SHOTGUN PROTEOMICS: IDENTIFICATION OF UNIQUE PROTEIN PROFILES OF APOPTOTIC BODIES FROM BILIARY EPITHELIAL CELLS

Abstract

Shotgun proteomics is a powerful analytic method to characterize complex protein mixtures in combination with multi-dimensional liquid chromatography-tandem mass spectrometry (LC-MS/MS). We have used this platform for proteomic characterization of apoptotic bodies in efforts to define the complex protein mixtures found in primary cultures of human intrahepatic biliary epithelial cells (HiBEC), human renal proximal tubular epithelial cells, human bronchial epithelial cells, isolated intrahepatic biliary epithelial cells from explanted primary biliary cirrhosis (PBC) and control liver, using a total of 24 individual samples. Further, as additional controls and for purposes of comparison, proteomic signatures were also obtained from intact cells and apoptotic bodies. The data obtained from LC-MS/MS, combined with database searches and protein assembly algorithms, allowed us to address significant differences in protein spectral counts and identify unique pathways that may be a component to the induction of the signature inflammatory cytokine response against BECs, including the Notch signaling pathway, IL8, IL6, CXCR2 and integrin signaling. Indeed there are 11 proteins that localize specifically to apoptotic bodies of HiBEC and 8 proteins that were specifically absent in HiBEC apoptotic bodies. In conclusion, proteomic analysis of BECs from PBC liver compared to normal liver are significantly different, suggesting that an immunological attack affects the repertoire of proteins expressed and that such cells should be thought of as living in an environment undergoing continuous selection secondary to an innate and adaptive immune response, reflecting an almost “Darwinian” bias.

We have previously postulated that the selective destruction of biliary epithelial cells (BECs) in PBC is related to the relationships between the apoptotic bodies of BEC, the innate immune system and mitochondrial autoantibodies (1–13). This model includes an MHC-independent mechanism as PBC can recur following liver transplantation, i.e. irrespective of HLA mismatch (14, 15). The bleb contents of biliary cells may contain components that stimulate pro-inflammatory responses or alternatively lack components that normally suppress such responses. Over the past decade, shotgun proteomics has emerged as a powerful analytical method to characterize complex protein mixtures (16–18). By combining multidimensional liquid chromatography-tandem mass spectrometry (LC-MS/MS) with database searches and protein assembly algorithms, shotgun proteomics platforms surpass other MS-based proteomics systems in the number and the diversity of proteins identified, and in the dynamic range of detection. Moreover, shotgun proteomic data sets from different tissue phenotypes are sufficiently rich in quantitative information that statistically significant differences in proteins spectral counts can identify differences in abundance of proteins in the samples (19, 20).

We report herein the proteomic characterization of the content of apoptotic bodies from human BECs. We identify 11 proteins localized specifically to apoptotic bodies of BECs but not similar apoptotic bodies from other epithelial tissues, and 8 proteins that were specifically absent in HiBEC apoptotic bodies. We also identify 6 proteins present in apoptotic blebs from BECs isolated from PBC patients when compared to those from BECs from healthy donors. These signature proteins and pathways may shed light on potential molecules for therapeutic intervention.

METHODS

Cell types and isolation of biliary epithelial cells

Primary cultures of normal HiBEC, Human Renal Proximal Tubular Epithelial cells (HRPTEpiC) and human bronchial epithelial cells (BrEPC) cells (ScienCell, San Diego, California) were primary cultures isolated from normal human tissue and cryopreserved immediately after purification. HiBEC were cultured in epithelial cell medium (ScienCell) supplemented with 2% fetal bovine serum (FBS), epithelial cell growth supplement (ScienCell) and 1% penicillin in flasks coated with poly-L-lysine (Sigma-Aldrich, St. Louis, MO). HiBEC were characterized using a previously described immunofluorescence microscopic method with antibodies to cytokeratin 7, cytokeratin 19 and vimentin, which labeled >90% of the cells (21, 22). The other epithelial cells were cultured under the same conditions without FBS. All experiments were performed using cells between passage 2 and 5 (8, 9).

