Keratin variants associate with progression of fibrosis during chronic hepatitis C infection


  • Pavel Strnad,

    1. Department of Medicine, Palo Alto VA Medical Center, Palo Alto, CA; and Stanford University Digestive Disease Center
    Search for more papers by this author
  • Tim C. Lienau,

    1. Department of Medicine, Palo Alto VA Medical Center, Palo Alto, CA; and Stanford University Digestive Disease Center
    2. Medizinische Klinik mit Schwerpunkt Hepatologie und Gastroenterologie, Charite, Universitätsmedizin Berlin, Campus Virchow, Berlin, Germany
    Search for more papers by this author
  • Guo-Zhong Tao,

    1. Department of Medicine, Palo Alto VA Medical Center, Palo Alto, CA; and Stanford University Digestive Disease Center
    Search for more papers by this author
  • Laura C. Lazzeroni,

    1. Department of Health Research and Policy —Biostatistics, Stanford University, Stanford, CA
    Search for more papers by this author
  • Felix Stickel,

    1. Institute for Clinical Pharmacology, University of Berne, Switzerland
    Search for more papers by this author
  • Detlef Schuppan,

    1. Department of Medicine I, University of Erlangen-Nürnberg, Erlangen, Germany
    2. Division of Gastroenterology, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA
    Search for more papers by this author
  • M. Bishr Omary

    Corresponding author
    1. Department of Medicine, Palo Alto VA Medical Center, Palo Alto, CA; and Stanford University Digestive Disease Center
    • Palo Alto VA Medical Center, Mail code 154J, 3801 Miranda Avenue, Palo Alto, CA 94304
    Search for more papers by this author
    • fax: 650-852-3259.

  • Potential conflict of interest: Nothing to report.


Keratins 8 and 18 (K8/K18) protect the liver from various forms of injury. Studies of liver explants from a large cohort of U.S. patients showed that K8/K18 mutations confer a risk to developing end-stage liver diseases, though which diseases are preferentially involved is unknown. We tested the hypothesis that K8/K18 variants are associated with chronic hepatitis C (CHC) and that their presence correlates with progression of fibrosis. Genomic DNA was isolated from peripheral blood of a well-characterized German cohort of 329 patients with CHC infection. Exonic regions were PCR-amplified and analyzed using denaturing high-performance liquid chromatography and DNA sequencing. Our findings showed: (1) amino acid altering keratin heterozygous variants in 24 of 329 CHC patients (7.3%) and non-coding heterozygous variants in 26 patients (7.8%), and (2) 3 new exonic K8 variants (T26R/G55A/A359T); 6 novel non-coding variants and one K18 coding variant (K18 S230T; 2 patients). The most common variants were K8 R341H (10 patients), K8 G62C (6 patients) and K8 I63V (4 patients). A novel and exclusive association of an intronic KRT8 IVS7+10delC deletion in all 10 patients with K8 R341H was observed. Notably, there was a significant association of exonic, but not of intronic K8 variants with increased fibrosis. In conclusion, previously described and novel K8 variants are present in a German population and collectively associate with progression of fibrosis in CHC infection. The unique 100% segregation of the most common K8 variant, R341H, with an intronic deletion suggests that one of these two genetic changes might lead to the other. (HEPATOLOGY 2006;43:1354–1363.)

Hepatitis C virus (HCV) is the major cause of chronic hepatitis in the Western world and the most common chronic blood-borne infection in the United States.1 HCV affects over 120 million individuals worldwide,1 and its clinical course is variable, leading to cirrhosis in 4% to 22% of chronically infected patients within 20 years whereas other affected individuals may never develop advanced liver fibrosis.2, 3 Due to its clinical variability, the limited success and availability of treatments, better predictability of HCV disease progression is vital to make treatment decisions and to tailor the appropriate therapy. Several environmental and host factors including male sex, older age at acquisition of infection, obesity and simultaneous infection with hepatitis B or HIV negatively affect the course of the disease.1, 4 Furthermore, genetic variations in genes encoding immunoregulatory proteins, cytokines, and fibrogenic factors act as important modifiers of disease progression.4 Moreover, polymorphisms in the keratin cytoskeletal proteins predispose their carriers to the development of end-stage liver disease5–7 but potential association of keratin variants with HCV infection or its progression is unknown.

