The role of macrophage migration inhibitory factor in autoimmune liver disease


  • Potential conflict of interest: Dr. Bucala is a coinventor on patent applications describing the utility of MIF genotype determination.

  • This work was supported by the National Institutes of Health (NIH; the clinical/translational, cell and molecular biology and morphology core facilities of P30 DK34989, the Yale Liver Center), an NIH Digestive Diseases Research Core Centers (DK034989-28) Liver Center Pilot Project Award (to D.A.), an American Liver Foundation Postdoctoral Autoimmune Hepatitis Grant (to D.A.), NIH T32 DK007356-33 (to D.A.), the German Research Foundation (DFG; BE1977/4-2 and SFB/TRR57-P07; to J.B.), NIH DK25636 (to J.L.B.), NIH AR050498 and AR049610 (to R.B.), and N01-HHSN272201100019C (to R.B.).


The role of the cytokine, macrophage migration inhibitory factor (MIF), and its receptor, CD74, was assessed in autoimmune hepatitis (AIH) and primary biliary cirrhosis (PBC). Two MIF promoter polymorphisms, a functional −794 CATT5-8 microsatellite repeat (rs5844572) and a −173 G/C single-nucleotide polymorphism (rs755622), were analyzed in DNA samples from over 500 patients with AIH, PBC, and controls. We found a higher frequency of the proinflammatory and high-expression −794 CATT7 allele in AIH, compared to PBC, whereas lower frequency was found in PBC, compared to both AIH and healthy controls. MIF and soluble MIF receptor (CD74) were measured by enzyme-linked immunosorbent assay in 165 serum samples of AIH, PBC, and controls. Circulating serum and hepatic MIF expression was elevated in patients with AIH and PBC versus healthy controls. We also identified a truncated circulating form of the MIF receptor, CD74, that is released from hepatic stellate cells and that binds MIF, neutralizing its signal transduction activity. Significantly higher levels of CD74 were found in patients with PBC versus AIH and controls. Conclusions: These data suggest a distinct genetic and immunopathogenic basis for AIH and PBC at the MIF locus. Circulating MIF and MIF receptor profiles distinguish PBC from the more inflammatory phenotype of AIH and may play a role in pathogenesis and as biomarkers of these diseases. (Hepatology 2014;59:580–591)




autoimmune hepatitis


alkaline phosphatase


alpha-smooth muscle actin


antinuclear Ab


alanine aminotransferase


chronic hepatitis B


confidence interval


enzyme-linked immunosorbent assay


extracellular signal-regulated kinase


genomic DNA


human leukocyte antigen


horseradish peroxidase


hepatic stellate cells


half-maximal inhibitory concentration




immunoglobulin G


monoclonal antibody


macrophage migration inhibitory factor


molecular weight


not significant


optical density


odds ratio


primary biliary cirrhosis


phosphate-buffered saline


systemic lupus erythematosus


single-nucleotide polymorphism


T-helper cell type 1.

Macrophage migration inhibitory factor (MIF) is a proinflammatory cytokine that mediates the host response to infection and stress by activating innate, and adaptive immune pathways.[1, 2] MIF is released both from immune and neuroendocrine cells and acts to counter-regulate the immunosuppressive effect of glucocorticoids.[3, 4] Recently, MIF has been associated with a number of autoimmune diseases, including rheumatoid arthritis, systemic sclerosis, and inflammatory bowel disease.[5-8] Two polymorphisms with clinical significance have been identified in the MIF promoter: a functional microsatellite repeat sequence −794 CATT5-8 (rs5844572), where higher MIF expression occurs with increased repeat numbers, and a −173 G/C (rs755622) single-nucleotide polymorphism (SNP), where the C allele is associated with risk of disease, most likely by linkage disequilibrium. MIF polymorphisms and circulating MIF levels in systemic lupus erythematosus correlate with disease susceptibility and multiorgan involvement.[9]

In the present study, we investigated MIF expression and MIF gene associations in patients with autoimmune hepatitis (AIH) and primary biliary cirrhosis (PBC). Our findings suggest a distinct MIF genetic and serum profile in AIH, compared to PBC. We also report on the characterization of a circulating form of the MIF receptor (CD74) that neutralizes MIF activity and differs in expression in AIH versus PBC patients. Furthermore, the ratio of serum CD74 and MIF values correlated with serum alanine aminotransferase (ALT) in patients with AIH who experienced relapse, as measured by serum liver tests. A soluble form of CD74 is released by human hepatic stellate cells (HSC) in vitro after interferon-gamma (IFN-γ) stimulation, and its production in vivo may contribute to the differences in inflammatory manifestations of these two autoimmune liver diseases.

Materials and Methods

Patient Cohorts

An AIH cohort (N = 52) of patients diagnosed by established criteria[10] was recruited from the Yale Liver Clinics (New Haven, CT). Two PBC cohorts comprised one from the Yale Liver Clinics (N = 42) and a second group with similar clinical, geographic, and ethnic characteristics from Tufts Medical Center (N = 267; Boston, MA). Overlap syndromes[11] were excluded. Sera (N = 71) or genomic DNA (gDNA; N = 286) from healthy controls were obtained from a Yale biospecimen repository. The study was approved by Yale's Human Investigation Committee.

