Article first published online: 22 FEB 2006
Copyright © 2006 American Association for the Study of Liver Diseases
Volume 43, Issue 3, pages 573–580, March 2006
How to Cite
Morishima, C., Paschal, D. M., Wang, C. C., Yoshihara, C. S., Wood, B. L., Yeo, A. E. T., Emerson, S. S., Shuhart, M. C. and Gretch, D. R. (2006), Decreased NK cell frequency in chronic hepatitis C does not affect ex vivo cytolytic killing. Hepatology, 43: 573–580. doi: 10.1002/hep.21073
Potential conflict of interest: Nothing to report.
See Editorial on Page 395
- Issue published online: 22 FEB 2006
- Article first published online: 22 FEB 2006
- Manuscript Accepted: 7 DEC 2005
- Manuscript Received: 8 APR 2005
- INBRE program of the National Center for Research Resources. Grant Number: 2P20RR16466
Prior studies have suggested that natural killer (NK) cell function might be impaired in chronic hepatitis C virus (HCV) infection. Circulating NK cell frequency and cytolytic activity were examined freshly ex vivo in HCV-infected and uninfected subjects. Surprisingly, the intrinsic cytolytic activity of peripheral blood NK-enriched cells was similar between HCV-infected and uninfected groups (P = .91). Although the percentage of circulating CD3−CD16/56+NK cells was 30% lower in HCV-infected compared with uninfected subjects (P = .02) paralleled by a decrease of CD56dim cytolytic NK cells (P = .02), overall K562 cytolysis by unfractionated peripheral blood mononuclear cells was not affected (P = .29). Analysis of the relationships between NK cytolytic activity and other clinical information revealed an inverse association with liver fibrosis stage (P = .035). In conclusion, NK cell cytolytic function does not appear to be impaired in chronic hepatitis C, but higher levels of NK cell cytolysis are associated with less liver fibrosis. (HEPATOLOGY 2006;43:573–580.)
Hepatitis C virus (HCV) is able to establish a chronic persistent infection in the vast majority of individuals exposed,1 making it likely that the virus uses multiple strategies to elude the immune response it elicits. Several groups have demonstrated that hepatitis C–specific adaptive immunity is attenuated in chronic disease.2–5 However, relatively little is known about the role of the innate immune response in hepatitis C, despite the fact that innate immunity has the ability to significantly influence the subsequent adaptive immune response.6–8
A number of investigators have studied natural killer (NK) cell function on exposure to HCV components in vitro, with intriguing results. Tseng et al.9 and Crotta et al.10 simultaneously reported that NK cell function could be diminished in vitro by interactions with HCV envelope protein E2, anti-E2 antibody (via CD81), and anti-CD16. Whereas one implication from these studies is that diminished NK cell function early in the course of acute infection may explain HCV persistence, the effects described by these investigators could also pertain to chronic infection. However, studies of ex vivo NK cell function in patients with chronic hepatitis C have yielded mixed results. Corado and colleagues first reported that NK cell cytotoxicity was compromised in chronic hepatitis C compared with healthy negative controls,11 using a well-established human NK cell target cell line, K562.12 Similar data were published by Par and colleagues,13 but Duesberg et al. recently demonstrated that cytotoxicity and antibody-dependent cell-mediated cytotoxicity against HT29 cells were not adversely affected in chronic hepatitis C.14 Moreover, Kawarabayashi et al. found no difference in K562 cytolysis between liver mononuclear cells containing a mean of 31% CD56+ NK cells between 12 HCV-negative and 12 HCV-infected subjects without cirrhosis.15 Thus, the existence of an in vivo NK cell defect in chronic HCV infection has not been firmly established.
We sought to determine whether the frequency of NK cells in the peripheral blood mononuclear cell (PBMC) population or intrinsic cytolytic activity by NK cells was reduced in chronically HCV-infected individuals. In addition, we examined associations among NK cell frequency, NK cytolysis in unfractionated PBMC, and clinical factors.