The cytoskeleton of eukaryotic cells consists of an extensive 3-dimensional network of microtubules, microfilaments, and intermediate filaments (IFs).8, 9 IFs make up a large family of proteins that are expressed in a tissue specific manner and include the nuclear lamins and cytoplasmic proteins such as keratins in epithelial cells, desmin in muscle and neurofilaments in neuronal cells.10–12 All IF proteins have the conserved structural feature of a central α-helical “rod” domain that links the non-α-helical N-terminal “head” and C-terminal “tail” domains.11, 12

Within IF proteins, keratins are the largest subfamily and consist of >20 unique gene products,13, 14 with types I and II corresponding to keratins 9-20 (genes KRT9-KRT20, proteins K9-K20) and keratins 1-8 (genes KRT1-KRT8, proteins K1-K8), respectively. Keratins are found as obligate type I and type II non-covalent heteropolymers (i.e., at least one type I and one type II keratin) with expression of unique pairs as the major cellular IF in a tissue specific manner.13, 14 This heteropolymeric organization explains the autosomal dominant phenotype that is seen in most IF-related diseases.15 In digestive organs, “simple” (i.e., single layered) epithelia express K8 together with variable levels of K7, K18, K19 and K20 depending on the tissue.9, 13, 16 Adult hepatocytes are unique among simple epithelial cells in that they express exclusively K8 and K18, in contrast to other glandular epithelia (e.g., ductal, intestinal) which express one or more additional keratins. The “exclusive” expression of K8/K18 in hepatocytes renders the liver the most sensitive digestive organ to K8/K18 mutation or ablation as shown in several transgenic and knockout animal models.17, 18

In humans, mutations in 13 of the >20 known keratins cause a wide range of epidermal, oral, and ocular diseases.15, 19 In contrast to non-simple epithelial keratin mutations which cause several skin and other tissue-specific diseases with near-complete to complete penetrance, K8/K18 variants predispose to, rather than cause, disease likely due to the location of the K8/K18 variants in non-conserved domains of the protein backbone.17, 20 The only established association of keratin mutations with digestive disorders is the association of K8/K18 variants with end-stage liver disease. This is based on several studies involving one large U.S. patient cohort derived from three transplant centers that noted 12.4% of 467 liver explants (with multiple liver disease etiologies) harbored K8/K18 variants as contrasted with a 3.7% variant frequency in a control group of 365 blood bank donors (P < 0.0001; prevalence OR = 3.8 with 95% CI = 2.1-7.1).7 It is not known if K8 or K18 variants have a preferential association with a specific liver disease since the U.S. patient cohort included liver diseases of multiple etiologies.

In contrast to the findings by Ku et al,5–7 two European studies involving German patients with chronic liver diseases of multiple causes and unspecified severity either did not observe any K8/K18 variants upon examination of the entire K8/K18 exonic regions21 or did not confirm an association with liver disease when only two K8 variants Y54H and G62C were examined.22 However, the inability to identify K8/K18 variants in the entire coding region analysis21 was due to limited sensitivity of the experimental conditions,23 while the lack of association of the K8 Y54H/G62C variants with liver disease in the K8 variant-specific study22 likely represented population differences, incomplete staging of analyzed liver disease or the limitation of focusing on only two keratin variants.7, 23 Therefore, we undertook this study to test for the presence of K8/K18 variants in a well-defined German patient cohort with chronic HCV (CHC) infection. This allowed us to address in a single study aspects that were not addressed in prior studies5–7, 21–23 including: potential population differences, a single disease entity, and whether K8 or K18 variants correlate with progression of liver disease.


HCV, hepatitis C virus; IF, intermediate filament; WHO, World Health Organization; BMI, body mass index; CHC, chronic hepatitis C; DHPLC, denaturing high-performance liquid chromatography; IVS, intervening sequence; K, keratin protein; KRT keratin gene; UTR, untranslated region.

Patients and Methods

Selection of Patients.

