Adaptive Subsets Limit the Anti-Tumoral NK-Cell Activity in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) is a global health burden with increasing incidence, poor prognosis and limited therapeutic options. Natural killer (NK) cells exhibit potent anti-tumoral activity and therefore represent potential targets for immunotherapeutic approaches in HCC treatment. However, the anti-tumoral activity of NK cells in HCC associated with different etiologies, and the impact of the heterogeneous NK cell subset, e.g., adaptive and conventional subsets, are not understood in detail. By comparatively analyzing the NK-cell repertoire in 60 HCC patients, 33 liver cirrhosis patients and 36 healthy donors (HD), we show in this study that the NK-cell repertoire is linked to HCC etiology, with increased frequencies of adaptive NK cells in Hepatitis B virus (HBV)-associated HCC. Adaptive NK cells exhibited limited anti-tumoral activity toward liver cancer cells; however, this was not a result of a specific NK-cell impairment in HCC but rather represented an intrinsic feature, since the characteristics of circulating and intra-tumoral adaptive NK cells were conserved between HD, HCC and liver cirrhosis patients. Hence, the expansion of adaptive NK cells with reduced anti-tumoral activity, detectable in HBV-associated HCC, may have implications for tumor surveillance and therapy.


Introduction
Hepatocellular carcinoma (HCC) is the most common form of primary liver cancer in adults and represents a major health problem due to its increasing incidence world-tissue-residency markers. These observations suggest that adaptive NK cells intrinsically harbor reduced anti-tumoral activity against liver cancer cells that is not further impaired in HCC. Taken together, the presence of adaptive CD56 dim NK cells may therefore contribute to the impaired anti-tumoral NK-cell activity in HCC, especially in association with HBV infection.

Study Cohort
A cohort of 60 HCC patients ( Table 1, Table S1), 36 healthy donors ( Table 1, Table S2), 33 patients with liver cirrhosis ( Table 1, Table S3) and 28 patients with chronic HBV infection (Table S4) was recruited at the Department of Medicine II of the University Hospital, Freiburg, Germany. Written informed consent was obtained in all cases and the study was conducted according to the Declaration of Helsinki (1975), federal guidelines and local ethics committee regulations (Albert-Ludwigs-University, Freiburg, Germany, approvals 474/14 and 152/17). Peripheral blood mononuclear cells (PBMCs), HCC tumor tissue and adjacent non-tumor liver tissue samples (Table S5) were obtained from the hepatologic and gastroenterologic biobank of the University Hospital, Freiburg (HBUF, banked since 2010) and non-tumoral liver samples (obtained from liver resections due to colon canceroriginated metastasis) from the Karolinska University Hospital, Stockholm (Table S6), and analyzed retrospectively. The donors were not analyzed for genetic polymorphisms in the KLRC2 gene. Due to the restricted availability of patient material, complete analyses including all parameters were not possible for all patients. A detailed list of performed analyses for each donor is included in Tables S1-S6.

Isolation of Lymphocytes from Tissue Samples
Lymphocytes were isolated from tissue samples by density-gradient centrifugation after mechanical dissociation in complete RPMI culture medium (RPMI 1640 supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin and 1.5% 1M HEPES (all Thermo Fisher, Waltham, MA, USA)).

HCMV Status
HCMV serostatus was determined via HCMV-IgG chemiluminescence immunoassay (DiaSorin LIAISON, Saluggia, VC, Italy) by the Department of Virology, University of Freiburg, for all donors recruited at the University Hospital Freiburg or as previously Cells 2021, 10, 1369 4 of 16 described by Marquardt et al. [37] for the donors of non-tumoral liver samples from the Karolinska University Hospital, Stockholm, where no plasma/PBMC samples (but mononuclear liver cells) were available. Briefly, 10 6 mononuclear liver cells were stimulated with 1 µg/mL CMV pp65 overlapping peptides (JPT Peptide Technologies, Berlin, BE, Germany) in the presence of Brefeldin A (GolgiPlug (final concentration 5 µL/mL, BD Biosciences, Franklin Lakes, NJ, USA) and Monensin (GolgiStop (final concentration 3.25 µL/mL, BD Biosciences, Franklin Lakes, NJ, USA) overnight (o/n). After 16 h, cytokine production was assessed by flow cytometry [37]. The HCMV seronegative (HCMV) individuals are only included in Figure S1.

