Keratin 19 demonstration of canal of hering loss in primary biliary cirrhosis: “Minimal Change PBC”?


  • Fahad M. Khan,

    1. Department of Medicine, Division of Digestive Diseases, Beth Israel Medical Center of Albert Einstein College of Medicine, New York, NY
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Arathi Rajendra Komarla,

    1. Department of Medicine, Division of Digestive Diseases, Beth Israel Medical Center of Albert Einstein College of Medicine, New York, NY
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Paulo G. Mendoza,

    1. Department of Pathology, Beth Israel Medical Center of Albert Einstein College of Medicine, New York, NY
    Search for more papers by this author
  • Henry C. Bodenheimer Jr,

    1. Department of Medicine, Division of Digestive Diseases, Beth Israel Medical Center of Albert Einstein College of Medicine, New York, NY
    Search for more papers by this author
  • Neil D. Theise

    Corresponding author
    1. Department of Medicine, Division of Digestive Diseases, Beth Israel Medical Center of Albert Einstein College of Medicine, New York, NY
    2. Department of Pathology, Beth Israel Medical Center of Albert Einstein College of Medicine, New York, NY
    • Division of Digestive Diseases, Beth Israel Medical Center, First Avenue at 16th St., New York NY 10003
    Search for more papers by this author
    • fax: 212-420-4373

  • Potential conflict of interest: Nothing to report.


Liver biopsy is important for diagnosing primary biliary cirrhosis (PBC). Prior investigations suggest that immunostaining for biliary keratin 19 (K19) may show the earliest changes suspicious for PBC, namely, loss of the canals of Hering (CoH). We aimed to study the clinical outcomes of patients whose biopsy specimens appeared histologically near normal or with minimal inflammatory changes, but in which K19 staining revealed widespread periportal CoH loss, a finding we termed “minimal change PBC.” Ten patients were identified prospectively as having nearly normal or mildly inflamed biopsy specimens without diagnostic or suggestive histologic features of PBC, but with near complete CoH loss; six had available follow-up clinical data, one had follow-up biopsy. Controls for clinical and/or K19 analysis included six normal livers and biopsy specimens from 10 patients with confirmed early PBC, 10 with early stage chronic hepatitis C (CHC), and nine with resolving, self-limited hepatitis (RSLH). Staining for K19 in normal controls, livers with “minimal change” PBC, CHC, and RSLH showed 9.2 ± 6.0, 0.44 ± 0.37 (P < 0.0001), 5.7 ± 4.6 (n.s.), 4.1 ± 2.1 (P < 0.02) CoH per portal tract, respectively. Patients with available clinical follow up, compared to patients with diagnostic early-stage PBC biopsies, showed identical treatment responses to ursodeoxycholic acid, similar rates and types of nonhepatic autoimmune diseases, and/or subsequent development of autoimmune hepatitis overlap syndrome. Conclusion: We suggest that CoH loss demonstrated by K19 immunostaining is an early feature in PBC. Clinical findings in the years following biopsy, including response to ursodeoxycholic acid, show identical changes to patients with biopsy confirmed PBC. We suggest that this “minimal change” feature may support a clinical diagnosis of PBC even in the absence of characteristic, granulomatous, duct destructive lesions. (HEPATOLOGY 2013)

Primary biliary cirrhosis (PBC) is a relatively rare autoimmune disease primarily affecting women with a prevalence of ∼65.4 per 100,000 persons versus 12.1 for men. The usual age of onset is between 30 and 65 years of age.1, 2 The diagnosis of PBC is often made when the patient is asymptomatic, but has abnormal serum liver test results. These abnormalities are mainly elevated serum alkaline phosphatase (AP) and/or a positive antimitochondrial antibody (AMA), the latter being positive in nearly 95% of such patients. When PBC is suspected, liver biopsies are an important confirmatory and staging tool, and are particularly important diagnostically when autoantibodies are negative.3