The study was performed in 336 CHC patients who underwent liver biopsy at the University of Erlangen-Nürnberg between the years 2001-2004. Most of these patients were part of a previous study that addressed glutathione S-transferase P1 and manganese superoxide dismutase polymorphims in liver disease.24 Seven of the 336 patients were excluded from further analysis since their isolated DNA could not be reliably amplified. The diagnosis of CHC was confirmed by a positive HCV RNA (Cobas Amplicor; Roche, Basel, Switzerland). Virus genotyping was performed using the Inno-LipQA HCV II test (Innogenetics, Ghent, Belgium). For inflammation grading and fibrosis staging, 4 μm-thick liver sections were stained with hematoxylin-eosin and Masson trichrome, respectively. The classification was done according to Ludwig25 which scores fibrosis and inflammation from 1-4 and largely concurs with the Metavir scoring system.26 Stained sections were independently scored by the hospital senior pathologist using established criteria25, 26 and by a second histopathologist who was not aware of the patient's clinical data except for the condition “chronic hepatitis C infection”.24 The duration of CHC was estimated based on the presence of a qualifying life event in most cases, and when such an event was lacking then by the dates that corresponded to the first signs of hepatitis. Absence of significant alcohol consumption was estimated in accordance to the World Health Organization (WHO) recommendations (<20 g alcohol/day for females and <60 g alcohol/day for males). Other causes of chronic liver disease including chronic hepatitis B virus infection, hereditary hemochromatosis, autoimmune liver disease, chronic cholestatic liver disorders were excluded by appropriate diagnostic means during the clinical work-up. The study protocol was approved by the Human Subjects Committees of the participating centers (University of Erlangen-Nürnberg in Germany and Stanford University in the United States).

Genetic Analysis.

Genomic DNA was isolated from peripheral blood using a DNeasy tissue kit (Qiagen; Valencia, CA). Primers were chosen based on the genomic sequences M34482.1 and AF179904.1 for KRT8 and KRT18, respectively and included the complete exonic regions of K8 and K18 with the adjacent exon-intron boundaries (Supplementary Table S1; Supplementary material for this article can be found on the HEPATOLOGY website ( DNA fragments were amplified with a “hot-start” Amplitaq Gold DNA Polymerase (Applied Biosystems; Foster City, CA) and the specificity of the amplification was confirmed by sequencing (ABI 377 sequencer; Applied Biosystems). PCR products were analyzed by denaturing high-performance liquid chromatography (DHPLC)27 using a WAVE DNA Fragment Analysis system as described.23 Samples with “shifted” elution pattern were purified with a Qiaquick PCR purification kit (Qiagen) and sequenced in both directions using an ABI 377 sequencer (Applied Biosystems). The location of K8/K18 variants was assigned based on the mRNA sequences NM 002273.2 and NM 000224.2.

Statistical Analysis.

For this retrospective study, we analyzed the occurrence of K8/K18 variants in patients using univariate and multivariate logistic regression (SPSS software, version 14; SPSS, Chicago, IL). Separate analyses were run for the outcome defined as the presence of any K8/K18 variant or as the presence of “likely biologically significant” K8/K18 variants. We ran separate univariate analyses for both outcomes and all variables in our study, yielding estimates and 95% confidence intervals for the unadjusted odds ratios for the presence of a variant given a one-unit change in the value of the predictor.

In an observational study such as this, multivariate analyses are often more reliable than univariate analyses, as multivariate analyses adjust for the simultaneous effects of other variables. Thus, we evaluated a multivariate logistic regression model with gender, age, fibrosis stage, inflammation grade, body mass index (BMI) and alcohol consumption as predictors. The remaining variables had missing values for several patients and were not included so as not to decrease the number of patients in the complete-case multivariate analyses. We computed estimates and 95% confidence intervals for the adjusted odds ratios for the presence of a variant given a one-unit change in the value of a predictor, and the other predictors in the model. All P values are based on standard Wald tests.


Patient Demographics and Identified Keratin Variants.