Materials and Methods
All study subjects were sequentially recruited at the University of Washington, Harborview Medical Center, Seattle, WA, after written informed consent was obtained through an Institutional Review Board–approved protocol. The 26 HCV-uninfected (or HCV-negative) subjects were hepatitis C antibody negative, had undetectable serum HCV RNA (<100 IU/mL) by either Roche COBAS Amplicor assay v. 2.0 or real time RT-PCR,16 and had normal alanine aminotransferase (ALT) levels (mean ALT = 15.7 ± 8.8). By contrast, all 42 HCV-infected treatment-naïve subjects had detectable serum or plasma HCV RNA with a mean viral titer of 3.4 × 106 IU/mL (median, 1.4 × 106; range, 0.000825 × 106 to 37.1 × 106) and mean ALT level of 87.9 U/L (median, 60; range, 16–466). Sixty-nine percent were infected with HCV genotype 1 and 31% with genotypes 2 or 3. The proportion of males among the HCV-infected and HCV-uninfected groups was similar (55% and 50%, respectively; P = .34); however, subjects in the HCV-infected group were older relative to the HCV-uninfected group (mean age, 43.5 and 37.7, respectively; P = .036). Liver biopsy data were available in 16 subjects. Fibrosis stage was scored from 0 to 4 using the Batts-Ludwig criteria.17
Flow Cytometric Analysis.
The frequency of CD45+CD3-CD16/56+ NK cells among fresh PBMCs was evaluated using flow cytometric analysis of 1% paraformaldehyde-fixed PBMC surface labeled with anti-CD45-PerCP, anti-CD3-FITC, anti-CD16-PE, and anti-CD56-PE (Becton Dickinson, San Jose, CA) and isotype matched controls after PBMC isolation by Ficoll density centrifugation (Ficoll-Paque Plus, Amersham Pharmacia Biotech, Piscataway, NJ). Four-color analysis was performed in later samples using anti-CD45-PerCP, anti-CD16-FITC, anti-CD56-PE (Becton Dickinson), and anti-CD3-APC (Beckman Coulter, Inc., Miami, FL). The data were collected using a Becton Dickinson FACS Calibur at the Puget Sound Blood Center (Seattle, WA) and analyzed using FlowJo software for Macintosh (Tree Star, Inc., Ashland, OR). The gating strategy for identifying the population of interest is shown in Fig. 1A and described in the figure legend. The percentage of CD16 and CD56-expressing cells amongst PBMC was calculated using these flow cytometric data, and best represents the cells that were manually counted in the hematocytometer as live cells, and used in concurrent cytolytic assays. The number of CD3-CD16/56+ NK cells was computed using manual cell counts obtained after isolation of PBMC and the percentage of live CD45+CD3−CD16/56+ NK cells as determined by flow cytometry.
NK Cell Isolation and Cytolytic Assay.
NK cell–enriched populations were freshly isolated from whole blood by Ficoll density centrifugation followed by enrichment using the StemSep™ human NK cell enrichment cocktail (Stem Cell Technologies, Vancouver, BC, Canada) containing antibodies specific for CD3, CD4, CD14, CD19, CD66b, and glycophorin A, or after a 20-minute incubation with antibodies specific for CD3, CD4, CD19, CD36, and CD66b in the NK Rosette Separation Kit (Stem Cell Technologies). Triplicate wells of highly enriched NK cells were co-cultured with a human NK target cell line, K562,12 at effector-to-target ratios of 25:1, 12.5:1, 6.25:1, 3.125:1, 1.56:1, and 0.78:1. Cytolysis was measured using a standard 4-hour 51-chromium (51Cr) release assay. The % specific lysis was calculated as follows: (mean of triplicate test wells − mean of background target lysis)/(mean of maximum target lysis − mean of background target lysis). When multiple measurements were available for the same HCV-negative subject (n = 12), the % specific lysis was averaged. Excluding single values, the range of means was 46.8% to 66.7% with a standard error range of 1.1 to 17.6.
The LD50 of cytolytic activity was defined as the half maximal degree of specific lysis estimated from a linear regression of each individual's NK assay results using cell numbers adjusted for the degree of purity of the NK cell populations. The LD50 measurement permitted the representation of all data points (from the different E:T ratios) by a single number to describe the ability of NK-enriched cells to induce cytolysis of K562 cells. Specifically, the input NK cell numbers used in the cytolytic assay (at each E:T ratio) were adjusted for purity by multiplying the input cell number by the percentage of NK cells in PBMC as assessed by flow cytometry and as calculated for Fig. 1. The log of the adjusted NK cell number (x-axis) was plotted against the log of the LD50-specific lysis (y-axis) obtained by using the formula: % specific lysis = mean lysis of triplicate test wells − mean minimum target lysis (background)/mean maximum target lysis − mean lysis of test wells. These data generate a line described by the formula, y = mx + b. At the y-intercept, where y = 0, x, or the LD50, is defined as −b/m. Thus, when the test lysis is half of the maximal lysis, then the % specific lysis is equal to 1 (where the y intercept = 0, because log101 = 0). Thus, because a higher LD50 value indicates that more cells are required to achieve the same level of lysis (half-maximal), a higher LD50 value corresponds with lower overall cytolysis.