Histologically, PBC is classically characterized by nonsuppurative, often granulomatous destruction of bile ducts followed by ductular reaction, progressive fibrosis, and cirrhosis.3 The destructive process preferentially involves the most proximal portion of the biliary tree, the smaller bile ducts.4 We previously suggested that the disease process of PBC involves loss of the very smallest, most proximal branches of the biliary tree, namely, the canals of Hering (CoH).4 The CoH connect hepatocellular bile canaliculi to interlobular bile ducts and are also considered a facultative stem cell niche of the liver.5-7 In normal liver tissue, CoH are unapparent by routine histochemical stains.8 They are made evident by immunostaining for biliary markers such as keratin 19 (K19) and epithelial cell adhesion molecule (EpCAM), which are positive in all cholangiocytes of the entire biliary tree.5, 7 The CoH are strings of small, K19-positive cholangiocytes that extend from variable locations in acinus zone 1 or 2 to where they link to ductules near or at the limiting plate.5 It is rare to see completely, longitudinally sampled CoH even with immunostains; most often the CoH appear in cross-section, as isolated cells or short strings of cells arrayed around the portal tract, but within the periportal parenchyma (Fig. 1).

Figure 1.

Normal portal tract with immunostain for K19. Numerous K19-positive CoH cross-sections are arrayed in the periportal region surrounding this normal portal tract (arrows) (DAB, hematoxylin counterstain, original magnification 10×).

In our prior study, liver biopsy specimens from patients with PBC were immunostained for K19.4 The number of CoH per portal tract were decreased in portal tracts containing injured or absent bile ducts, but were most strikingly absent (<1 CoH profile/portal tract) around normal-appearing portal tracts without any histologic features of bile duct injury or loss. These findings suggested that CoH loss might be a particularly early event in the development of PBC, perhaps even preceding overt bile duct injury. However, data regarding the clinical outcomes of these patients was lacking.

In our present study, we aimed to determine the clinical outcomes of patients with a clinical picture suspicious for PBC, but whose liver biopsy specimens contained no histologic evidence of bile duct destruction or loss typical of PBC by routine stains, were otherwise histologically normal or with minimal necroinflammatory changes, but in whom K19 staining revealed widespread CoH loss. We suggest that this finding be termed “minimal change PBC” and hypothesize that the loss of CoH alone in such patients may represent an immunohistochemical test for earlier biopsy confirmation of PBC.


AASLD, American Association for the Study of Liver Diseases; AMA, antimitochondrial antibody; ANA, antinuclear antibody; AP, alkaline phosphatase; CHC, chronic hepatitis C; CoH, canals of Hering; C/P ratio, CoH to portal tract ratio; DAB, diaminobenzidine; EpCAM, epithelial cell adhesion molecule; GGT, gamma-glutamyl transpeptidase; K19, keratin 19; K7, keratin 7; PBC, primary biliary cirrhosis; UDCA, ursodeoxycholic acid.

Materials and Methods

Case Selection.

We prospectively identified all liver biopsy specimens sent to Beth Israel Medical Center, Division of Digestive Diseases, between 2003 and 2008 for evaluation for the suspected clinical diagnosis of PBC that showed features of “minimal change PBC” as follows: patients clinically suspected as having PBC underwent liver biopsy, the specimens from which were processed routinely with three hematoxylin and eosin (H&E)-stained slides and four slides stained with Masson trichrome, Prussian blue, periodic acid-Schiff after diastase digestion and silver stain for reticulin; if a biopsy specimen was defined as nondiagnostic, without overt duct destruction or duct loss, and characterized by, at most, focal, scant, portal mononuclear infiltrates or rare foci of necroinflammation. When the clinical impression was PBC, but typical features were absent, routine practice of the study pathologist (N.D.T.) has been to order six H&E-stained additional slides and one slide immunostained for biliary marker K19 (techniques described below); if these additional slides still showed no PBC-typical features, then CoH loss as determined with K19 stain (see below) was recorded in the report (in searchable format so that the cases could be retrieved at a later date).