Serum MIF and CD74 Enzyme-Linked Immunosorbent Assay

Serum MIF was measured by sandwich enzyme-linked immunosorbent assay (ELISA), as previously described.[9] Competitive sandwich ELISAs for circulating CD74 and MIF-CD74 complexes were developed (see Supporting Materials). The ratio of CD74/MIF was calculated by dividing the molar serum concentration ratio of circulating CD74 (20 kDa) by MIF (12.5 kDa).

Histologic Analysis

Paraffin-embedded, stored biopsy tissue was stained with anti-CD74 (LN-2; Santa Cruz Biotechnology, Santa Cruz, CA) as primary antibody (Ab) and labeled with anti-IgG (immunoglobulin G) alkaline phosphatase (ALP). The tissue was double-stained for alpha-smooth muscle actin (α-SMA) and revealed with 3,3′-diaminobenzidine. For MIF tissue staining, the primary Ab was anti-MIF IIb in phosphate-buffered saline (PBS), and secondary Ab was anti-rabbit IgG horseradish peroxidase (HRP; Dako, Carpinteria, CA).[12] Isotype control Ab-stained tissues were generated using nonimmunized mouse IgG.

MIF Polymorphism Analyses

DNA was extracted from serum samples using the Easy-DNA Kit (Invitrogen, Carlsbad, CA) with MIF −794 CATT5-8 (rs5844572) polymorphism analysis performed as previously described.[9] MIF −173 G/C SNP (rs755622) analysis was performed by pyrosequencing.[9] Figure 1 illustrates the MIF gene.

Figure 1.

Diagram of the MIF gene showing the −794 CATT5-8 microsatellite repeat (rs5844572), the −173 G/C SNP (rs755622), its exonic structure, and predicted transcription-factor-binding sites. Numerical prefixes refer to nucleotide distance (in base pairs) upstream from the transcription start site.

Methods for the cell-culture stimulation of HSCs, whole-liver lysate analysis, and assays for MIF-CD74 epitope and competitive MIF-CD74-binding assays are described in the Supporting Materials.

Characterization of Circulating CD74

The molecular weight of serum-circulating CD74 was estimated by western blotting and compared with full-length CD741-232 expressed in COS-7 cells.[13] CD74 was immunoprecipitated from serum by incubation with 12 μg/mL of anti-CD74 for 3 hours at 4°C, followed by adding 50 μL of Protein G Sepharose beads (GE Life Sciences, Piscataway, NJ), overnight incubation at 4°C, and PBS wash. Loading buffer (NuPage LDS; Invitrogen) with 10% β-mercaptoethanol was added, the sample run by sodium dodecyl sulfate polyacrylamide gel electrophoresis, and proteins revealed by western blotting with anti-CD74 (clone LN-2; catalog no.: SC-6262 Santa Cruz and IgG HRP-linked secondary Ab (catalog no.: 7076S; Cell Signaling Technology, Danvers, MA). Sera or COS-7-expressed recombinant CD741-232 were deglycosylated before immunoprecipitation by incubation of 250 μL serum with 30 μL of a deglycosylation enzyme mix (New England Biolabs, Ipswich, MA) containing PNGase F glycerol free, endo-α-N-acetylgalactosaminidase, neuraminidase, β1-4 galactosidase, and β-N-acetylglucosaminidase. AIH and PBC serum were also analyzed by western blotting with an Ab-specific for the intracellular domain of the molecule (CD74 PIN.1 Ab; catalog no.: SC-47742; Santa Cruz Biotechnology).

Neutralization of Serum MIF by CD74

Test sera were incubated with human primary skin fibroblasts (3.6 × 105 cells/well), with or without neutralizing anti-CD74 Ab, anti-MIF monoclonal Ab (mAb; IIID.9; 100 μg/mL), or recombinant CD7473-232 (100 ng/mL), followed by the addition of recombinant human MIF (2.7 nM) protein. For immunoadsorption of CD74, two AIH sera and two PBC sera (100 μL per sample) were preincubated with anti-CD74 Ab (clone LN2; Santa Cruz Biotechnology) and then immunoprecipitated with protein G beads. Serum was also absorbed with a control IgG (Hb49). MIF signal transduction was measured by MIF-dependent extracellular signal-regulated kinase (ERK)1/2 phosphorylation as previously described.[13, 14]

Statistical Analysis

Serum ELISA MIF and CD74 values were analyzed as continuous variables. Because of a non-normal distribution of CD74 and MIF concentrations, Mann-Whitney's nonparametric U test was performed for all serum values. For the MIF polymorphism data, 2LD[15] was used to calculate the linkage disequilibrium between the CATT and SNP in healthy controls. Both markers were found to be in Hardy-Weinberg's equilibrium by PEDSTATS.[16] A chi-square test was performed to test the overall association of the alleles in CATT between case and controls. Implemented in R software (, a logistic regression was fitted with the disease status as response variable, and for the −794 CATT5-8 and −173 G/C SNP alleles as predictor variables. The SNP and each CATT allele (5, 6, and 7) were coded as separate variables in an additive model. The CATT8 allele was omitted because of its extremely low frequency. To address multiple testing, calculations were made based on 10,000 permutation tests that control the family-wise error rate in multiple comparisons.