Purity of isolated populations was defined by the percentage of CD45+CD3−CD16/56+ cells with antibodies specific for CD45, CD3, and CD16/56 or CD16 and CD56 separately (Becton Dickinson) as described. The mean purities of the NK-enriched populations were 84.3% (range, 67.5%-93%) and 77.1% (range, 46%-90%) for the HCV-negative and HCV-infected groups, respectively. Among CD45+ lymphocytes, both groups had mean purities of greater than 90%. Most specimens from HCV-infected subjects were tested in paired fashion with an HCV-negative specimen. For many HCV-negative subjects, multiple measurements on the same individual were averaged before comparing the two groups.
The data were summarized using appropriate descriptive statistics, and groups were compared using the chi-square test or Student t test for unequal variances where appropriate. Correlations were assessed using the Spearman correlation coefficient. The effect of a predictor on a continuous outcome was determined using linear regression with robust variance estimates. A P value less than .05 was considered significant.
The Percentage of Peripheral Blood NK Cells Is Decreased in Hepatitis C–Infected Subjects.
The percentage of live CD3−CD16/56+ NK cells in PBMC populations was determined by flow cytometry (Fig. 1). Cell surface labeling with anti-CD45 was employed to better distinguish lymphocytes from contaminating red blood cells in these freshly isolated cell populations, whereas anti-CD3, anti-CD16, and anti-CD56 were used to define the NK cells of interest (Fig. 1A). Early experiments used the mixture of CD16 and CD56 antibodies conjugated to the same fluorochrome to more comprehensively identify NK cells. These analyses indicated that the average percentage of NK cells was approximately 30% lower (95% CI, 15%-44% lower) in the chronically HCV-infected subjects (n = 28) compared with the HCV-negative subjects (n = 24, 2-tailed t test, P = .02, Fig. 1B). This observation held when CD56 alone was used to identify NK cells rather than CD16 and CD56 together (data not shown; P = .035). Comparison of the overall calculated CD3−CD16/56+ NK cell frequencies in the HCV-infected and uninfected populations also demonstrated lower NK cell numbers on average in the HCV-infected group (mean HCV-infected, 55,297 vs. mean HCV-uninfected, 111,443, P = .0003).
The Frequency of Peripheral Blood Cytolytic CD56dim NK Cells Is Decreased in Hepatitis C–Infected Subjects.
A CD3-CD56dim subset of NK cells has been described by several other groups to be primarily responsible for the cytolytic function of human NK cells.18–20 In HIV infection, a diminished CD3-CD56dim subset has been described to correlate with an overall decrease in cytolytic activity.21 We therefore wondered whether the number of CD3−CD56dim NK cells was measurably decreased in the HCV-infected group compared with the HCV-negative group and whether this might correspond to a difference in cytolytic function. A subset of subjects (21 HCV-uninfected and 17 HCV-infected) had PBMC analyzed using fluorescently labeled antibodies for CD3, CD56, and CD16 separately, and CD45. Typical data from a representative experiment are shown in Fig. 2. The mean percentage of live CD3-CD56dim NK cells was approximately 35% lower in the HCV-infected group (95% CI: 19%-50% lower; P = .020 using a two-sided t test) compared with the HCV-negative group (Table 1), corresponding with the overall decrease in the percentage of CD3−CD56+NK cells in the circulation of HCV-infected individuals. The strong correlation between CD56dim NK cell percentages and overall NK cell percentages (r = 0.91, P < .0001) is not surprising, given that the CD56dim subset comprises 90% or more of circulating NK cells in our cohort and in other published reports.18, 22 Thus, our data indicate that percentages of CD3−CD56+ or CD3−CD16/56+ NK cells and the cytolytic CD3−CD56dim NK cell subset are all decreased in HCV-infected subjects compared with HCV-uninfected subjects.
|% CD56dim ± SD||7.9 ± 4.3||5.2 ± 2.4|
|95% CI||5.9 to 9.9||3.9 to 6.4|
Cytolysis of K562 Cells by NK-Enriched Populations Does Not Differ Between HCV-Infected and Uninfected Subjects.