Human BECs were isolated from PBC end stage liver tissue removed at transplantation and from non-diseased liver tissue that were surplus to transplant requirements (23). All tissue was obtained from fully informed patients with written consent. Ethical approval was granted by the South Birmingham Research Ethics Committee. Liver tissue underwent mechanical and enzymatic digestion with Collagenase type 1A (Sigma Aldrich, Dorset, UK) and density gradient centrifugation on 33%/77% Percoll. BECs were further purified by immunomagnetic isolation using antibodies against the cholangiocyte-specific receptor HEA 125 (Progen, Heidelberg, Germany). The cells were cultured in Dulbecco’s Modified Eagle medium, Hams F12 plus 10% heat-inactivated human serum, glutamine (2 mM), hepatocyte growth factor (10 ng/ml, Peprotech, London, UK), epidermal growth factor (10 ng/ml, Peprotech), hydrocortisone (2 µg/ml), cholera toxin (10 ng/ml, Sigma Aldrich), tri-iodo-thyronine (2 nM, Sigma Aldrich) and insulin (0.124 U/ml) with penicillin (100 IU/ml) and streptomycin (100 µg/ml). Cells were cultured to confluence in tissue culture flasks coated with rat tail collagen.

Apoptosis induction and isolation of apoptotic bodies

Apoptosis was induced as described (9); briefly HIBEC, HRPTEpiC and BrEPC were incubated in medium containing 1mM sodium glycochenodeoxycholate at 37°C for 4h. To establish optimal conditions for the induction of apoptosis, we previously incubated all cells types at 37°C for 1, 2, 3, and 4 hours using serial concentrations (100 µM, 500 µM, 1 mM, and 2mM) of GCDC added to normal culture medium, and in absence of serum and growth factors (24). The degree of apoptosis was confirmed for each experiment.

Apoptotic blebs were isolated as previously described (9) from normal primary cultures of HIBEC, HRPTEpiC and BrEPC, as well as from BECs isolated from PBC patients and healthy controls. Briefly, the cell culture supernatant fluid was collected after apoptosis induction. Two additional centrifugation steps (500g for 5 minutes) were performed to remove remaining cells. The supernatant fluid was thence passed through a 1.2 µm non-pyrogenic, hydrophilic syringe filter. After centrifugation at 100,000g for 30 minutes, the pellet containing apoptotic bodies was harvested and lysed in RIPA buffer (Thermo Scientific Pierce, Rockford, IL) supplemented with Complete Protease Inhibitor Cocktail (Roche, Indianapolis, IN). Protein concentration was determined by the BCA assay (Thermo Scientific Pierce). A total of 24 samples of apoptotic bodies were studied, including 4 independent preparations of HiBEC cells, 6 independent preparations of HRPTEpiC, 6 human BrERC, 3 independent preparations of BECs from PBC patients, and 5 from healthy controls. Six different samples of apoptotic blebs of each cell type were isolated. Finally to compare the expression level of proteins found in apoptotic bodies with those in intact cells, we ran an extensive proteome characterization of HiBEC, HRPTEpiC, and BrEPC.

Comparative Shotgun Proteomics

Protein content of the samples was determined by the BCA (bicinconic acid) assay and using a Nanodrop ND-1000 ultraviolet-visible Spectrophotometer (NanoDrop Technologies, Wilmington, DE). Purified apoptotic bodies containing normalized levels of protein were lysed in urea buffer (7M urea, 1M thiourea, 4% CHAPS, 4 mg/mL DTT, pH 8.0). Protein was precipitated using 25% TCA (w/v), washed with ice cold acetone, and re-suspended in 8M urea 100 mM Tris, pH 8.0. Proteins were reduced with TCEP, alkylated with iodoacetemide, diluted to 2M urea, and then digested with trypsin. The resulting peptide mixtures were split into duplicates and analyzed via MudPIT (25). Briefly, peptides were loaded onto a biphasic pre-column fitted using an Upchurch M-520 filter union. This 150 um fused silica microcapillary column was packed with 4 cm of 5-µm strong cation-exchange resin followed by 4 cm of C18 reverse-phase resin. Once loaded, it was then placed in-line with a 100 µm X 20 cm, C18 packed emitter tip column coupled to an LTQ ion trap mass spectrometer nanospray source. Separations were accomplished using 25–1000 mM pulses of ammonium acetate followed by a 115 min reversed phase gradient of 0–40% acetonitrile 0.1% formic acid. Tandem mass spectra were collected in a data dependent manner using dynamic exclusion. Spectra were extracted using ScanSifter and searched against the mouse Uniprot database that also contained reversed versions of the proteins using Sequest. Peptide to protein matches and spectral counts were assembled using IDPicker (9, 26) using a false discovery rate target of 5%. All shotgun proteomics experiments were performed at the Proteomics Laboratory at Vanderbilt University.