The demographic data and clinical characteristics of the cohort are summarized in Tables 1 and 2, respectively. More than 95% of the patients were Caucasian (Supplementary Table S2), 62% were males and 21% met WHO criteria for significant alcohol consumption (Table 1). The majority of the patients had HCV genotype 1 (74%) and presented with mild-moderate inflammation (grades 0-2, 88%) which is typical for chronic hepatitis C, and 37% of the patients had advanced fibrosis (stages 3 and 4) (Table 2). A search for keratin variants in the entire coding regions of K8 and K18 identified a total of eight K8 and one K18 heterozygous variants that lead to an amino acid change in 24 patients (Table 3). Among the 9 K8/K18 mutations, three were novel K8 polymorphisms that had not been described previously and included K8 Thr26→Arg (T26R) (Fig. 1A, Table 3), K8 G55A and A359T (Table 3). A total of 20 patients had single and 4 had compound variants (two K8 A319S+R341H, one K8 R341H+A359T and one K8 I63V+K18 S230T) (Table 3). Not included in Table 3 or in any subsequent analysis is a previously described silent and common heterozygous K8 L227L variant7, 21 that we observed in 51% of the patients (not shown).

Table 1. Demographics of Patient Cohort
Patient DemographicsMaleFemaleTotal
  • Abbreviation: BMI, body mass index.

  • *

    The cutoff criteria for alcohol consumption were set <20g/day for females, and <60g/day for males in accordance to WHO recommendations.

  • **

    Duration of the disease was not known in 30 patients.

Patients, n (%)205 (62%)124 (38%)329 (100%)
Age, mean ± SD, years46.8 ± 13.546.9 ± 12.246.8 ± 13.0
BMI, mean ± SD25.8 ± 4.024.4 ± 4.325.3 ± 4.2
Alcohol consumption yes/no*50/155 (24%)19/105 (15%)69/260 (21%)
Duration, mean ± SD, years**13.4 ± 9.314.1 ± 8.813.7 ± 9.1
Table 2. Histology and HCV Genotype Characteristics of Patient Cohort
CharacteristicsPatients, n (%) [N = 329]
Viral genotype 
1243 (74)
29 (3)
367 (20)
1 + 32 (1)
Unknown8 (2)
Fibrosis stage 
1108 (33)
298 (30)
345 (14)
478 (23)
Inflammation grade 
021 (7)
1131 (40)
2135 (41)
330 (9)
44 (1)
Unknown8 (2)
Table 3. Distribution of Exonic Keratin Variants and Fibrosis Stage in CHC Patients
GeneVariantsFibrosis Stage (# of Patients)Total # of Variants
NucleotideAmino Acid1 (108)2 (98)3 (45)4 (78)
  • NOTE. The table lists the number of patients for each keratin variant per fibrosis stage (stages 1-4). The location of K8/K18 variants was assigned based on the mRNA sequences NM 002273.2 and NM 000224.2. The amino acid number includes the first codon.

  • *

    Variants highlighted with an asterisk are considered “polymorphisms” rather than “mutations” that are likely to have biologic relevance based on previous association studies.7 These polymorphisms (K8 I63V/A319S and K18 S230T) were not included in the analysis of “significant variants”.

  • **

    Brackets indicate the number of compound variants.

  • Newly identified variants which were not previously described.

 187A→GI63V*0004 [1]**4 [1]
 955G→TA319S*1 [1]001 [1]2 [2]
 1022G→AR341H3 [1]2 [1]14 [1]10 [3]
 1075G→AA359T01 [1]001 [1]
KRT18689G→CS230T*00022 [1]
# patients with any variant (%)4 (3.7%)4 (4.1%)3 (6.7%)13 (16.7%)24 (7.3%)
# patients with significant variants (%)*4 (3.7%)4 (4.1%)3 (6.7%)8 (10.3%)19 (5.8%)
Figure 1.

Identification of the variants K8 T26R and KRT8 IVS7+10delC. PCR-amplified genomic material was analyzed by DHPLC and samples with a shifted chromatogram (not shown but see Strnad et al.23) were analyzed further by DNA sequencing as described in Patients and Methods. Shown are DNA sequencing results of amplified genomic material from patients with wild-type K8 exon 1 and a heterozygous K8 T26R mutation (A), or with wild type KRT8 exon 7 and a heterozygous KRT8 IVS7+10delC variant (B). The heterozygous nature of the mutation is evident by the presence of wild-type and mutant allele sequences.

Characterization of the Exonic K8/K18 Variants.

