelamanchi, S. D.; Jayaram, S.; Rajagopalan, P.; Sharma, J.; Murthy, K. R.; Syed, N.; Goel, R.; Khan, A. A.; Ahmad, S.; Dey, G.; Mudgal, K.; Chatterjee, A.; Huang, T. C.; Zhong, J.; Wu, X.; Shaw, P. G.; Freed, D.; Zahari, M. S.; Mukherjee, K. K.; Shankar, S.; Mahadevan, A.; Lam, H.; Mitchell, C. J.; Shankar, S. K.; Satishchandra, P.; Schroeder, J. T.; Sirdeshmukh, R.; Maitra, A.; Leach, S. D.; Drake, C. G.; Halushka, M. K.; Prasad, T. S.; Hruban, R. H.; Kerr, C. L.; Bader, G. D.; Iacobuzio-Donahue, C. A.; Gowda, H.; Pandey, A. A draft map of the human proteome. Nature 2014, 509 (7502), 575−81. (2) Wilhelm, M.; Schlegl, J.; Hahne, H.; Gholami, A. M.; Lieberenz, M.; Savitski, M. M.; Ziegler, E.; Butzmann, L.; Gessulat, S.; Marx, H.; Mathieson, T.; Lemeer, S.; Schnatbaum, K.; Reimer, U.; Wenschuh, H.; Mollenhauer, M.; Slotta-Huspenina, J.; Boese, J. H.; Bantscheff, M.; Gerstmair, A.; Faerber, F.; Kuster, B. Mass-spectrometry-based draft of the human proteome. Nature 2014, 509 (7502), 582−7. (3) Norris, S.; Collins, C.; Doherty, D. G.; Smith, F.; McEntee, G.; Traynor, O.; Nolan, N.; Hegarty, J.; O’Farrelly, C. Resident human hepatic lymphocytes are phenotypically different from circulating lymphocytes. J. Hepatol. 1998, 28 (1), 84−90. (4) Li, Z.; Diehl, A. M. Innate immunity in the liver. Curr. Opin. Gastroenterol. 2003, 19 (6), 565−71. (5) Young, B.; Lowe, J. S.; Stevens, A.; Heath, J. W. Wheater’s Functional Histology, 5th ed.; Elsevier: China, 2006. (6) Uhlen, M.; Fagerberg, L.; Hallstrom, B. M.; Lindskog, C.; Oksvold, P.; Mardinoglu, A.; Sivertsson, A.; Kampf, C.; Sjostedt, E.; Asplund, A.; Olsson, I.; Edlund, K.; Lundberg, E.; Navani, S.;

1628

DOI: 10.1021/acs.jproteome.6b00073 J. Proteome Res. 2016, 15, 1623−1629

Article

Journal of Proteome Research Szigyarto, C. A.; Odeberg, J.; Djureinovic, D.; Takanen, J. O.; Hober, S.; Alm, T.; Edqvist, P. H.; Berling, H.; Tegel, H.; Mulder, J.; Rockberg, J.; Nilsson, P.; Schwenk, J. M.; Hamsten, M.; von Feilitzen, K.; Forsberg, M.; Persson, L.; Johansson, F.; Zwahlen, M.; von Heijne, G.; Nielsen, J.; Ponten, F. Proteomics. Tissue-based map of the human proteome. Science 2015, 347 (6220), 1260419. (7) Uhlen, M.; Oksvold, P.; Fagerberg, L.; Lundberg, E.; Jonasson, K.; Forsberg, M.; Zwahlen, M.; Kampf, C.; Wester, K.; Hober, S.; Wernerus, H.; Bjorling, L.; Ponten, F. Towards a knowledge-based Human Protein Atlas. Nat. Biotechnol. 2010, 28 (12), 1248−50. (8) Cornish, T. C.; Chakravarti, A.; Kapoor, A.; Halushka, M. K. HPASubC: A suite of tools for user subclassification of human protein atlas tissue images. J. Pathol Inform 2015, 6, 36. (9) Azimifar, S. B.; Nagaraj, N.; Cox, J.; Mann, M. Cell-type-resolved quantitative proteomics of murine liver. Cell Metab. 2014, 20 (6), 1076−87. (10) Safran, M.; Dalah, I.; Alexander, J.; Rosen, N.; Iny Stein, T.; Shmoish, M.; Nativ, N.; Bahir, I.; Doniger, T.; Krug, H.; Sirota-Madi, A.; Olender, T.; Golan, Y.; Stelzer, G.; Harel, A.; Lancet, D. GeneCards Version 3: the human gene integrator. Database 2010, 2010, baq020. (11) The Gene Ontology Consortium. Gene Ontology Consortium: going forward. Nucleic Acids Res. 2015, 43 (Databaseissue), D1049− 56. (12) Gaudet, P.; Michel, P. A.; Zahn-Zabal, M.; Cusin, I.; Duek, P. D.; Evalet, O.; Gateau, A.; Gleizes, A.; Pereira, M.; Teixeira, D.; Zhang, Y.; Lane, L.; Bairoch, A. The neXtProt knowledgebase on human proteins: current status. Nucleic Acids Res. 2015, 43 (Database issue), D764−70. (13) Han, X.; Gui, B.; Xiong, C.; Zhao, L.; Liang, J.; Sun, L.; Yang, X.; Yu, W.; Si, W.; Yan, R.; Yi, X.; Zhang, D.; Li, W.; Li, L.; Yang, J.; Wang, Y.; Sun, Y. E.; Zhang, D.; Meng, A.; Shang, Y. Destabilizing LSD1 by Jade-2 promotes neurogenesis: an antibraking system in neural development. Mol. Cell 2014, 55 (3), 482−94. (14) O’Brien, K. P.; Tapia-Paez, I.; Stahle-Backdahl, M.; Kedra, D.; Dumanski, J. P. Characterization of five novel human genes in the 11q13-q22 region. Biochem. Biophys. Res. Commun. 2000, 273 (1), 90− 4. (15) Aird, W. C. Phenotypic heterogeneity of the endothelium: II. Representative vascular beds. Circ. Res. 2007, 100 (2), 174−90. (16) McCall, M. N.; Kent, O. A.; Yu, J.; Fox-Talbot, K.; Zaiman, A. L.; Halushka, M. K. MicroRNA profiling of diverse endothelial cell types. BMC Med. Genomics 2011, 4, 78. (17) Aird, W. C. Phenotypic heterogeneity of the endothelium: I. Structure, function, and mechanisms. Circ. Res. 2007, 100 (2), 158−73. (18) Fritz, G.; Gnad, R.; Kaina, B. Cell and tissue-type specific expression of Ras-related GTPase RhoB. Anticancer Res. 1999, 19 (3A), 1681−8.

1629

DOI: 10.1021/acs.jproteome.6b00073 J. Proteome Res. 2016, 15, 1623−1629