Patient Demographics

Demographic data for the clinically characterized cohorts of AIH and PBC patients are presented in Table 1. Most patients were Caucasian (AIH, 85%; PBC, 91%) and female (AIH, 83%; PBC, 91%), in agreement with typical U.S. patient characteristics.[17, 18] Most patients had blood collected as outpatients during routine follow-up care. The AIH cohort received a median of 5 mg of prednisone, daily. No PBC patients received glucocorticoids.

Table 1. Demographic Data of Patients With AIH and PBC and Healthy Controls
Demographic DataAIH N = 52 (Range)PBC N = 42 (Range)Controls N = 71 (Range)
  1. Medication doses and liver test values correspond to the time of serum collection.

  2. Abbreviations: IAHG, International Autoimmune Hepatitis Group[10]; AMA, antimitochondrial Ab; NA, not available.

Age at blood collection, median54 (20-79)62 (30-84)45.0 (21-87)
Female, %82.790.570.4
Caucasian, %84.690.573.2
Age at diagnosis, median41 (1.5-71)52 (26-77)NA
Cirrhosis (imaging or biopsy), %55.89.5NA
IAHG score, median17 (6-25)NANA
PBC stage, medianNA2 (0-4)NA
ANA positive, %77.033.3NA
AMA positive, %3.878.6NA
Prednisone, median daily mg5.0 (0-60)NANA
Azathioprine, median daily mg50 (0-150)NANA
Ursodeoxycholic acid, median daily mgNA1,000 (0-2,000)NA
ALT, median U/L27 (5-383)29 (10-104)NA
ALP, median U/L66 (34-294)116 (62-423)NA
Total bilirubin, median mg/dL0.51 (0.5-37.1)0.59 (0.25-50)NA

Serum Levels of MIF Elevated in Patients With AIH and PBC

The median serum MIF level was significantly elevated, but similar, in both the AIH and PBC cohorts, compared to healthy controls (11.1 and 9.58 versus 2.63 ng/mL, respectively; P < 0.001; Fig. 2A). Median serum MIF levels were slightly, but not significantly, different between patients taking 0-5 mg, compared to >5 mg, of daily prednisone (9.69 [N = 29] versus 11.44 [N = 23] ng/mL; P = not significant [NS]). Neither stage of fibrosis nor autoAbs correlated with serum MIF values in AIH or PBC (data not shown).

Figure 2.

(A) Median serum MIF levels in healthy controls and patients with AIH and PBC (AIH, 11.1 ± 9.55 ng/mL; PBC, 9.58 ± 12.12 ng/mL; controls, 2.63 ± 3.45 ng/mL; *P < 0.001 for AIH versus controls and for PBC versus controls). There was no significant difference between serum MIF among the two patient cohorts (P = NS). Bottom, middle, and top lines of the box demarcate the 25th, 50th, and 75th percentiles, respectively, and the vertical lines show the maximum and minimum values. (B) Immunohistochemistry staining of MIF in human liver tissue from a patient with AIH, a patient with PBC, and a healthy control. The AIH section displays dark MIF staining (dark brown) in 100% of counted hepatocytes, whereas the PBC section has lighter MIF staining in 100% of counted hepatocytes, with more-prominent biliary staining. There was no hepatocyte staining in control tissue. All three sections revealed mild biliary epithelial staining. Images shown are representative of 11 patients examined (AIH = 2, PBC = 4, and control = 5).

MIF and the MIF Receptor Expression Increased in Liver Tissue

MIF is expressed by hepatocytes and infiltrating mononuclear cells in chronic hepatitis B (CHB) infection.[19] Intracellular MIF staining in hepatocytes, biliary epithelial cells, and infiltrating mononuclear cells was increased in AIH and PBC. In AIH, there was more-consistent staining in hepatocytes, (Fig. 2B). Only biliary epithelia stained in control sections. The MIF receptor, CD74, and α-SMA, were expressed in the sinusoids, but not in hepatocytes, in control sections (Fig. 3A). Distinct, but faint, perimembrane CD74 staining was noted in the hepatic sinusoids in mild chronic AIH, whereas diffuse sinusoidal and prominent hepatocyte staining of CD74 was present in severe acute AIH. Analysis of PBC tissue revealed distinct CD74 staining in sinusoids and granulomatous tissue. These findings indicate that both MIF and its receptor protein, CD74, are expressed in AIH and PBC liver, with variable cellular staining between mild versus severe AIH.

Figure 3.