Statistical Analysis and Pathway definition

The data comparing the hepatic blebs to each of the control groups were individually plugged into Quasitel, a graphical interface software package that reads standard output from protein assemblies created by IDPicker, and analyzed with a quasi-likelihood Generalized Linear Modeling (GLM) system (19). All reported proteins were identified with a minimum of 2 distinct peptides. The fold change of each protein was calculated by comparing the log2 of total spectra per protein in blebs from HiBEC compared to the log2 ratio of spectra in HRPTEpiC or BrEPC. The results were then sorted by a metric (quasi-P) that takes into account both the difference between the two groups and how consistent the values are within that group. Only p<0.05 were considered significant. To handle the complications presented by simultaneously testing thousands of proteins, we applied the False Discovery Rate (FDR) method (27). The top 250 protein groups by spectral count were selected and unsupervised hierarchal clustering was performed using MeV (Multiple Experiment Viewer – http://www.tm4.org/mev/).

Pathway analysis was performed using The Pathway Interaction Database National Cancer Institute (http://pid.nci.nih.gov) (28) as well as the STRING (Search Tool for the Retrieval of Interacting Genes) software (v.9.0) (http://string-db.org/) (29). STRING is a large database of known and predicted protein interactions. All the edges were supported by at least a reference from the literature or from canonical information stored in the STRING dataset. A confidence score was fixed at 0.4 (medium level). Cluster networks were created using the MCL algorithm which is included in the STRING website and a value of 2 was selected for all the analyses. To simplify large datasets, all proteins that met the filter criteria by IDPicker groups were filtered to include proteins containing 25–300 total spectral counts. Only proteins exclusively detected in HiBEC blebs when compared to both HRPTEpiC and BrEPC were retained.

RESULTS

Characterization of the proteome of apoptotic bodies from primary cultures

Firstly, we note that samples were isolated and run independently. We identified a total of 40,843 distinct peptides and 6,160 protein groups within apoptotic bodies from HiBEC, BrEPC, and HRPTEpiC (Table 1A) and similar numbers in those from BECs from PBC patients and controls (Table 1B). To eliminate proteins that had too few spectral counts to be statistically different, we reduced the data set to protein groups that were identified by at least 5 peptides spectra. An unsupervised clustering of the top 250 proteins by spectral count was able to appropriately separate the samples (Figure 1), suggesting consistent results and distinguishing characteristics between groups.

Table 1.

Number of unique peptides, spectra, and proteins identified for a given run of apoptotic bodies isolated from HiBEC, BrEPC, and HRPTEpiC (A), and of apoptotic bodies from BECs isolated from PBC patients and healthy controls (B).

A.	Distinct Peptides	Distinct Matches	Filtered Spectra	Protein Groups
BrEPC	7572	9120	14699	2881
	7979	9148	14456	3003
	8842	10131	14444	3247
	10069	11531	16635	3361
	10528	12275	19153	3310
	10978	12761	18375	3190
HiBEC	7005	8082	14270	2733
	6074	7111	13849	2601
	5766	6833	10922	2544
	4305	5114	10493	2085
HRPTEpiC	9875	11540	19218	3392
	7747	9208	15159	2919
	7245	8713	14065	2919
	8858	10563	16675	3191
	9711	11038	14455	3105
	7265	8688	14245	2730
B.	Distinct Peptides	Distinct Matches	Filtered Spectra	Protein Groups
Healthy controls	3743	4638	17479	1128
	3920	4652	15153	1177
	8090	9929	21127	1843
	6852	8182	20275	1709
	5504	6790	17151	1466
PBC patients	4923	6586	18949	1264
	4848	6271	16567	1276
	6702	8557	20781	1655

Figure 1.