Cell Lines
The Huh7 cell line was kindly provided by Volker Lohmann (University Hospital Heidelberg, Germany). The HepG2 cell line was kindly provided by Michael Nassal (University Hospital Freiburg, Germany). The K562 cell line was kindly provided by Hanspeter Pircher (University Hospital Freiburg, Germany). All cell lines were tested for mycoplasma.

Assessment of NK-Cell Function
Degranulation and cytokine production of NK cell subsets were determined after cytokine stimulation o/n with IL-12 (10 ng/mL, Sigma Aldrich, St. Louis, MO, USA) and IL-18 (5 ng/mL, MBL, Woburn, MA, USA), after CD16-crosslinking or after co-culture with the cell lines K562, Huh7 and HepG2 and autologous CD8 + T cells (pre-activated with ImmunoCult TM Human CD3/CD28 T Cell Activator (Stemcell Technologies, Vancouver, BC, Canada) according to the manufacturer's instructions). More detailed information is provided in the supplementary information.

Multiparametric Flow Cytometry
A detailed list of antibodies and reagents used for flow cytometry analysis is included in the supplementary information.

Profiles of Adaptive NK Cells Are Conserved in HCC
CD56 dim and CD56 bright NK-cell frequencies within the circulating NK-cell population (defined as CD3 − CD56 + ) were similar in patients with HCC compared to patients with liver cirrhosis and HD (irrespective of HCMV serostatus) with a clear dominance of CD56 dim NK cells (Figure S1A-C, Table 1 and Tables S1-S3). A minor but significant reduction of CD56 dim NK cells was detectable in cirrhotic patients compared to HD (Figure S1B,D). Frequencies of CD56 dim and CD56 bright NK-cell populations were not affected by HCMV infection ( Figure S1D,E), which was apparent in the majority of our HCC cohort ( Figure S1F).
Next, we analyzed the presence of FcεRIγ − , Syk − , PLZF − , Helios − , NKG2C + and CD57 + ( Figure 1A-F) cells within CD56 dim NK-cell populations in HCMV + HCC patients, to assess the HCMV-associated appearance of adaptive NK cells. HCMV-driven CD56 dim NK-cell diversification was also evident in HCMV + HCC patients; however, no significant difference in the frequencies was detectable in comparison to HCMV + patients with liver cirrhosis and HCMV + HD ( Figure 1A-F, marked in grey).
Next, we analyzed the presence of FcεRIγ -, Syk -, PLZF -, Helios − , NKG2C + and CD57 + ( Figure 1A-F) cells within CD56 dim NK-cell populations in HCMV + HCC patients, to assess the HCMV-associated appearance of adaptive NK cells. HCMV-driven CD56 dim NK-cell diversification was also evident in HCMV + HCC patients; however, no significant difference in the frequencies was detectable in comparison to HCMV + patients with liver cirrhosis and HCMV + HD ( Figure 1A-F, marked in grey).  Each dot represents an HCMV + individual. Bars indicate median with IQR. Statistical significance for cohort comparison was assessed by using the Kruskal-Wallis test (A-F) and for subset comparison mixed-effect analysis (G,H). Patients with less than 10% adaptive FcεRIγ − CD56 dim NK cells were excluded (G,H). HCC: patients with hepatocellular carcinoma, HD: healthy donors, cirrhosis: patients with liver cirrhosis, HCMV: human cytomegalovirus. *: p < 0.05; ****: p < 0.0001. Next, we assessed whether the combined phenotypic profile of adaptive NK cells is also conserved in HCMV + HCC patients. For this, the presence of adaptive CD56 dim NK cells was defined by more than 10% FcεRIγ − cells among CD56 dim NK-cell populations ( Figure S1G), and we then comparatively analyzed the expression of PLZF and NKG2C ( Figure 1G,H) in FcεRIγ − CD56 dim NK cells. Lower frequencies of adaptive FcεRIγ − CD56 dim NK cells from HCC patients compared to conventional FcεRIγ + CD56 dim NK cells expressed PLZF ( Figure 1G), whereas higher frequencies expressed NKG2C ( Figure 1H), similar to HCMV + control cohorts. Additional phenotypic analysis of CD2, CD7, Siglec-7, CX3CR1 and CXCR3 expression ( Figure S2) also showed a conserved adaptive NK cell profile in HCMV + HCC patients. T-SNE analysis of adaptive FcεRIγ − CD56 dim NK cells revealed that the majority of the cells intermingle between all three cohorts ( Figure 1I). Thus, HCMV-driven adaptive NK-cell diversification is conserved in HCMV + HCC, compared to HCMV + patients with liver cirrhosis or HCMV + HD.