(A) Immunohistochemical staining for the MIF receptor, CD74 (blue), and α-SMA (brown) in liver tissue sections from 2 patients with AIH (mild chronic versus severe acute), 1 patient with PBC, and control liver tissue. All three sections revealed CD74 in the sinusoidal areas; however, in severe AIH, there was prominent staining diffusely in hepatocytes. Images shown are representative of 15 patient sections examined (AIH = 2, PBC = 5, and control = 6). (B) Detection of serum circulating CD74 by ELISA. Plot of immunoreactive, circulating CD74 median concentrations (controls, 14.2 ± 107.75 ng/mL; AIH, 38.13 ± 430.24 ng/mL; PBC, 171.57 ± 1,195.98 ng/mL; *P < 0.001 for AIH versus controls and for PBC versus controls; **P < 0.001 for PBC versus AIH). Bottom, middle, and top lines of the box demarcate the 25th, 50th, and 75th percentiles, respectively, and vertical lines show the maximum and minimum values. (C) Calculated median CD74/MIF ratio based on conversion of the serum concentrations (ng/mL) ratio into molar concentrations (controls, 2.32 ± 33.54; AIH, 3.13 ± 82.56; PBC, 15.25 ± 69.38; *P < 0.0001 for PBC versus AIH; **P < 0.0001 for PBC versus controls). (D) The serum CD74/MIF ratio correlates with serum ALT in patients with active disease. Negative correlation was found between CD74/MIF and ALT in longitudinal samples from 4 patients with AIH and at least one elevation in ALT > 2 × ULN (24 time points). Coefficient of −0.4762; P = 0.0338.

Association Between MIF Gene Polymorphisms and AIH Versus PBC

MIF is encoded in a functionally polymorphic locus (22q11.2) that is linked to the susceptibility and clinical severity of several autoimmune diseases[5-9] (Fig. 1). To determine whether the two MIF promoter polymorphisms (−794 CATT5-8 and −173 G/C SNP) correlated with disease susceptibility or phenotype, we analyzed gDNA from 306 patients with PBC and 45 with AIH. To eliminate population stratification,[20] analysis was limited to Caucasian patients. DNA and serum protein was analyzed from the Yale cohort, whereas the Tufts Medical Center cohort was used only for genotype analysis. Allelic data were compared to a healthy Caucasian control population (N = 286) from a repository at Yale.

MIF −794 CATT5-8 Alleles

Two hundred and sixty-two of the three hundred and six PBC patients were genotyped at the CATT locus along with 41 AIH and 286 healthy controls. Overall, we found a statistically significant difference of the allele frequencies in PBC, compared to healthy controls (chi-square test: p = 0.0123; Table 2). We also compared the allele frequency for each individual CATT5-8 allele. In particular, the low-expression, MIF −794 CATT5 allele was more frequent in PBC patients (31.1%) than in healthy controls (23.8%), with an odds ratio (OR) for PBC risk of 1.34 (confidence interval [CI]: 1.05-1.7; P = 0.0165). The P value on permutation testing was 0.0331. Conversely, a decrease in frequency of the high-MIF-expression CATT7 allele was evident in the PBC patients (OR, 0.628; CI, 0.43-0.93; P = 0.0188), with a P value on permutation testing of 0.0389. A trend toward this difference could also be shown for PBC by genotype analysis of each allele.

Table 2. Allele and Genotype Frequencies of MIF −794 CATT5-8 and −173 G/C SNP for Caucasian Patients With AIH and PBC, Compared to Healthy Caucasian Controls
 Healthy Controls (%)AIH (%)PBC (%)
  1. Adjusted P values were calculated by permutation testing to account for multiple comparisons.

  2. 1. Allele:

  3. AIH or PBC versus healthy controls:

  4.  CATT5: *PBC versus healthy controls (OR, 1.34; CI, 1.05-1.7; P = 0.0165, adjusted P = 0.0331.

  5.  CATT7: AIH versus healthy controls (OR, 1.68; CI, 1.0-2.84; P = 0.051; adjusted P = 0.114).

  6. PBC versus healthy controls (OR, 0.63; CI, 0.43-0.93; P = 0.0188, adjusted P = 0.0389).

  7. AIH versus PBC:

  8.  CATT5: AIH versus PBC (P = 0.627; adjusted P = 0.899).

  9.  CATT7: AIH versus PBC (P = 0.00102; adjusted P = 0.0013).

  10. 2. Genotype:

  11. AIH or PBC versus healthy control:

  12.  CATT55: §PBC versus healthy controls (OR, 1.81; CI, 1.09-3.01; P = 0.022; adjusted P = 0.099).

  13.  CATT77: AIH versus healthy controls (OR, 3.83; CI, 1.24-11.85; P = 0.0196; adjusted P = 0.0704).

  14.  PBC versus healthy controls (OR, 0.43; CI, 0.13-1.38; P = 0.156; adjusted P = 0.639).

  15. AIH versus PBC:

  16.  CATT55: AIH versus PBC (P = 0.774).

  17.  CATT77: AIH versus PBC (P = 0.00158; adjusted P = 0.0074).

−794 CATT5-8(N = 286)(N = 41)(N = 262)
−173 C/G(N = 286)(N = 45)(N = 304)

The allele frequencies between the PBC and AIH patients overall were significantly different (chi-square test: P = 0.0105). When the frequencies for each allele were compared, the high-expression MIF −794 CATT7 allele was 2.59 times less frequent in PBC than in AIH patients (7.3% versus 20.5%; P = 0.00102). The P value on permutation testing was 0.0013 (Table 2). Additionally, we found a significant decrease in the MIF −794 CATT77 genotype frequency in PBC patients compared to AIH patients (1.5% versus 11.4%; P = 0.00158) with a P value on permutation testing of 0.0074. There also was a trend toward increased CATT77 genotype in AIH patients, compared to healthy controls. These findings suggest that the MIF genetic profile for PBC may be distinct from AIH and healthy controls, with a notable decreased frequency of the high-expression MIF −794 CATT7 allele in PBC.