For this report of these patients with “minimal change PBC,” patient charts were retrieved from archives, reviewed, and updated clinical data were then collected, when possible. In addition, 10 case control patients were selected who had overt histologic features of early (Scheuer stage 1 or 2) PBC on liver biopsy. Histologic and immunohistochemical assessments of CoH loss were also performed on two groups of control patient biopsy specimens. The first group was comprised of specimens from patients with chronic hepatitis C (CHC). These were prospectively selected by the study pathologist, i.e., the next CHC biopsy specimens received after initiation of review by our study group (n = 11). The second case control group comprised biopsy specimens from retrospectively identified patients with features of resolving, self-limited hepatitis (RSLH), namely, clustered, pigment-laden macrophages in the absence of active acute or chronic hepatitis and with mildly elevated, but declining serum transaminases (n = 9).9 Biopsy specimens from six normal control patients were identified with no known liver disease.

Clinical information gathered included an evaluation of symptoms and signs, treatment given, review of clinical outcomes (including eventual diagnosis), and serologic test results (serum autoantibodies, liver enzymes, immunoglobulins). Additional information gathered included a detailed clinical history consisting of age, gender, weight, past medical history, social history, family history, medication history, and allergy information. Further laboratory and radiologic data gathered included thyroid function tests, vitamin D levels, viral hepatitis panels, and results of abdominal sonograms, computed tomography of the abdomen and pelvis, and magnetic resonance imaging (MRI) of the abdomen. Pathologic assessment of minimal change biopsy specimens included assessment of the presence/absence of granulomas, pigment-laden macrophages, and ductular reactions. Portal inflammation, interface and lobular hepatitis, and confluent necrosis were assessed according to the Hepatitis Activity Index (HAI) of the Ishak staging scheme developed for chronic hepatitis.9

Immunohistochemical Stains and Quantification of CoH.

Immunostaining was performed on study group specimens, normal controls, and case controls with CHC or RSLH according to standard techniques.5, 7 In brief, specimens were fixed in 10% buffered formalin and paraffin-embedded. Sections were cut at 4 μm. After deparaffinization and antigen retrieval in high PH, tissues were stained for K19 (monoclonal antibody RCK108; Dako, Carpinteria CA; dilution 1:100), K7 (monoclonal antibody OV-TL 12/30, Dako; dilution 1:100), and for EpCAM (monoclonal antibody VU-1D9, Leica Microsystems, Buffalo Grove, IL; prediluted by manufacturer). Colorization of the stain was accomplished with diaminobenzidine (DAB) and counterstaining was performed with Mayer's hematoxylin.

The total number of portal tracts and CoH were counted in all biopsy specimens. A single K19-positive cell or cell cluster or a single K19-positive linear string was counted as one unit. The CoH to portal tract (C/P) ratio was calculated for all normal tissues, PBC specimens, and CHC specimens. Stains for K7 and EpCAM were assessed semiquantitatively: 0, absent; 1+, occupying <10% portal tract circumferences; 2+, 11%-25%; 3+, 26%-50%; 4+, 51%-100%.

Data Analysis.

All data were placed in a database with names of patients and other identifying information removed for confidentiality to the extent permitted by law. Institutional Review Board approval was obtained prior to study commencement. Statistical analysis of comparisons between laboratory data among both subject and control patients was performed using unpaired t tests.


Clinical and Pathologic Data.

Pathology findings in the 10 biopsy specimens from all prospectively identified minimal change cases are shown in Table 1. A retrospective chart review was then conducted of the 10 prospectively identified subject patients and six were identified who had retrievable clinical data. All 10 PBC control patients had retrievable clinical data. The average length of follow-up was 2 years. Baseline characteristics and clinical data on the subject and PBC control patients are summarized in Tables 2-5, respectively. There were no statistically significant differences between baseline characteristics or laboratory values before and after treatment, among both sets of patients using paired t-test analysis. In addition, total bilirubin levels (not presented in tables) among both sets of patients were within normal limits with no statistically significant differences before or after treatment. No exposures to known hepatotoxins (prescription or non-prescription) were identified in the patients upon chart review. Study patients had an age distribution of 52 ± 7 years; PBC control patients had an age distribution of 52 ± 12. All suspected or diagnosed PBC patients were female. Clinical data for the CHC patients showed a male:female gender distribution of 5:6 and age distribution of 48 ± 9 years. These age differences are not statistically significant. Patient 1 presented initially with symptoms of fatigue and pruritus. On laboratory evaluation the patient's AP and gamma-glutamyl transpeptidase (GGT) levels were found to be elevated for at least 1 year. The patient also had a positive AMA, as well as mildly elevated aminotransferases. Sonographic evaluation of the liver did not reveal any abnormalities. Due to ongoing suspicion that the patient had PBC, a liver biopsy was performed that was nondiagnostic for PBC; however, immunostain for K19 highlighted focal bile duct loss and widespread loss of CoH (Table 1). The patient was subsequently started on 15 mg/kg daily dose of ursodeoxycholic acid (UDCA). During the follow-up period of 4 years, the patient's AP, GGT, and aminotransferase levels normalized. The patient also responded symptomatically and reported resolution of complaints of pruritus and fatigue following initiation of treatment. There were no follow-up liver biopsies performed. Currently, the patient is still being treated and continues to be asymptomatic, with normal laboratory findings.