Adaptive NK Cells in Tumor and Non-Tumor Liver Tissue Exhibit Conserved Profiles
As the human liver is enriched with NK cells, we analyzed non-tumor liver tissue (Tables S5 and S6) and HCC tumor samples (Table S5) Figure S3D) or CD57 ( Figure S3E) in non-tumoral liver tissue and HCC tumor tissue. To assess whether adaptive NK cells isolated from the tumor are phenotypically different, we compared the expression of PLZF, Helios, NKG2C and CD57 in FcεRIγ − CD56 dim NK cells from matched blood, adjacent non-tumor liver tissues, and HCC tumor tissues obtained from HCMV + HCC patients with more than 10% FcεRIγ − adaptive CD56 dim NK cells in their blood. All analyzed markers were equally expressed in FcεRIγ − adaptive CD56 dim NK cells from the three matched compartments of the respective donors, suggesting conserved profiles even within HCC tumors (Figure 2A-D). This finding was further corroborated by t-SNE analysis of FcεRIγ − adaptive CD56 dim NK, showing no clear separation between the cells of the different compartments ( Figure 2E). To further address whether FcεRIγ − CD56 dim NK cells are circulating or liver-infiltrating cells, or rather, represent tissue-resident NK cells, we analyzed the expression of the tissue-resident marker molecules CD69, CXCR6 and CD49a on CD56 dim NK-cell subpopulations and CD56 bright NK cells. FcεRIγ − adaptive CD56 dim NK cells, isolated from liver tissue irrespective of non-tumor or tumor origin, displayed a low expression of these tissue-resident markers ( Figure 2F-H), whereas CD56 bright NK cells obtained from liver tissue expressed higher levels of CD69, CXCR6 and CD49a ( Figure S4). These findings were confirmed by cross-sectionally analyzing blood, non-tumor liver and HCC tumor tissues ( Figure S5). Hence, adaptive CD56 dim NK cells found in HCC lesions are most probably circulating or tumor-infiltrating lymphocytes.

Limited Anti-Tumoral Activity of FcεRIγ -Adaptive NK Cells
To assess the anti-tumoral capacity, we performed co-culture experiments of circulating FcεRIγ-based NK-cell populations of HCMV + HCC patients (>10% FcεRIγadaptive CD56 dim NK cells) with the hepatoma cell lines Huh7 and HepG2. CD107a expression, as a surrogate marker for degranulation ( Figure 3A,B), and MIP-1ß ( Figure  S6A,B) production after stimulation with Huh7 or HepG2 cells, were significantly reduced in FcεRIγadaptive NK cells compared to conventional FcεRIγ + NK cells, with the same , adjacent non-tumoral liver (n = 4 for (A,C), n = 6 for (B), n = 3 for (D)) and HCC tissue (n equal to blood) of HCMV + HCC patients. T-SNE analysis of concatenated flow cytometry data obtained from FcεRIγ − CD56 dim NK cells in blood (n = 5), adjacent non-tumoral liver (n = 4) and HCC tissue (n = 5) of HCMV + HCC patients (E). Expression levels of NKG2C and PLZF are depicted in the t-SNE plots (left). CD69 (F), CXCR6 (G) and CD49a (H) expression of FcεRIγ − CD56 dim NK cells in blood (n = 3 for (F,H), n = 4 for (G)), non-tumor (n = 2) and HCC tissue (n equal to blood) of HCMV + HCC patients. Each point represents a single HCMV + patient, and lines connect the samples from one patient. Bars indicate median with IQR. Statistical significance was tested using the Kruskal-Wallis test (A-D) and mixed-effect analysis (F-H). HCC: hepatocellular carcinoma.