We classified the variants we identified as simple polymorphisms that might not have biological relevance (K8 I63V/A319S and K18 S230T), and as likely mutations with suggested biological relevance as we and others have described previously (K8 T26R/Y54H/G55A/G62C/R341H/A359T).5–7, 28 This classification takes into consideration the biochemical property of the variant when known (e.g., its effect on keratin solubility or crosslinking), its frequency in the general population, or its predicted likelihood to disrupt keratin filament organization.7, 28 Among the variants we identified, K8 R341H was the most common and was found in 10 patients followed by K8 G62C and K8 I63V with six and four affected individuals, respectively. Interestingly, the single K8 Y54H variant (Table 3) was found in a patient who was from the United States with African American origin, thereby lending support to the association of this variant with individuals of African descent.7 A close inspection of the distribution of exonic keratin variants showed a higher association of keratin variants with advanced fibrosis stages (Table 3) and inflammation grades (Table 4) (see Logistic Regression). Given that most of our patients were of German Caucasian background, most of the variants were also found in this group of patients (Supplementary Table S2).

Table 4. Distribution of Exonic Keratin Variants and Inflammation Grade in Patients With CHC
GeneVariantsInflammation Grade (# of Patients)Total # of Variants
NucleotideAmino Acid0 (21)1 (131)2 (135)3 (30)4 (4)
  • NOTE. The table lists the number of patients for each keratin variant per inflammation grade.

  • *

    Variants highlighted with an asterisk are considered “polymorphisms” rather than “mutations” that are likely to have biologic relevance based on previous association studies.7 These polymorphisms (K8 I63V/A319S and K18 S230T) were not included in the analysis of “significant variants”.

  • **

    Brackets indicate the number of compound variants.

  • Newly identified variants which were not previously described.

  • One individual carrying the K8 R341H/A319S variant is excluded from this table since he was among the 8 patients in whom the grade of inflammation was not known.

 187A→GI63V*013 [1]004 [1]
 955G→TA319S*010001 [1]
 1022G→AR341H03 [1]321 [1]**9 [2]
 1075G→AA359T000011 [1]
KRT18689G→CS230T*011 [1]002 [1]
# patients with any variant (%)0 (0%)8 (6.1%)11 (8.1%)3 (10.0%)1 (25%)23 (7.2%)
# patients with significant variants (%)*0 (0%)6 (4.6%)8 (5.9%)3 (10.0%)1 (25%)18 (5.6%)

Characterization of the Non-Coding K8/K18 Variants.

In addition to the heterozygous exonic variants shown in Table 3, we also identified 9 heterozygous polymorphisms in non-coding sequences that are within the analyzed parts of K8/K18 introns, promoter and 3′-untranslated region in a total of 26 patients (Table 5). Three of these non-coding variants were previously described: KRT8 IVS7+10delC (i.e., non-coding deletion found 10 nucleotides downstream from the 3′-end of exon 7 in intervening sequence 7)29; KRT8 IVS6+46A→T23 and KRT18 IVS1-4C→G (i.e., non-coding substitution found 4 nucleotides upstream from the 5′-end of exon 2 in intervening sequence 1).21 Eleven patients had compound exonic/non-coding variants (ten K8 R341C plus KRT8 IVS7+10delC; and one K8 G62C plus KRT18 nt-15C→T) in proximity to the exon-intron boundary (Table 5). Notably, the most common non-coding variant, KRT8 IVS7+10delC (Fig. 1B) was completely and exclusively associated with the K8 R341H polymorphism and all 10 identified patients carried both variants. There was no apparent association between the non-coding keratin variants and the severity of fibrosis (Table 5) or inflammation (not shown).

Table 5. Distribution of Non-coding Keratin Variants and Fibrosis Stage in Patients With CHC
GeneVariantsFibrosis stage (# of Patients)LocationTotal # of variants
Nucleotide1(108)2 (98)3 (45)4 (78)
  • NOTE. The location of K8/K18 variants was assigned based on the genomic sequences M34482.1 and AF179904.1. IVS, intervening sequence; nt, nucleotide.

  • *

    The frequency of non-coding variants excluding compound exonic-intronic variants.

  • **

    Numbers within brackets indicate the number of compound variants. The IVS7+10delC intronic variant associates in all cases with the presence of K8 R341H in the same individuals.

  • Newly described intronic variants that have not been previously reported.