MIF −173 G/C Alleles

A total of 304 PBC and 45 AIH patients and 286 healthy controls were genotyped at the −173 G/C SNP. The −173 G/C SNP is only 621 bases upstream from the −794 CATT5-8 microsatellite site and most likely reflects MIF promoter functionality on the basis of linkage disequilibrium.[9] We found evidence of linkage disequilibrium between the −794 CATT5-8 and −173 G/C SNP loci in our control group, with D′ of 0.63 (P < 1e-10), in accord with earlier reports in Caucasian populations.[20] There were no significant differences between the −173 G/C alleles or genotypes among the cohorts, (Table 2). There was also no relationship between serum MIF concentrations and different MIF alleles (data not shown). Such correlations have not been uniformly observed in other diseases,[21] possibly because the plasma compartment may not accurately reflect tissue-specific differences in MIF expression.

A Soluble Form of the MIF Receptor (CD74) Identified in Autoimmune Liver Disease

Because the MIF receptor, CD74, is expressed by several cell types within the inflamed liver (Fig. 3A) and cytokine receptor shedding is a well-described immunoregulatory mechanism for the control of cytokine pathways,[22] we developed a two-Ab, competitive sandwich ELISA to test for the presence of circulating CD74 in serum and analyzed samples from the cohorts used previously for assessing MIF serum concentrations. A significant increase in the median level of immunoreactive, circulating CD74 protein in patients with AIH was observed, when compared to controls (38.13 versus 14.2 ng/mL; P < 0.001; Fig. 3B). As with serum MIF, median serum CD74 levels were not significantly different between patients taking 0-5 mg, compared to >5 mg, of daily prednisone (44.8 [N = 29] versus 33.1 [N = 23] ng/mL; P = NS.

Surprisingly, significantly higher levels of circulating CD74 were present in PBC sera, compared to both the AIH cohort as well as controls (171.57 versus 14.2 ng/ml; P < 0.001). Although we did not find a correlation between serum MIF and circulating CD74 in AIH (P = NS), the correlation coefficient in the PBC cohort was 0.456 (P = 0.002). The marked serological difference between AIH and PBC was further emphasized when calculated CD74/MIF ratios were expressed after converting CD74-to-MIF serum concentration (ng/mL) ratios into molar values (Fig. 3C). Furthermore, the CD74/MIF ratio negatively correlated with serum ALT values in a small sample of patients with AIH who experienced a relapse measured by liver enzymes and had serial serum measurements (Fig. 3D). No correlation was identified between serum ALT and the CD74/MIF ratio in the PBC cohort (data not shown).

Immunoblottings confirmed the presence of a circulating CD74 protein from AIH, PBC, and controls. CD74 is a type II receptor protein expressed in four possible isoforms that depend on alternative translation initiation and the insertion of an additional C-terminal exon.[23] Using cell lysates from COS-7 cells expressing recombinant full-length CD741-232 protein as a molecular-weight (MW) standard,[13] CD74 was immunoprecipitated from human serum. Sera contained a single, immunoreactive band of approximately 25 kDa, compared to the 31-kDa band for recombinant CD741-232 (Fig. 4A). Given that full-length CD741-232 has a calculated MW of 26.5 kDa, the additional 4.5 kDa present in the 31-kDa control band is consistent with known post-translational glycosylation of CD74 on its extracellular domain.[23] To confirm that post-translational glycosylation occurs in circulating CD74 and recombinant CD741-232, we treated serum with deglycosylases before immunoprecipitation. This step produced a clear shift in MW of serum CD74 from 25 to approximately 20 kDa (Fig. 4B, left panel). The recombinant CD741-232 protein also exhibited a 5-kDa band shift with deglycosylation (Fig. 4B, right panel). Therefore, the 6-kDa difference in size between CD741-232 and circulating serum CD74 cannot be attributed to glycosylation and is likely the result of additional proteolysis of full-length CD74 to yield a truncated protein. Proteolytic cleavage at the transmembrane portion of CD74 during regulated intramembrane proteolysis in B lymphocytes is known to produce a 6-kDa (58-amino-acid) cytosolic fragment[24, 25] (shown in Fig. 4C). Such cleavage also may account for the difference in band size observed between serum-circulating CD74 and full-length CD741-232. These results suggest that circulating serum CD74 is a membrane-truncated protein that may be released after processing by proteolysis. We also tested for the presence of a circulating CD74 intracellular fragment in serum by performing western blotting of AIH and PBC serum using an anti-CD74 Ab specific for the protein's intracellular domain, but found no evidence of this fragment in the circulating (data not shown).

Figure 4.