Unsupervised clustering. The top 250 protein groups by spectral count were selected and unsupervised hierarchal clustering using MeV (Multiple Experiment Viewer – http://www.tm4.org/mev/) was performed. The figure shows clustering of apoptotic bodies isolated from HiBEC, BrEPC, and HRPTEpiC (A), and of apoptotic bodies from BECs isolated from PBC patients and healthy controls (B).

Both data sets were tested for possible differences. First, we compared the proteome of apoptotic bodies from commercial primary cultures (i.e. HiBEC, BrEPC, and HRPTEpiC); which yielded a total of 589 protein database entries when apoptotic bodies from BrEPC were compared to those from HiBEC and a total of 663 protein database entries when apoptotic bodies from HRPTEpiC were compared to those from HiBEC, with a quasi p-value of less than 0.05 and at least a 4-fold difference in spectral counts (log2(rate1/rate2), or rate ratio, higher than 2 or lower than −2). This list includes a number of proteins with artificially low p-values either because no spectra were identified in one of the groups (reported with “cv” value of “NA”).

We also looked for proteins that were specifically identified in apoptotic bodies from HiBEC when compared to apoptotic bodies from both bronchial and renal epithelial cells (Table 2). Only 11 proteins were found to be specific for apoptotic bodies of HiBEC with a quasi p-value of less than 0.05 and at least a 4-fold difference in spectral counts. Table 3A summarizes the function of the identified proteins and the related genes. Further, we also investigated the presence of proteins unique to apoptotic bodies from BrEPC and HRPTEpiC, and absent from HiBEC. We identified eight such proteins and they are listed in Table 3B.

Table 2.

List of proteins specifically located in apoptotic bodies from HiBEC compared to epithelial controls.

	BrEPC vs HiBEC			HRPTEpiC vs HiBEC
Proteins	2log (λ1/λ2)	quasi p- value fdr	Total Counts	2log (λ1/λ2)	quasi p- value fdr	Total Counts
A6NN80	−3.05	0.000961	1061	−2.21	0.002345	1189
ANXA6	−3.09	0.000880	1064	−2.18	0.002086	1204
B4DN38	−2.39	0.018616	81	−2.39	0.016801	81
GPC6	−4.16	0.007293	39	−1.39	0.017383	66
HSPB6	−6.62	0.004719	67	−2.20	0.001517	57
LRP1	−2.42	0.008753	55	−6.62	0.003810	67
PAPS2	−6.35	0.001450	111	−2.27	0.002121	97
Q6ZR44	−4.67	0.005526	72	−2.75	0.002872	55
RAB11A	−2.31	0.005947	160	−1.16	0.002963	95
SERPH	−2.41	0.000674	396	−1.23	0.013308	95
VGFR3	−5.76	0.015027	521	−6.98	0.011321	513

Top Ranked Proteins with differential spectral counts that are uniquely present in apoptotic bodies from HiBEC when compared to apoptotic bodies from both bronchial (BrEPC) and renal (HRPTEpiC) cells.

Table 3.

Encoded protein, gene and description of the specific HiBEC apoptotic bodies proteome (A) and of proteins located in apoptotic bodies from both bronchial (BrEPC) and renal (HRPTEpiC) cells but absent in HiBEC apoptotic bodies (B).