To directly assess NK cytolytic function in chronic hepatitis C, PBMC depleted of non-NK cells were used in standard 51Cr release assays at 6 Effector:Target (E:T) ratios as described in Materials and Methods. Specific lysis measurements were available on 24 of the HCV-negative subjects and 29 chronically HCV-infected subjects. The mean purities of the NK-enriched populations amongst PBMC were 84.3% (median, 86.7; range, 67.5%-93%) and 77.9% (median, 80; range, 46%-94%) for the HCV-negative and HCV-infected groups, respectively (t test, P = .017). In addition, the LD50, which took into account both the purity of the NK cell enrichment and the data from all 6 E:T ratios tested was calculated.
As shown in Fig. 3, no significant difference of K562 cytolysis was found between the HCV-infected and the HCV-uninfected groups (P = .91 and .64 for the LD50 and % specific lysis at an E:T ratio of 25:1, respectively). In addition, using linear regression analysis adjusted for NK cell purity, no difference in cytolytic killing by NK-enriched populations was found between the HCV-infected and uninfected groups (P = .55). Finally, when the analysis was limited to samples with NK purities of 75% or greater, the mean purities were 86.4% for the HCV-negative (n = 21) and 84.3% for the HCV-infected groups (n = 20), respectively (P = .19). Again, no differences were found in LD50 or % specific lysis between the 2 groups (P = .37 and P = .51, respectively). This is the first comparison of NK cytolytic function using NK-enriched cell populations from HCV-infected and uninfected subjects.
The same cytolytic assays were performed in the presence of 10% HCV-infected serum or 10% uninfected human serum, to determine whether the presence of HCV during the cytolytic assay would alter NK cell function. However, the results of 5 separate experiments did not show any consistent effect on NK cell function (data not shown).
PBMC From HCV-Infected and Uninfected Subjects Induce Similar Levels of K562 Cytolysis.
We hypothesized that the decreased frequency of CD3−CD16/56+ NK cells might result in decreased K562 cytolysis induced by the overall PBMC population from HCV-infected subjects. We therefore performed K562 cytolysis as described above using unfractionated PBMC as effectors. However, no difference in K562 cytolysis was found between PBMC isolated from the HCV-infected and uninfected groups (Fig. 4A, P = .29). Because a wide range of specific lysis was found in the HCV-infected group, the subjects were further subdivided arbitrarily according to whether their specific lysis results were greater or less than 10% specific lysis at an E:T ratio of 25:1. These two subsets of HCV-infected subjects were similar with respect to demographic and clinical characteristics, and circulating NK cell frequency (data not shown). However, those with higher K562 cytolysis by PBMC had higher cytolysis when NK populations were enriched (t test, P = .006 and .002 for the LD50 and % specific lysis at an E:T ratio of 25:1, respectively).
The Degree of Cytolytic Function Appears to Be Poorly Related to the Frequency of NK Cells.
Although NK cell frequency was decreased in hepatitis C–infected subjects, overall cytolysis by PBMC was not significantly decreased. We wondered whether any association could be detected between NK cell frequency and cytolytic killing. To maximize the sample size, these data were analyzed for all subjects together in Fig. 4B. The Spearman correlation coefficient of 0.33 was quite poor, and thus suggests only a very weak relationship between the percentage of CD3−CD16/56+NK cells among PBMC and overall K562 cytolysis by PBMC when all subjects were considered (Fig. 4B). Additional analyses using larger data sets will be required to better understand this relationship.
NK Cytolytic Activity Predicts Fibrosis Stage, But Not Grade.
Associations between the degree of NK-enriched cytolytic activity and demographic and clinical data (age, sex, ALT levels, HCV RNA level, HCV genotype, and liver biopsy) were sought using linear regression. No demographic or laboratory data were associated with NK-enriched cytolysis.