 IVS7+10delC3214Intronic10 [10]**
KRT18nt-15C→T0001Promoter1 [1]
Total variants (%)9 [3] (5.6%)*6 [2] (4.1%)*4 [1] (6.7%)*7 [5] (2.6%)* 26 [11] (4.6%)*

Logistic Regression Analysis of Exonic K8/K18 Variants.

Univariate analysis (Table 6) for the presence of any variant showed a strong association of K8/K18 exonic variants with increasing fibrosis (P = .002). We also noted nonsignificant trends for an association of these variants with increasing inflammation and with the MnSOD A16V polymorphism (P = .07 for both; Table 6).

Table 6. Univariate and Multivariate Analysis of K8/K18 Exonic Variant Associations
Factor (Incremental comparative unit*)Total variantsSignificant variants
OR (95% CI)P ValueOR (95% CI)P Value
  • *

    Incremental comparative unit represents the unit of comparison for a given odds ratio or P value. For example, a patient with fibrosis stage 2 has ∼2.1 higher risk of having a keratin variant using multivariate analysis than a patient with fibrosis stage 1, and a patient with fibrosis stage 3 has ∼2.1 higher risk of harboring a keratin variant than a patient with fibrosis stage 2.

Univariate analysis:    
Gender (male:female)1.51 (0.61–3.76)0.371.75 (0.61–4.97)0.30
Age (1 year)1.00 (0.97–1.03)0.860.98 (0.95–1.02)0.36
Duration (1 year)1.00 (0.95–1.05)0.871.00 (0.95–1.05)0.91
Fibrosis (1 stage)1.85 (1.26–2.70)0.0021.48 (0.99–2.21)0.05
Inflammation (1 stage)1.61 (0.96–2.70)0.071.79 (1.01–3.18)0.05
BMI1.06 (0.97–1.16)0.221.04 (0.94–1.15)0.44
Alcohol (yes:no)1.28 (0.49–3.36)0.621.01 (0.32–3.13)0.99
Virus genotype (III vs. I)0.77 (0.29–2.02)0.860.70 (0.24–2.04)0.81
MnSOD A16V variant1.84 (0.95–3.54)0.071.59 (0.78–3.25)0.21
GstP1 I105V variant0.99 (0.49–2.00)0.980.93 (0.42–2.04)0.86
Multivariate analysis:    
Gender (male:female)1.26 (0.48–3.30)0.641.44 (0.48–4.30)0.52
Age (1 year)0.96 (0.92–1.00)0.060.94 (0.90–0.99)0.03
Fibrosis (1 stage)2.05 (1.33–3.15)0.0011.67 (1.04–2.67)0.03
Inflammation (1 stage)1.50 (0.88–2.59)0.141.78 (0.97–3.26)0.06
BMI1.10 (0.99–1.21)0.081.09 (0.97–1.22)0.14
Alcohol (yes:no)1.22 (0.44–3.36)0.700.90 (0.27–2.95)0.86

Multivariate logistic regression (Table 6) showed that the association with fibrosis was independent of other variables (P = .001) with an estimated 2.05 fold increase in the odds of a keratin variant for each one-stage increase in the fibrosis scale. Inflammation became more nonsignificant (P = .14) in the multivariate analysis, suggesting that its univariate association might be related to that of other variables in the model. The results for age (P = .06) and BMI (P = .08) approached, but did not reach, significance in the multivariate model.

Separate univariate analyses using stricter criteria for variant inclusion (i.e., analysis of the likely relevant variants K8 T26R/Y54H/G55A/G62C/R341H/A359T) showed a marginally significant association with inflammation (P = .048) and a marginally nonsignificant association with fibrosis (P = .054). The adjusted multivariate analysis, controlling for the effect of other variables, yielded significant associations of likely keratin mutations with both fibrosis (P = .03) and age (P = .03). A nonsignificant trend (P = .06) was seen for inflammation.