Identification and characterization of circulating CD74. (A) Recombinant CD741-232 was electrophoresed as control, and serum circulating CD74 was immunoprecipitated from human samples. A single protein band was identified at 31 kDa for CD741-232 and at 25 kDa in serum samples. (B) Serum circulating CD74 (left panel) and recombinant CD741-232 (right panel) before and after deglycosylation. (C) Schematic diagram and MWs of CD74 illustrating its cytosolic, transmembrane (TM) and extracellular domains, as well as predicted extracellular glycosylation sites.[23] The measured circulating CD74 peptide (25 kDa) is shown alongside the known cytosolic peptide that is produced by regulated intramembrane proteolytic (RIP) cleavage. (D) Western blotting analysis of supernatants obtained over time from HSCs stimulated with or without IFN-γ (200 U/mL). (E) Detection of full-length (arrow) and truncated CD74 peptides in the hepatic microenvironment. Western blotting analysis of human whole-liver lysate, using anti-CD74 detection Ab.

We next investigated a potential cellular source of circulating CD74. HSCs express high levels of CD74 in response to IFN-γ,[26] which is a relevant inflammatory signal in autoimmune hepatitis.[27] CD74 also mediates MIF inflammatory signaling on HSCs.[28] Accordingly, we hypothesized that HSCs might be a source of circulating CD74. Stimulation of cultured immortalized human HSCs with IFN-γ produced a single anti-CD74 reactive band at 72 and 96 hours, similar in size to full-length, recombinant CD741-232 (Fig. 4D). This result indicates that HSCs can release the MIF receptor in response to the T-cell cytokine, IFN-γ. It is possible that full-length CD74 is released by HSCs through an alternate, exosomal mechanism as occurs in other soluble cytokine-receptor pathways.[22] Finally, a whole-liver lysate was generated from healthy human liver tissue and analyzed by western blotting. We detected both full-length (31 kDa) and lower MW, immunoreactive CD74 peptides in the hepatic microenvironment (Fig. 4E). Our findings do not exclude the possibility that other cell types might also contribute to the release of full-length and truncated CD74 into the circulation.

Serum CD74 Neutralizes MIF Bioactivity

The CD74 ectodomain (CD7473-232) binds to MIF with nM affinity.[14] To assess the structural basis of this interaction, we prepared a recombinant CD7473-232-Fc fusion protein to model circulating CD74 and tested its reactivity in an MIF peptide epitope scan. An overlapping series of 6- to 10-mer MIF peptides were synthesized, immobilized, and incubated with a recombinant CD7473-232-Fc fusion protein. Bound CD7473-232-Fc was detected by fluorescently labeled, anti-Fc IgG. The two highest scoring peptides corresponded to MIF79-88 (Fig. 5A), which is located within MIF's solvent accessible, second α2-helix.[29] We confirmed reactivity at this site by a competitive, MIF/CD7473-232-binding assay in which the minimally reactive peptide, MIF79-86, was tested for competition of biotinylated MIF binding to the immobilized CD74 ectodomain (CD7473-232; Fig. 5B). This synthetic peptide inhibited MIF interaction with CD74 with an half-maximal inhibitory concentration (IC50) of 2.24 μM.

Figure 5.

(A) MIF epitope scan for reactivity of the CD74 ectodomain versus overlapping MIF peptides. Peptide microarray fluorescence image analysis of a portion of the epitope scan showing high anti-IgG Cy5 reactivity with bound CD7473-232-Fc. (B) Confirmation of MIF79-86 peptide reactivity with CD7473-232 by competition for biotinylated MIF binding to plate-bound CD7473-232 (IC50, 2.24 μM). Control peptide was scrambled MIF79-86. (C-E) Circulating CD74 neutralizes MIF. Phosphorylated ERK1/2 (pERK1/2) and total ERK1/2 activity was measured by densitometry scanning of three gels and presented as a ratio as previously described.[13] Statistically significant differences were noted (*P < 0.05; **P < 0.01; ***P < 0.001). (C) Human skin fibroblasts (3.6 × 105 cells/well) were incubated with human MIF (2.7 nM), anti-MIF, human CD7473-232, and AIH, PBC, or control serum, as described in the Materials and Methods. (D) AIH sera were incubated with or without control IgG or preadsorbed with anti-CD74 Ab, followed by incubation with human MIF (2.7 nM). (E) PBC sera were incubated with or without control IgG or preadsorbed with anti-CD74 Ab, followed by incubation with human MIF (2.7 nM).

We next assessed the ability of circulating CD74 to neutralize MIF proinflammatory activity by measuring MIF-stimulated ERK1/2 phosphorylation in human primary skin fibroblasts cultured in the presence of AIH or control serum. MIF-dependent ERK1/2 phosphorylation was inhibited by AIH and PBC serum but not by control serum (Fig. 5C-E). Moreover, the MIF-neutralizing capacity of CD74 in both AIH and PBC sera was eliminated by preabsorption of serum with anti-CD74. Furthermore, there was a greater degree of MIF neutralization in PBC versus AIH serum (P = 0.02; Fig. 5C). These data support the conclusion that circulating, serum CD74 both binds and neutralizes MIF in serum of AIH and PBC patients.