Table 1. Baseline Biopsy Specimen Characteristics of “Minimal Change PBC” Patients
  Ishak Hepatitis Activity Index8      
Study Pt. #GranulomasPortal Mononuclear Inflammation (possible 0 to 4)Interface Hepatitis (possible 0 to 4)Lobular Hepatitis (possible 0 to 4)Confluent Necrosis (possible 0 to 6)Pigment Laden Mø'sBile Duct to Portal Tract RatioCoH to Portal Tract RatioDRsCholate StasisK7
  1. Mø's: Macrophages; CoH: canals of Hering; DRs: ductular reactions; K7: Periportal keratin 7 hepatocyte expression (semiquantitative grading: 0, absent; 1+, occupying <10% portal tract circumferences); PT: portal tract; bx: biopsy.

  2. *Elevated due to extensive tangential and longitudinal sampling without portal trends in the specimen.

101 (scant)000Focal1.00000
201 (scant)00001.90001+
303 (single PT)00002.00000
4 (bx 1)01 (scant)01 (rare)0013*0001+
4 (bx 2)Scattered parenchymal, noncaseating2222Focal0.61.0001+
5Single, portal, noncaseating3 (two PT)1 (rare)1 (rare)002.21.3001+
603 (single PT)00009*1.1000
703 (single PT)00001.11.2000
801 (scant)1 (rare)0002.10000
901 (scant)00001.80000
1001 (scant)00001.21.0000
Table 2. Baseline Clinical Characteristics of Subject Patients
Pt. #GenderAgeRacePMHxFHxETOHSmoking
  1. PMHx, past medical history of an autoimmune or liver disease; FHx, family history of any autoimmune or liver disease.

1F42HispanicGrave's diseaseNoneOccasionalNo
3F54HispanicHypothyroidismSister-PBCNo1 pack year; quit 8 years ago
6F54CaucasianDiabetes mellitus, hypothyroidism, ITPNoneOccasionalNo
Table 3. Baseline Characteristics of PBC Control Patients
Pt. #GenderAgeRacePMHxFHxETOHSmoking
  1. PMHx, past medical history of an autoimmune or liver disease; FHx: family history of any autoimmune or liver disease.

11F41HispanicAutoimmune thyroiditisNoneNoNo
16F53HispanicLupus (SLE), Raynaud'sNoneNoNo
Table 4. Clinical Follow-up of Subject Patients
 Tests at DiagnosisPost-UDCA Treatment
Pt. #ALT/AST (U/L)Alk Phos (U/L)AutoAbsGGT (U/L)IgM/IgG mg/dlALT/AST (U/L)Alk Phos (U/L)AutoAbsGGT (U/L)IgM/IgG mg/dL
  1. N/A, not available; AutoAbs, autoantibodies; ANA and ASMA are negative when not listed.

Table 5. Clinical Follow-up of PBC Control Patients
 Tests at DiagnosisPost-UDCA Treatment
Pt. #ALT/AST (U/L)Alk Phos (U/L)AutoAbsGGT (U/L)IgM/IgG mg/dlALT/AST (U/L)Alk Phos (U/L)AutoAbsGGT (U/L)IgM/IgG mg/dL
  1. N/A, not available; AutoAbs, autoantibodies. ANA and ASMA are negative when not listed.