Liver histology was available on 16 HCV-infected subjects. Three (18%) had fibrosis stage 0, 7 (44%) had stage 1, 5 (31%) had stage 2, and 1 (6%) had stage 3. This subset of 16 subjects with liver biopsy data was not different from the rest of the cohort with respect to age (P = .25), sex (P = .10), ALT level (P = .13), or serum HCV RNA level (P = .90). Using the Mantel Haenszel chi square test, NK-enriched specific lysis (E:T, 25:1) was inversely associated with liver fibrosis stage (P = .035, Fig. 5). A significant association was also found between the LD50 and liver fibrosis stage (P = .048). These data suggest that the presence of cytolytically active peripheral NK cells may be protective for liver disease progression, as suggested by data from a murine model of liver fibrosis.23
The potential impairment of NK cell function in hepatitis C was suggested by early experiments testing unfractionated PBMC for killing of NK cell target cell lines.11 Two subsequent studies used NK cells from healthy subjects treated in vitro with anti-CD16, HCV E2, and anti-E2 to recapitulate NK cell activation by HCV, and found diminished NK cell function in those situations.9, 10 A third study suggested that dendritic cells from HCV-infected individuals were impaired and unable to activate NK cells for maximal function.24 Not all studies agree that NK function is compromised in chronic hepatitis C.14, 15 Moreover, no one has previously examined the contributions of both NK cell frequency and intrinsic cytolytic function to the NK cytolytic activity found in PBMC.
It was striking that cytolysis by NK-enriched cells, tested freshly ex vivo, did not appear to be attenuated by hepatitis C virus infection. Cytolytic function was remarkably similar between the HCV-infected and uninfected subjects, using either the measurement of specific lysis at an E:T of 25:1, or the LD50, which accounted for purity of the NK enrichment and cytolysis at all E:T ratios (P = .64 and P = .91, respectively).
The purity of the NK-enriched populations was meticulously monitored by flow cytometric analysis following the depletion procedure for each assay performed. These analyses indicated that both the HCV-infected and uninfected NK-enriched populations contained some non-NK cells. Basophils identified by surface labeling of CD203c were found to be a component of this contaminating population. The remaining cells were not specifically identified, other than that they had similar size and granularity with lymphocytes, and expressed the lymphocyte surface marker CD45, but not CD3, CD56, CD16, or CD203c. These non-NK and non-T cells obtained by the NK enrichment procedure were directly tested in a K562 assay for cytolytic activity, but none was found (data not shown).
It is notable that the LD50 calculation, which adjusted the input cell numbers for the killing assay based on the purity measurements, showed no difference in cytolytic ability between the HCV-infected and uninfected groups. Because the NK cell populations from the HCV-infected subjects were less pure than those from the uninfected subjects, the most likely finding would have been that of less cytolysis associated with HCV infection. However, the finding that this is not the case argues even more strongly that intrinsic NK killing is not impaired in our sample. Furthermore, we compared the two groups using samples selected for highest NK cell purity. In all cases, the results were identical to the original finding in that no difference in cytolytic function was identified between the HCV-infected and uninfected groups. Thus, we do not believe that the level of purity of the NK-enriched populations significantly affected our results.
Our data indicate that HCV-infected subjects have fewer circulating peripheral blood cytolytic (CD56dim) NK cells as well as total NK cells, consistent with results published recently by Meier et al.25 Two earlier studies did not detect a decrement in NK cell percentages in hepatitis C. The first group used CD16 alone to identify their NK cell population,11 a clear difference in methodology that could affect the results, because not all NK cells express CD16. Although the second group did not find a decrease in the percentage of total CD56+CD3− NK cells, they did detect a decrease in the percentage of CD56dim NK cells in their HCV-infected subjects,26 as we did. Different methodologies, including the use of cryopreserved cells, lack of CD45 labeling, and smaller sample sizes, might explain why some of the findings in these studies differed from our own.