The major findings of our study are: (1) characterization of germline amino-acid-altering K8/K18 heterozygous variants in 24 of 329 (7.3%) patients with CHC (excluding the common silent K8 L227L polymorphism) and non-coding variants in 26 patients (7.9%) 11 of whom overlap with the 24 patients that harbor exonic variants (summarized in Fig. 2); (2) the identification of 3 novel exonic K8 variants (T26R, G55A, A359T) and 6 novel non-coding variants (Table 5); (3) K8 R341H is the major K8/K18 variant in German patients with HCV (10 patients; 3% of all patients) followed by K8 G62C and I63V that were found in six and four patients, respectively (Fig. 2); (4) the exclusive segregation of K8 R341H with the intronic KRT8 IVS7+10delC deletion and, importantly; (5) the significant association of exonic K8 variants with progressive fibrosis in patients with HCV which is consistent with our hypothesis that K8/K18 variants predispose to liver injury. The number of variants we detected may be somewhat underestimated due to potential experimental condition limitations. For example, DHPLC can easily miss homozygous variants under the analysis conditions we used, but the frequency of such variants is likely to be rare due to the low frequency of individual heterozygous K8/K18 variants.23, 27 No association of K8/K18 variants with disease duration was noted (Table 6), but this result might be caused by the inherent imprecision of such measurements. A prospective study that accurately follows disease progression will be needed to address this question.

Figure 2.

Distribution of the keratin variants within the KRT8 and KRT18 genes. The amino acid positions of K8 and K18 protein domains (head, rod and tail) and subdomains [IA, IB, II, L (linker) 1, L1-2, L2] are indicated. These protein domains are conserved in all IF proteins. The shaded regions at the beginning and end of the rod domain correspond to mutation “hot spots” for epidermal keratins which notably have not been found in any of the hitherto identified K8 or K18 variants. Previously characterized K8/K18 variants are highlighted above each protein backbone schematic while the variants characterized in this study (coding variants in black color and non-coding variants in light gray color) are displayed below each schematic. The regions immediately upstream (promoter) and downstream (3′-UTR) of the KRT8 and KRT18 genes with their respective variants are shown. The approximate location of the intronic variants, in relationship to the coding sequences is also shown although for schematic and simplification purposes the introns themselves are not shown. Amino acids are denoted by the single-letter abbreviations, intervening sequences by “IVS” and nucleotides by “nt.” The most common K8/K18 variants that were found in prior studies are marked as 1-4 (one being most common) and those characterized in the current study as 1 and 2 (unmarked variants are found in much lower frequency). Note the three novel K8 coding variants (T26R/G55A/A359T) and several non-coding K8 and K18 variants which were not previously described (based on their presence below but not above the protein backbone schematic). Also of note, the K8 intronic variant IVS7+10delC was previously described,23, 29 but its exclusive association with the exonic K8 R341H variant is one of the findings described in this study.

In the multivariate model (Table 6), the age effect (P = .06 for “total variants” and P = .03 for “significant variants”) appears to represent a predictive power for the presence of a mutation that adds to the linear effect of fibrosis on the log odds. There are at least four possible explanations of an age effect in this analysis. It may be that mutations are associated with differential age-of-onset or age-related disease progression. It is also possible that the age effect represents a correlation of age with clinical characteristics, not otherwise included in our analysis, that are associated with the presence of the mutation. The age effect could also represent additional effects of fibrosis, if the relationship between age and fibrosis is non-linear. Lastly, it is possible that there is some cohort effect involving demographic changes in the population due to migration. In this retrospective study, we have inadequate information to clearly differentiate among these possible explanations.

The findings of this study not only demonstrate an association of keratin variants with a specific and highly prevalent liver disease (i.e., CHC) and with its progression in a German patient cohort, but are also in line with the frequency of K8/K18 variants in a U.S. cohort where liver explants were analyzed.7 Hence, our conservative estimate of likely relevant mutations shows a frequency of 10.3% in patients with stage 4 fibrosis (Table 3) which compares with the 12.4% prevalence seen in liver explants (i.e., harboring end-stage liver disease/cirrhosis) from multiple liver disease etiologies.7 The overall frequency of 5.8% (19 patients with “relevant” exonic variants from 329 patients, Table 3) is also consistent with our hypothesis since our cohort includes a broad range of disease severity, and is significantly higher than the frequency of these variants that is seen in a U.S. blood bank donor pool (3.7%).4 Similar to previous studies, a common heterozygous K8 L227L variant7, 21 was detected in 51% of patients. This variant is unlikely to have biologic significance, and was not included in the analysis, since it was common in all fibrosis stages (not shown) and does not cause an amino acid substitution.