Patient 2 also initially complained of pruritus. Laboratory analysis revealed markedly elevated AP levels, elevated GGT levels, elevated IgM levels, and minimally elevated aminotransferases. The patient tested negative for AMA. These abnormal laboratory results persisted for ∼6 years and sonographic evaluation of the liver revealed possible fatty liver disease or slight chronic/diffuse disease with no evidence of cholelithiasis. A liver biopsy was performed that was nondiagnostic for PBC; however, immunostain for K19 showed no duct loss, but widespread loss of CoH (Table 1). The patient was then started on 15 mg/kg treatment of daily UDCA and reported resolution of her pruritus. The patient was followed after treatment for ∼1 year, during which her alkaline phosphate levels decreased by ∼20% to around 240 U/L but never normalized, GGT levels were reduced by 50% to around 32 U/L but never normalized, and aminotransferases decreased by 50% and did normalize. Immunoglobulin M (IgM) levels remained elevated and the AMA remained negative.

Patient 3 initially complained of fatigue, pruritus, and symptoms of dry eyes. Laboratory evaluation revealed that the patient's AP and aminotransferase levels were elevated. The patient was found to be AMA-negative. A hepatic sonogram and MRI of the abdomen did not reveal any pathology. A liver biopsy was performed, which was nondiagnostic for PBC; however, immunostain for K19 highlighted focal bile duct loss and widespread loss of CoH (Table 1). The patient was started on 15 mg/kg of daily UDCA. After treatment, the patient's AP and aminotransferase levels normalized and remained normal over a year and half of follow-up. The patient's symptoms also subjectively improved.

Patient 4 initially had symptoms of fatigue and pruritus. Laboratory evaluation revealed elevated AP, negative AMA, positive antinuclear antibody (ANA), and elevated aminotransferase levels. Both a sonogram and MRI of the liver did not reveal any radiographic evidence of hepatic pathology. The first liver biopsy was performed and it was nondiagnostic for PBC; however, immunostain for K19 highlighted no bile duct loss and widespread loss of CoH (Table 1). The patient was initially started on treatment with 15 mg/kg of daily UDCA. However, after treatment of UDCA alone the patient's laboratory abnormalities initially improved but then started to increase again after 2 years. A second biopsy was done 2 years later which was compatible with an autoimmune “overlap” syndrome inclusive of features strongly suggestive of PBC, namely, duct loss, focal ductular reactions, and parenchymal noncaseating granulomas (Table 1). The patient was then continued on UDCA, started on a prednisone taper, and azathioprine. With this treatment, the patient's laboratory abnormalities normalized within 3 months and remained stable during the 1-year follow-up period. The patient's symptoms also improved.

Patient 5 initially complained of pruritus and fatigue. Biochemical properties were significant for a positive AMA, elevated AP, and elevated aminotransferases. Hepatic ultrasound and MRI did not reveal any abnormalities. A liver biopsy was performed which was nondiagnostic for PBC; although a single, portal, noncaseating granuloma was present, this did not contain an injured bile duct and so could not be classed as a diagnostic, duct-destructive lesion. Immunostain for K19 highlighted no bile duct loss and widespread loss of CoH (Table 1). The patient was started on 15 mg/kg of daily UDCA. After a year of follow-up the patient became AMA-negative. The aminotransferase levels were also reduced significantly but did not completely normalize (ALT was reduced by 65% to around 48 U/L, and AST reduced by 50% to around 46 U/L). AP levels were also reduced by 25% but remained elevated around 172 U/L. The patient also reported improvement of symptoms with treatment over the year and half of follow-up.

Patient 6 initially had no symptoms but had markedly elevated AP, GGT, and aminotransferase levels. The patient was AMA-negative but ANA-positive. Ultrasound of the liver revealed mildly heterogeneous echogenic hepatic parenchyma compatible with mild steatosis or possible chronic diffuse liver disease. Liver biopsy was subsequently performed which showed no significant bile duct loss or steatosis (Table 1). K19 immunostaining revealed an almost complete absence of CoH. The patient was started on 15 mg/kg of daily UDCA for a presumed diagnosis of PBC, but was lost to follow-up and no subsequent data are available.