The cause for the decreased NK cell frequency we found in chronic hepatitis C is unknown and may not be significant if such a decrease has no effect on overall function. However, possible explanations include sequestration of NK cells in the infected liver, increased death or turnover of NK cells, or decreased production of NK cells. Although sequestration of NK cells in the liver is an attractive hypothesis, the published data do not support this possibility. Two groups have demonstrated no difference in NK cell percentages in normal compared with HCV-infected liver,15, 27 a third group demonstrated a statistically significant positive correlation between peripheral NK cell percentage and intrahepatic NK cell percentage,28 and a fourth showed similar ratios of CD56bright:CD56dim cells in five HCV-infected subjects with paired peripheral blood and intrahepatic data. This last report also suggested that lower levels of circulating interleukin-15 may adversely affect NK cell survival.25 Further studies will be important for establishing the mechanism(s) by which circulating cytolytic NK cell frequencies are diminished in chronic hepatitis C.
Similar to the findings of Duesberg and colleagues,14 cytolysis by unfractionated PBMC was not substantially decreased in our cohort of HCV-infected subjects when compared with healthy seronegative controls (P = .29). Given the wide range of specific lysis obtained using unfractionated PBMC as effectors, it seems likely that sample size could greatly affect the conclusions drawn from the results of such studies and may help to explain the disparate findings of previously published reports.11, 13, 14 No associations were found between PBMC-specific lysis and the clinical or demographic factors reported in this study, making it difficult to identify differences in cohort composition that could have contributed to differences in the results. The correlation coefficient relating NK cell percentage in PBMC to PBMC-specific lysis was also very low at 0.33, indicating only a weak relationship between these two parameters. Thus, our data indicate a lack of an effect of lower numbers of circulating CD3−CD16/56+ or CD3−CD56dim NK cells on overall cytolysis by PBMC. These results suggest 2 main possibilities: either that decreased numbers of CD3−CD16/56+ or CD3−CD56dim cytolytic NK cells to the extent found in our HCV-positive subjects is not sufficient to significantly affect overall cytolytic function by unfractionated PBMC, or that another group of non-NK cells present in the heterogeneous PBMC populations contributes to overall cytolysis of cells devoid of HLA class I molecules. Future studies will be important for assessing these possibilities.
A recent study has suggested that NK cell function may be important in the spontaneous resolution of acute HCV infection,29 although this has not been directly tested. While we found no significant effect of chronic HCV infection on NK cytolysis, it is possible that one or more aspects of NK cell function could be impaired in acutely infected subjects. Additional detailed analyses of activated NK cells in the peripheral blood or liver could demonstrate impaired cytolytic function, as suggested by Jinushi et al.,24 although we did not find such an impairment in circulating NK cells tested freshly ex vivo. Of the circulating cytolytic CD56dim NK cells examined in our study, the activation marker CD69 was expressed at similar percentages in both HCV-infected (mean, 10.1% ± 13.8, n = 12) compared with uninfected subjects (mean, 8.8% ± 9.2, n = 14, P = .79). This similarity in CD69 expression was consistent with the similarity of intrinsic cytolytic activity between the two groups. Finally, it is possible that chronic HCV infection could adversely affect γ-interferon or other cytokine secretion by NK cells without affecting cytolytic function, and this will be explored in future studies.
In conclusion, our data indicate that circulating peripheral blood NK cells, enriched and tested freshly ex vivo from HCV-infected subjects, do not have impaired cytolytic ability compared with NK cells from HCV-negative subjects. Although we found that the frequency of NK cells was significantly decreased in HCV infection, these decreased numbers did not affect overall cytolytic killing by unfractionated PBMC in our cohort. Finally, our observation that increased NK cytolytic activity is associated with less severe liver fibrosis stage should be confirmed in additional, larger cohorts of hepatitis C–infected subjects. Such studies would clarify whether NK cell function could be protective in HCV disease pathogenesis.
The authors thank Rachel B. Life and Minjun Chung for help with coordination of the study; Michelle Gano, Megan Allison, and Nicholas Lejarcegui for technical support; Mark H. Wener and Julie McElrath for helpful discussions; and Jaime Manasala and Terri Mathisen for help with subject enrollment.
- 13Decrease in CD3-negative-CD8dim(+) and Vdelta2/Vgamma9 TcR+ peripheral blood lymphocyte counts, low perforin expression and the impairment of natural killer cell activity is associated with chronic hepatitis C virus infection. J Hepatol 2002; 37: 514-522., , , , , , et al.
- 28Simultaneous evaluation of lymphocyte subpopulations in the liver and in peripheral blood mononuclear cells of HCV-infected patients: relationship with histological lesions. Clin Exp Immunol 2002; 130: 518-525., , , , , .