A.
Protein	Gene	Description
A6NN80	ANXA6	Putative uncharacterized protein
ANXA6	ANXA6	Annexin A6, Lipocortin VI, p68, p70, Protein III, Chromobindin-20, 67 kDa calelectrin, Calphobindin-II
B4DN38	TSC22D2	highly similar to Glucosamine--fructose-6-phosphateaminotransferase (isomerizing)
GPC6	GPC6	Secreted glypican-6
HSPB6	HSPB6	Heat shock protein beta-6, Heat shock 20 kDa-like protein p20
LRP1	LRP1	Low-density lipoprotein receptor-related protein 1, Alpha-2-macroglobulin receptor, Apolipoprotein E receptor, Low-density lipoprotein receptor-related protein 1 intracellular domain
PAPS2	PAPSS2	3’-phosphoadenosine 5’-phosphosulfate synthase 2, isoform CRAb, FLJ92929, highly similar to Homo sapiens 3’-phosphoadenosine 5’-phosphosulfate synthase 2
Q6ZR44	C20orf151	highly similar to Thioredoxin reductase
RAB11A	RAB11A	RAB 11A, member oncogene family; rab-11; ras-related protein Rab-11A
SERPH	SERPINH1	highly similar to Collagen-binding protein 2, Serpin peptidase inhibitor, clade H (Heat shock protein 47), member 1, (Collagen binding protein 1), isoform CRAa
VGFR3	SERPINB4	Tyrosine-protein kinase receptor
B.
Protein	Gene	Description
ANXA3	ANXA3	Annexin A3
H2AW	H2AFY2	Core histone H2A Inositol 1,4,5-trisphosphate receptor type 3, Type 3 inositol
ITPR3	ITPR3	1,4,5-trisphosphate receptor, Type 3 InsP3 receptor, IP3 receptor isoform 3, InsP3R3
K1C19	KRT19	Keratin, type I cytoskeletal 19, Cytokeratin-19, Keratin-19
K2C7	KRT7	Keratin, type II cytoskeletal 7, Cytokeratin-7 Laminin subunit gamma-2, Laminin 5 gamma 2 subunit, Kalinin/nicein/epiligrin 100 kDa subunit, Laminin B2t chain,
LAMC2	LAMC2	Cell-scattering factor 140 kDa subunit, Large adhesive scatter factor 140 kDa subunit
PYGB	PYGB	Glycogen phosphorylase, brain form
Q9HA11	ACAD9	Histone H2A

Proteome of intact cells versus apoptotic bodies

The presence of specific proteins within the apoptotic bodies was compared to the global proteome of naïve non apoptotic cells from the three epithelial cell types included in this study (HiBEC, HRPTEpiC, and BrEPC). We identified a total of 3,152 protein groups within HiBECs, HRPTEpiCs, and BrEPCs. Of the 11 proteins uniquely found in the apoptotic bodies of HiBEC cells, 4 of the 11 (ANXA6, LRP1, PAPS2, and SERPH) were found to be present in all three intact cell lines. One protein, HSPB6, was found only in intact HiBEC, but not intact HRPTEpiC or BrEPC. Interestingly, the other 6 proteins that were uniquely found in the blebs of HiBEC (A6NN80, B4DN38, GPC6, Q6ZR44, RAB11A AND VGFR3), were not found in intact cells. Finally, of the 3,152 protein groups, there were only 3 proteins found in intact HiBEC cells, but not in HiBEC apoptotic bodies (ANXA3, PYGB, and ITPR3). Hence, the presence of proteins specifically located in the blebs of isolated HiBEC cells appear secondary to apoptosis.

Proteome of apoptotic bodies from BECs from patients and controls

Comparison of the proteome of BECs from PBC patients and healthy controls yielded a total of 165 protein database entries, with a quasi p-value of less than 0.05 and at least a 4-fold difference in spectral counts (log2(rate1/rate2), or rate ratio, higher than 2 or lower than −2). We looked for proteins that were specifically located in apoptotic bodies from PBC patients compared to apoptotic bodies from healthy controls (Table 4). Six proteins were found to be specific for apoptotic bodies of PBC patients with a quasi p-value of less than 0.05 and at least a 4-fold difference in spectral counts. Further, we also investigated the presence of proteins unique to apoptotic bodies from healthy controls that are absent in those from PBC patients. We identified only 2 such proteins and they are listed in Table 4. Table 5 summarizes the function of the proteins and related genes.