To determine whether CD74 can circulate in a complex bound to MIF, we established a two-Ab sandwich ELISA to capture and quantify MIF-CD74 protein complexes in serum using wells precoated with anti-CD74 and detection with an anti-MIF mAb. MIF was detectable in CD74 captured by plate-bound anti-CD74 in sera from controls and both PBC and AIH patients. In controls (N = 15), the mean MIF-CD74 complex reading was 0.1 ± 0.2 optical density (OD; mean free MIF: 6.3 ng/mL; free CD74: 134 ng/mL; calculated CD74/MIF ratio: 19). In AIH (N = 16), the mean MIF-CD74 complex reading was 0.04 ± 0.03 OD (mean free MIF: 9.9 ng/mL; free CD74: 241.9 ng/mL; calculated CD74/MIF ratio: 24.4). In PBC (N = 16), the mean MIF-CD74 complex reading was 0.09 ± 0.14 OD (mean free MIF: 15.1 ng/mL; free CD74: 831 ng/mL; calculated CD74/MIF ratio: 55.2). Although there was a trend toward a lower MIF-CD74 complex level in AIH, we found no significant difference in serum MIF-CD74 complex levels between the three cohorts (P = NS). However, in the AIH cohort, MIF-CD74 complex levels correlated with free CD74 levels (r = 0.710; P = 0.002), suggesting that, in AIH, the level of circulating CD74 may drive the complex formation and subsequent neutralization of free MIF.


This is the first report of MIF expression and MIF gene associations in autoimmune liver disease. MIF was significantly elevated in AIH and PBC sera, compared to healthy controls, and both the expression of MIF and its receptor, CD74, was increased in liver tissues. Compared to patients with AIH, patients with PBC had a lower frequency of the high MIF expression, −794 CATT7 allele. We also report the identification and biochemical characterization of a circulating form of the MIF receptor ectodomain (CD74) and show that it binds MIF and neutralizes its signal transduction activity. Circulating serum levels of CD74 were markedly elevated in PBC, compared to AIH, and stimulation of human HSCs with IFN-γ resulted in release of CD74 in vitro. These results suggest distinct genetic and serum profiles of MIF and its circulating receptor, CD74, in AIH and PBC, which may explain in part their different immunopathologic profiles.

MIF activates both innate and adaptive immune systems, and in particular promotes T-helper cell type 1 (Th1) immunity, which is a key element of AIH pathophysiology.[27, 30] Expression of MIF in liver parenchymal cells further supports its active participation in the inflammatory cascade. Elevated serum MIF concentrations in patients with AIH and PBC are consistent with studies of other autoimmune disorders characterized by circulating autoAbs, such as systemic lupus erythematosus (SLE) and rheumatoid arthritis. In these disorders, circulating MIF also correlates with disease activity.[31, 32] We speculate that the elevated MIF concentrations in our patients represent a continuous, but often subclinical, autoimmune proinflammatory state. Though many patients with AIH were treated with low-dose glucocorticoids at the time of sample collections, MIF levels did not differ from those off steroids and elevated MIF levels were similar in PBC patients who were not managed with steroids. Although serum MIF levels were previously described in patients with CHB infection,[19] the identification of serum MIF and CD74 in AIH patients is novel and can add to the understanding of its pathophysiology. The negative correlation between the calculated serum ratio of CD74/MIF and serum ALT in AIH patients who experienced a relapse, as measured by serum liver tests (Fig. 3D), suggests that the concentration of MIF, relative to its neutralizing receptor, may be a potential marker of immunogenic inflammation in liver in AIH. This was not found in the PBC cohort, possibly becaue of the lack of a relationship between serum enzymes and the disease status in that condition. Given the need for markers to predict recrudescence of inflammation in the management of chronic AIH, this initial description of MIF and CD74 in AIH suggests that MIF, CD74, or MIF-CD74 complexes may be candidate immunologic biomarkers of disease activity. Further longitudinal studies will be needed.

The pathogenesis of autoimmune liver diseases is believed to be dependent on genetic predisposition compounded by environmental stimuli.[17, 18] Previous reports have identified associations with human leukocyte antigen (HLA) alleles, particularly DRB1*0301 and DRB1*0401 in AIH,[33, 34] and DR8 and DQB1 in PBC,[35, 36] and a number of non-HLA genetic polymorphisms also have been identified, including cytotoxic T lymphocyte antigen 4 in AIH and PBC,[37, 38] and the interleukin-12A locus in PBC.[36] Here, we present evidence of genetic associations with the −794 CATT5-8 microsatellite polymorphism in the MIF promoter, highlighting a genetic-phenotypic relationship between known MIF functional alleles and the clinical profile of AIH versus PBC. Importantly, these alleles are prevalent in the population,[20] and further studies should investigate discrete clinical phenotypes to understand whether these MIF promoter alleles predict clinical progression of disease, as they do in SLE.[9] The OR of 1.34 (P = 0.0331) for the low-MIF-expression CATT5 allele in PBC versus controls is very similar to the recently described genetic associations with HLA and non-HLA polymorphisms.[36]

The higher frequency of the −794 CATT7 allele (P = 0.0013) and CATT77 genotype (P = 0.0074) in AIH, compared to PBC, patients is consistent with the Th1-induced and IFN-γ-mediated proinflammatory phenotype of AIH. AIH also is more responsive to immunosuppression. Because our genetic data indicate that AIH, but not PBC, is linked to a high-expression MIF allele, elevated MIF expression in AIH may be driven more directly by the MIF locus, whereas in PBC, increased MIF levels may reflect the secondary stimulatory action of other inflammatory or immunologic pathways. Finally, the unique MIF genetic profile of AIH patients is worth comparing with SLE because each are antinuclear Ab (ANA)-positive diseases. Indeed, the −794 CATT77 genotype frequencies were much higher in AIH, compared to recent studies in SLE.[9] This may reflect different host responses to environmental triggers for these diseases or the greater clinical heterogeneity of SLE. However, caution is needed when interpreting the CATT77 genotype data because of the relatively small number of studied patients.