CoH Quantification and Other Immunostaining.

Liver biopsy specimens were deemed adequate by length and number of portal tracts sampled in the minimal change group (2.9 ± 0.8 cm, 23 ± 8 portal tracts per specimen) and comparison PBC (3.0 ± 1.1 cm, 25 ± 9 portal tracts per specimen), CHC controls (2.5 ± 0.7 cm, 20 ± 7 portal tracts per specimen), and RSLH (2.8 ± 0.9 cm, 26 ± 10 portal tracts per specimen).

In the minimal change cases only rare portal tracts were without bile ducts, although the average of bile ducts per portal tract (Table 1) is often less than the reported standard of many normal portal tracts having two bile duct profiles. C/P ratios demonstrated profound loss of K19-positive CoH in the 10 minimal change cases as compared to both normal controls and CHC disease controls, as summarized in Fig. 2. The minimal change PBC subjects showed 0.41 ± 0.57 CoH per portal tract (range: 0 to 3; median: 0) compared to normal controls 9.2 ± 6.0 CoH per portal tract (range: 3.1 to 16.2; median: 9; P < 0.0001). Early stage PBC control cases showed 3.3 ± 1.4 CoH per portal tract (range: 1.2 to 7.2; median 2.9; not statistically different than either normal controls or minimal change PBC cases). CHC biopsy specimens showed C/P ratios of 5.7 ± 4.6 (range: 2.0 to 9.0; median: median 5.6) (P < 0.0002 compared to suspected PBC subject group; no significant difference compared to normal). RSLH specimens showed C/P ratios of 4.1 ± 2.1 (range: 1.8 to 8.1; median 3.8). Figure 3A shows the H&E of a sample subject patient suspected of having PBC but with normal-looking portal tracts. Figure 3B for the same patient with K19 staining highlights CoH loss around a portal tract.

Figure 2.

Quantitation of CoH per portal tract. There is near complete loss of CoH in “minimal change” PBC biopsy specimens compared to normal control livers and to disease controls of early stage hepatitis C and RSLH.

Figure 3.

Minimal change, near-normal biopsy specimen from patient with clinically suspected primary biliary cirrhosis. (A) Normal-appearing portal tract with intact, uninjured bile ducts easily apparent. Mononuclear infiltration is scant at best, and neither granulomatous duct destructive lesions nor ductular reactions are present (H&E, original magnification 20×). (B) Immunostain for K19 confirms presence of small interlobular bile ducts without injury, but only one CoH remains (arrow) (DAB, hematoxylin counterstain, original magnification 10×).

K7 staining highlighted the same biliary structures as did K19 staining, but also, in some specimens (Table 1), showed very focal periportal hepatocyte positivity which is common (although usually more widespread) in typical PBC biopsy specimens. EpCAM showed identical patterns of staining to K19, differing in subcellular localization as expected (EpCAM is membranous, K19 is cytoplasmic), whereas no EpCAM-positive hepatocytes were seen in any normal, minimal change PBC, or CHC specimens (data not shown).


In current American Association for the Study of Liver Diseases; (AASLD) Practice Guidelines3 diagnosis of PBC can be made with compatible biochemical tests (esp. elevated AP), positive AMA (found in ∼95% of patients), and/or a compatible histological liver biopsy specimen. However, no standard case definition has been universally accepted.10 We have previously suggested that marked CoH loss (recognized by immunostaining for K19 or other cholangiocyte marker) is an earliest change in PBC,4 a finding confirmed by others,11 although no studies have assessed clinical outcomes of patients with suspected PBC, who have minimally injured liver biopsies by routine stain, but marked CoH loss with immunostaining, a finding we term “minimal change PBC.”