Table 4.

List of proteins specifically located in apoptotic bodies isolated from BECs from PBC patients and controls.

A.
Proteins	2log (λ1/λ2)	quasi p-value fdr	Total Counts
CPN1	−2.28	0.00017	52
ITIH2	−2.10	0.00134	1493
C9	−2.44	0.00180	288
FGG	−2.96	0.00180	192
FGA	−2.69	0.01155	155
SERPINF2	−2.08	0.01752	94
B.
Proteins	2log (λ1/λ2)	quasi p-value fdr	Total Counts
APOC2	32.44	0.007316	16
DRG1	32.03	0.015156	12

Top ranked proteins with differential spectral counts that are uniquely present in apoptotic bodies from BECs from PBC patients when compared to blebs from BECs from healthy donors (A) and uniquely present in apoptotic blebs from BECs from healthy donors but absent in those from PBC patients (B).

Table 5.

Gene, encoded protein and description of proteins located in apoptotic bodies from BECs from PBC patients and healthy controls.

Protein	Gene	Description
CPN1	CPN1	Carboxypeptidase N catalytic chain
ITIH2	ITIH2	Inter-alpha (Globulin) inhibitor H2
C9	C9	Complement component C9
FGG	FGG	Fibrinogen gamma chain
FGA	FGA	Fibrinogen alpha chain
SERPINF2	SERPINF2	Alpha-2-antiplasmin
APOC2	APOC2	Apolipoprotein C-II
DRG1	DRG1	Developmentally-regulated GTP-binding protein 1

Pathway analysis

Using the STRING database a proteome-scale interaction network was created. This database includes interactions from the published literature describing interactions as well as genome analysis based on domain fusion, phylogenetic profiling and gene neighborhood concepts. Accordingly, a confidence score for every protein-protein association was assigned to the network. To minimize false positives as well as false negatives, all interactions tagged as “low-confidence” (<0.4) in STRING database were eliminated from this study. The analysis resulted in the Protein interaction network for LRP1 (also known as A2MR), RAB11A and ANXA6 using the STRING software as illustrated in Figure 2. Stronger associations are represented by thicker lines. We also examined the pathways found in apoptotic bodies using The Pathway Interaction Database, NIH. This proteomic pathway analysis identifies inflammation pathways (Table 6).

Figure 2.

LRP1, ANXA6, RAB11A, FGG, and FGA network display. Nodes are either colored (if they are directly linked to the input protein) or white (nodes of a higher iteration/depth). Edges, i.e. predicted functional links, consist of up to eight active prediction methods: neighborhood, gene fusion, co-occurrence, co-expression, experiments, databases, textmining.

Table 6.

Pathway analysis of the BEC apoptotic bodies proteome.

Protein	Pathway
ANXA6	- EGFR/Ras signaling pathway
LRP1	- amb2 Integrin signaling
	- PDGFR-beta signaling pathway
	- Urokinase-type plasminogen activator (uPA) and uPAR-mediated signaling
	- ERK pathway
	- nuclear factor-kappa B
RAB11A	- Notch signaling pathway
	- Thromboxane A2 receptor signaling
	- Arf6 downstream pathway
	- IL8- and CXCR2-mediated signaling events
FGG	- IL6-mediated signaling events
	- Beta1, beta2, and beta3 integrin cell surface interactions
	- Urokinase-type plasminogen activator (uPA) and uPAR-mediated signaling
FGA	- Beta1, beta2, and beta3 integrin cell surface interactions
	- Urokinase-type plasminogen activator (uPA) and uPAR-mediated signaling
SERPINF2	- fibrinolysis pathway
APOC2	- metabolism of lipids and lipoproteins

Candidate proteins were uploaded into The Pathway Interaction Database of the National Cancer Institute after manually filtering for protein groups as described.