CD74 is a type II transmembrane receptor and initiates MIF signal transduction by recruitment of CD44 or in complex with chemokine receptors.[13, 14, 39] CD74 is expressed constitutively on the surface of antigen-presenting cells, as well as epithelial and different stromal cells. We demonstrated CD74 staining within the liver of patients with AIH and PBC, predominantly in the hepatic sinusoids, but with increased involvement of hepatocytes in severe AIH, suggesting that CD74 may be involved in the autoimmune response to hepatic antigen.

Our identification of a bioactive circulating form of the MIF receptor adds new insight into the modulation of MIF's cytokine properties. The 25-kDa circulating protein is 6 kDa smaller than the full-length, transmembrane CD74 protein, but remains modified by post-translational glycosylation present in the extracellular domain.[23] Extensive literature exists describing the role of the invariant chain, which is the intracellular expressed form of CD74, in antigen processing.[40-42] A report by Eynon et al. described the role of a secreted 25-kDa invariant chain fragment in the inhibition of T-cell activation.[42] Although the circulating CD74 fragment we identified has the same MW, it is unclear whether the mechanism of secretion and the biologic properties are similar. Our findings indicated that the circulating CD74 peptide is truncated, compared to full-length CD74, and we suggest that the difference in MW may be accounted for by regulated intramembrane proteolysis, which has been reported to generate a 6-kDa intracytosolic peptide fragment that up-regulates nuclear factor kappa B transcription.[24]

Maubach et al. previously reported on up-regulation of CD74 in IFN-γ-stimulated HSCs.[26] Here, we demonstrated the release of a nontruncated CD74 peptide into the supernatants of human HSCs after stimulation with IFN-γ, which is a product of activated T cells. These data support the known role of T cells in autoimmune hepatitis[27] and suggest a model whereby IFN-γ drives the production of circulating CD74 from HSCs in vivo. The proteolytic mechanism responsible for creating the single, 25-kDa CD74 peptide that circulates in serum remains to be elucidated, and we identified several smaller CD74 immunoreactive bands in whole-liver lysate analysis (Fig. 4E). Therefore, it is possible that additional cell types in the liver can release CD74 in AIH and PBC.

We also showed that circulating CD74 is bioactive and neutralizes MIF signal transduction activity, suggesting that circulating CD74 modulates MIF activity in vivo. MIF epitope scanning studies indicate that the CD74 ectodomain binds MIF in its α-2 helical domain. How this domain interacts with MIF's N-terminal region, which also has been implicated in receptor binding by site-mediated mutagenesis studies,[43] remains to be elucidated. That circulating CD74 binds MIF in vivo was confirmed by ELISA detection of circulating MIF-CD74 complexes. The existence of this complex is expected, given that the KD for MIF with the CD74 ectodomain of 1.4 × 10−9 M.[44]

Finally, we found differential levels of circulating CD74 between AIH and PBC with higher median levels in PBC, compared to AIH. Because CD74 neutralizes MIF function, the calculated molar ratio of circulating CD74/MIF may be a clinical indicator of the proinflammatory activity of MIF. The ratio of CD74/MIF clearly distinguished AIH from PBC, where patients with AIH have a much lower CD74/MIF ratio, compared to patients with PBC. This finding may have additional clinical significance, because PBC is a less-active inflammatory disease than AIH. Whether CD74 circulates in other autoimmune inflammatory diseases, and if the CD74/MIF ratio may distinguish clinical subtypes of disease or predict disease progression, remains to be studied. Further basic and clinical investigation is needed to evaluate the hypothesis that CD74 functions to regulate MIF inflammatory activity on a cellular and systemic level.

In conclusion, our findings suggest the existence of differences in both serum MIF and CD74 levels, as well as in functional MIF polymorphism distributions in patients with AIH and PBC that reflect known phenotypic differences between these two disorders. Furthermore, we have identified an endogenous modulator of MIF activity in the form of circulating CD74. Last, the unique relationship between CD74 and MIF in AIH versus PBC suggests that these proteins might be suitable candidates for evaluation as biomarkers of disease activity.


The authors gratefully acknowledge the assistance of Amos Brooks (Yale Research Histology) with the immunohistochemical staining of human liver tissue and Dr. M. Eric Gershwin for assistance in obtaining Dr. Kaplan's PBC sample repository. The authors also thank the patients of Yale Liver Clinics for their participation in clinical research. Finally, the authors dedicate this work to the memory of our coauthor and colleague, the late Dr. Marshall M. Kaplan.