We prospectively identified patients suspected to have PBC, but who did not meet definitive, nonbiopsy criteria for diagnosis. Five of six study patients had detectable serum autoantibodies, two of whom were AMA-positive prior to treatment (becoming AMA-negative with treatment) and four of six had either antismooth muscle antibody (ASMA) or ANA, serologic findings typical for AMA-negative PBC.12 The four AMA-negative patients (67%) is a higher rate than has been previously reported, which was 1/10 (10%) in our control group, in keeping with other published reports.13, 14 The difference may relate to our small cohort size and not have meaning; however, it could indicate that PBC identifiable only by CoH loss is such an early disease stage that perhaps AMA have not yet become detectable by standard assays.

To exclude the possibility that CoH loss is a nonspecific finding, unrelated to PBC, we compared the study group to two disease control specimen sets, both of which had preserved CoH. Compared to CHC, we confirm that CoH loss is not a nonspecific reaction to a chronic, portal, and peri-portal inflammation. In order to exclude the possibility that the study patients were simply recovering from an acute injury that would have spontaneously improved even without treatment, we assessed biopsy specimens from patients with chronic elevated serum liver tests for greater than 6 months, but without clinical data to indicate inciting infections or toxins. Biopsy specimens showed clustered macrophages indicating recent hepatic activity, but ongoing hepatitis was absent. All of these patients eventually normalized their liver test profiles (within 6 to 14 months). CoH appearance or number in this group also remained indistinguishable from normal.

All six subject patients who had CoH loss and clinical data available for review were started on treatment with UDCA. Posttreatment clinical follow-up was available for five of these six patients, all of whom showed normalization of abnormal serum liver tests and diminished pruritis, results indistinguishable from our 10 PBC control patients. These results, for both subject and PBC control groups, are similar to what is expected to be the known biochemical response rate to treatment for PBC.1-3, 12

One patient each, in study and control groups, had a repeat biopsy showing “autoimmune hepatitis overlap syndrome.” Both patients were initially diagnosed with PBC based on blood test results and were treated with a favorable response for at least a year. Then, with newly rising transaminases, they underwent repeat biopsy and were found to have emergent autoimmune hepatitis overlapping with the previously assessed PBC. A recent study has shown that 2.4% of PBC patients develop an acute autoimmune hepatitis on top of their PBC.17 Nonetheless, the development of overlap syndrome after the initial biopsy finding as “minimal change” further supports that the patient already had, at the time of first biopsy, an evolving autoimmune hepatopathy. Other concomitant autoimmune diseases in our study patients further support them as having PBC: one had a sister with PBC, three had thyroid dysfunction which is commonly associated with PBC, often predating the liver diagnosis18 and most strongly associated with AMA-negative PBC.19

Where did the CoH go? Prior research suggests two hypotheses. The first is that they were destroyed by immune attack, a possibility supported by our prior finding that they, like the bile ducts in PBC, show de novo human leukocyte antigen DR (HLA-DR) expression and may thus be targets of immune attack.4 An alternate hypothesis is that the cells lining the CoH are not destroyed, but “disappear” by undergoing hepatocellular differentiation, as suggested by bromodeoxyuridine (BRDU), label-retaining cell assays in a murine model of acetaminophen toxicity.20 In that study, stem/progenitor cells within the CoH appeared to differentiate directly into hepatocytes without cell division. Other reports indicate that K19-positive stem/progenitor cells in the CoH can produce K19 negative/EpCAM-positive hepatocytes.7, 21 For this reason we immunostained our specimens for EpCAM as well; however, no EpCAM staining was seen in any case. This question thus remains unanswered.

We conclude that diffuse CoH loss demonstrated by immunostaining for K19 can be considered a “minimal change” diagnostic biopsy feature of PBC in specimens without overt histological features classically associated with PBC. This conclusion is supported by the clinical findings in such patients of concomitant, nonbiliary autoimmune diseases; subsequent, biopsy-confirmed development of overt immune-mediated liver disease; and PBC-like response to UDCA treatment including loss of autoantibodies, trend toward normalization of serum liver tests, and remission of pruritus and/or fatigue. We thus propose that K19 staining be further investigated as a diagnostically useful test to be applied to biopsy specimens from patients clinically suspected to have PBC when classical histologic features are absent.