DISCUSSION

We report the use of shotgun proteomics to identify proteins that might be part of the induction of a signature inflammatory cytokine response in PBC. The cells most closely similar to the true target of disease are normal HiBECs from unaffected individuals. These are the most abundant available source of cells that would mimic biliary cells at the time of induction of disease. Importantly it is apoptotic bodies derived from these normal HiBECs that stimulate a pro-inflammatory response (8, 12), but there may be aspects of their proteome that differ from those found in apoptotic bodies isolated from HiBECs of patients at the commencement of disease. Such patients are difficult to identify and there are ethical issues with obtaining sufficient tissue for experimentation. However as an alternative we also examined apoptotic bodies present in the residual biliary epithelial cells that have survived immune destruction in PBC patients. These patients are many years after the disease induction event and their liver has been under sustained immunological attack for decades. Thus these HiBECs may now be somewhat different from such cells at the time of disease initiation. Further, unfortunately the extent of biliary destruction in PBC patients is quite severe and the low number of such surviving cells makes it technically impossible to derive sufficient apoptotic blebs to determine if such late stage cells are capable of inducing a pro-inflammatory response, but there are enough cells to provide a sample for proteomic analysis. The fact that proteomic analysis of apoptotic bodies from these cells suggest that they differ from apoptotic bodies of normal human HiBECs is noteworthy.

Immunological attack appears to have affected the repertoire of proteins expressed. Indeed, these cells may be thought of as having lived in an environment that would continuously select against production of molecules that would be stimulatory to the immune response. This almost “Darwinian” selection may explain why such cells do not produce the same set of unique proteins present in the apoptotic bodies of normal HiBECs. The unique proteins they do produce may be candidates for proteins that down regulate or ameliorate immune attack in such a way as to allow survival of these cells. We chose bile salts for apoptosis induction to mirror hepatic physiology. Bile salts accumulating in the liver during cholestatis trigger liver injury and subsequent fibrosis. A constituent of the hydrophobic bile salt, GCDC, induces apoptosis of hepatocytes at a concentration of 50 µM (30). GCDC has also an apoptotic effect in cholangiocytes (24). Future studies should take advantage of a variety of other epithelial cells and including comparisons of large versus small bile ducts from explants of PBC and non-PBC liver diseases.

Through a rigorous and sensitive comparison of the proteome of apoptotic bodies from HiBEC and non-stimulatory epithelial apoptotic bodies, we detected 11 proteins specifically located within apoptotic bodies from HiBEC. These proteins are largely involved in immune (and in the right context, autoimmune) reactions including NF-kB activation, ERK pathway, Notch signaling pathway, and IL8- and CXCR2-mediated signaling events (31–34). Of note, four proteins out of these eleven proteins were detected in all three intact cells (HiBEC, HRPTEpiC and BrEPC) and one protein was detected in only intact HiBEC. This finding reinforces our thesis that apoptotic bodies play a pathogenic role in biliary inflammation (35). Interestingly we were not able with this technique to detect PDC-E2 within HiBEC apoptotic bodies. It may be that the signature peptides that lead to the identification of PDC-E2 are not soluble, or are degraded/eliminated by the protocol utilized since there is limited knowledge on the susceptibility of peptides to such procedures.

In this context, among the proteins located within HiBEC apoptotic bodies, GPC6 looks particularly interesting as it is a secreted protein that has marked functional effects in the CNS (38). Six members of the glypicans family have been identified in mammals (GPC1-GPC6). Glypicans act as regulators of various cytokines, including Wnts, Hedgehogs, and bone morphogenetic proteins (39–41). No functional information on GPC6 is yet available. Data available from public databases (http://www.proteinatlas.org/index.php) note moderate immunostaining in hepatocytes, but prominent staining in the epithelial lining of the gall bladder (ie. cholangiocytes). Hedgehog signaling pathways are well known to be involved in the fibroproliferative response upon cholestatic biliary epithelial cell injury in PBC (43).