Some differences in specific variant frequency do exist in terms of what has been described to date7, 23 versus what we observe in this study (Fig. 2). For example, we describe herein coding and non-coding variants that have not been previously found. The non-coding variants may affect keratin splicing and consequently protein expression levels, a possibility that remains to be tested. In addition, some of the previously described variants have not been seen in our patient cohort (and vice versa) and this applies in particular to K18 exonic variants where we observed only K18 S230T. Inability to detect most of the K18 variants that have been previously described may represent population differences, including ethnic background (e.g., some variants are seen primarily in African Americans7 and our cohort had only two such patients), or disease association or severity differences (given that we looked in this study at a pure disease entity with a range of disease severity).

With regard to the coding and non-coding K8/K18 variants we identified, several additional points deserve mention. First, of the three novel exonic variants we report herein, K8 T26R and G55A were found in patients with stage 4 liver fibrosis and the carrier of K8 R341H plus K8 A359T compound polymorphism exhibited severe inflammation, which was otherwise rarely observed in our CHC patient cohort (Tables 3 and 4). K8 G55 and A359 are highly conserved among K8 from other species and other type II keratins, while K8 T26 is conserved (as Thr or Ser) throughout species but is substituted by other amino acids in other type II keratins (Supplementary Fig. S1). Interestingly, although it is not known if K8 Thr26 is phosphorylated, its mutation to Arg removes a potential protein kinase C phosphorylation site (with consensus sequence 23RXXTSGP) and introduces a new potential c-AMP-dependent phosphorylation site at Ser27 (with consensus sequence 23RXXRSGP)30. Second, the K8 I63V variant was found only in patients with severe liver fibrosis (Table 3) and this variant was described in patients with inflammatory bowel disease and may have an effect on keratin filament organization28 although it was overrepresented in a blood bank donor group.7 Additional studies will be needed to assess the biologic relevance of this variant. Third, is the exclusive association of the intronic variant KRT8 IVS7+10delC with the exonic K8 R341H. Of note, KRT8 IVS7+10delC was not seen in any genomic DNA from any other patient with or without any other keratin coding or non-coding mutation (not shown). Such a strong association of these two K8 intronic and exonic variants is intriguing since it suggests the possibility of a causal relationship between these two variants. We propose that one variant triggers the generation of the other during DNA replication, a hypothesis that remains to be tested.

The potential mechanisms by which K8 and K18 variants may predispose to liver injury include their established role in hepatocyte cytoprotection, as clearly documented by findings of several transgenic mouse studies that express K8/K18 mutants or that lack K8/K18.17, 18, 31 Interestingly, keratin protein and RNA induction during liver injury, including viral hepatitis infection in animals, is reminiscent of a stress protein response.17, 31–33 Keratins may also serve to sequester oxidatively damaged proteins34 and alter the threshold towards oxidative injury.35 In addition, K8 and K18 serve as anti-apoptotic proteins and their ability to do so is hampered when keratins are absent or mutated.36–38 Such keratin-mediated anti-apoptotic function may be highly relevant given the importance of apoptosis and oxidative stress in CHC infection.39, 40 In addition, keratin mutation may interfere with K8/K18 filament assembly,5, 28 decrease keratin solubility (particularly the Cys mutants, such as K8 G62C, likely due to formation of –S-S- crosslinks),7 or may alter the ability of K8/K18 to serve as physiologic kinase substrates as demonstrated in vitro7 and in transgenic mice (unpublished observations). At this stage it is an open question whether one or more of these keratin-mediated effects and/or whether virus replication/propagation effects (e.g., HCV core protein interact with keratins in transfected cells)41 are most impacted by keratin mutation during HCV infection. Regardless, the involvement of KRT8 and KRT18 as likely susceptibility genes for HCV progression highlights the unique involvement of epithelial-specific cytoskeletal proteins in such predisposition. KRT8/KRT18 involvement in liver disease adds these genes to the few better characterized genetic risks in CHC which include immunoregulatory proteins, such as cytokines, fibrogenic factors and enzyme polymorphisms.1, 4, 24


We are grateful to Nahid Madani for assistance with DNA sequencing, Kris Morrow for expert assistance with figure preparation, and all the patients and their families who participated and made this work possible.