It is interesting to note and discuss LRP1 and its potential role in the pathogenesis of PBC. Firstly, we should emphasize that PBC reoccurs following liver transplants and therefore there does not appear to be a PBC-specific biliary phenotype. Although this intense proteomic approach provides significant new data on biliary biology, we cannot specifically address etiology, including the potential role of either environmental agents or infections. However, lipoprotein receptors from the low density lipoprotein receptor (LRP) family are membrane proteins with multifunctional scavenger and signaling receptor functions (44). The largest member of this family, the LDL receptor-related protein 1 (LRP1), also known as cluster of differentiation 91 (CD91), serves as a multifunctional endocytic receptor with a broad range of ligands. CD91 is pivotal in the regulation of heat shock proteins (HSPS) mediated immune co-stimulation (45). CD91 and members of the scavenger receptor family efficiently direct Hsp70:peptide complexes into the MHC class II presentation pathway and thus enhance antigen-specific CD4+ T cell responses (46).

RAB11A is part of the small GTPase superfamily, Rab family. RAB11A is involved in both constitutive and regulated secretory pathways, and takes part in protein transport. RAB11A is a well-known marker for protein trafficking, sorting, and recycling in the endosomal pathway. RAB11A is also an important regulator of membrane delivery during cytokinesis. Importantly, the ability of toll-like receptors (TLRs) to activate innate immunity depends on their transport to pathogen-containing organelles.

Although a specific factor in apoptotic bodies of HiBECs may be the factor responsible for the innate immune response in PBC patients (8), it is possible that it is the loss or absence of down modulating proteins found in the other tissues that may be part of the pathophysiology. Therefore, we also investigated the presence of proteins unique to apoptotic bodies from BrEPC and HRPTEpiC, but absent in those from HiBEC. Among these molecules, IPR3, which regulates the release of Ca(2+) from intracellular stores, might play an important role in the immune response and uptake of apoptotic bodies. Furthermore, it is also involved in both naïve CD4 T cells cytokine release (51) and FcγR-mediated phagocytosis (52). Other findings of interest in this set of proteins are cytokeratines 19 and 7, as well as laminin subunit gamma-2, which are known to be located in epithelial cells (including BrEPC and HRPTEpiC) in high concentration. An accumulation of the linker-histone (histone 1) as well as the core-histones (histone 2A, histone 2B, histone 3, histone 4) have been demonstrated to be widely located in apoptotic bodies (53), so their absence from HiBECs is unexpected. Finally, we note that Annexin A3 is enriched in the apoptotic bodies of renal tubules and bronchial epithelium, but it is annexin A6 that is enriched in biliary epithelial cells.

Our data reflect the results of a detailed differential analysis of the proteome of apoptotic bodies of HiBEC that has led to the identification of proteins uniquely expressed by HiBECs with a potential pathogenic role in PBC. Indeed this data highlights the role of the innate immune system activation in PBC, in which NF-kB pathway involvement has been strengthened in recent GWAS studies (54–56). Further, these results also suggest that the protein content within apoptotic bodies from HiBEC could be involved in biliary proliferation. After liver injury, Notch signaling pathway activation leads to biliary regeneration and proliferation through bipotent hepatic progenitor cells polarization (57). Finally, IL-8 and CXCR2 activation may lead to inflammatory cell recruitment (58).

The current results therefore extend our knowledge about the immunological properties of HiBECs, which indicate that they are more than an innocent victim in the pathogenesis of PBC. However, it is important to emphasize that proteomic pathway analysis is primarily a tool to generate a hypothesis on the mechanisms and protein involvement in the local inflammatory response and will require further testing. Future work will address whether the activity of these proteins in apoptotic bodies and their ability to directly or indirectly activate and/or modulate macrophage function.

Abbreviations

PBC

Primary biliary cirrhosis

PDC-E2

E2 subunit of the pyruvate dehydrogenase complex

BECs

biliary epithelial cells

HiBECs

human intrahepatic biliary cells

LC-MS/MS

liquid chromatography-tandem mass spectrometry

HRPTEpiC

Human Renal Proximal Tubular Epithelial cells

BrEPC

and human bronchial epithelial cells

FBS

fetal bovine serum

GLM

Generalized Linear Modeling

FDR

False Discovery Rate

LRP

low density lipoprotein receptor

LRP1

LDL receptor-related protein 1

CD91

cluster of differentiation 91

HSPS

heat shock proteins

TLRs

Toll-like receptors