Limited Anti-Tumoral Activity of FcεRIγ − Adaptive NK Cells
To assess the anti-tumoral capacity, we performed co-culture experiments of circulating FcεRIγ-based NK-cell populations of HCMV + HCC patients (>10% FcεRIγ − adaptive CD56 dim NK cells) with the hepatoma cell lines Huh7 and HepG2. CD107a expression, as a surrogate marker for degranulation ( Figure 3A,B), and MIP-1ß ( Figure S6A,B) production after stimulation with Huh7 or HepG2 cells, were significantly reduced in FcεRIγ − adaptive NK cells compared to conventional FcεRIγ + NK cells, with the same trends seen for IFN-γ ( Figure S7A,B). This observation indicates reduced anti-tumoral activity against liver cancer cells. We also observed the same effect after stimulation with the leukemia cell line K562 ( Figure 3C, Figures S6C and S7C).
Cells 2021, 10 8 of 16 trends seen for IFN-γ ( Figure S7A,B). This observation indicates reduced anti-tumoral activity against liver cancer cells. We also observed the same effect after stimulation with the leukemia cell line K562 ( Figure 3C, Figures S6C and S7C). In addition, FcεRIγadaptive NK cells exhibited reduced IL-12 + IL-18 responsiveness mirrored by cytokine secretion (Figures S6D and S7D) and also by diminished degranulation ( Figure 3D).
However, CD16-mediated functionality, tested by CD107a and cytokine expression after cross-linking, was comparable between adaptive FcεRIγand conventional FcεRIγ + CD56 dim NK cells in HCMV + HCC patients ( Figure 3E, Figures S6E and S7E). Finally, to analyze the immunoregulatory potential of FcεRIγadaptive NK cells compared to FcεRIγ + conventional NK cells toward CD8 + T cells, we co-cultivated FcεRIγ-based NKcell populations from HCC patients with autologous, activated CD8 + T cells. Degranulation ( Figure 3F) by FcεRIγ -CD56 dim NK cells was significantly reduced compared to FcεRIγ + CD56 dim NK cells, with a similar trend with respect to cytokine production ( Figures S6F and S7F), suggesting a reduced immunoregulatory activity of adaptive NK cells toward CD8 + T cells. Representative negative controls for all assays are shown in Figure S8. In summary, while reduced immunomodulatory function may potentially lead to indirect support of the CD8 + T cell-based anti-tumoral response in HCC, the direct anti-tumoral activity mediated by adaptive NK cells is limited. Patients with less than 10% adaptive FcεRIγ − CD56 dim NK cells were excluded. The following statistical analyses were performed: two-tailed Wilcoxon Test. *: p < 0.05; **: p < 0.01.
However, CD16-mediated functionality, tested by CD107a and cytokine expression after cross-linking, was comparable between adaptive FcεRIγ − and conventional FcεRIγ + CD56 dim NK cells in HCMV + HCC patients ( Figure 3E, Figures S6E and S7E). Finally, to analyze the immunoregulatory potential of FcεRIγ − adaptive NK cells compared to FcεRIγ + conventional NK cells toward CD8 + T cells, we co-cultivated FcεRIγ-based NKcell populations from HCC patients with autologous, activated CD8 + T cells. Degranulation ( Figure 3F) by FcεRIγ − CD56 dim NK cells was significantly reduced compared to FcεRIγ + CD56 dim NK cells, with a similar trend with respect to cytokine production ( Figures S6F  and S7F), suggesting a reduced immunoregulatory activity of adaptive NK cells toward CD8 + T cells. Representative negative controls for all assays are shown in Figure S8. In summary, while reduced immunomodulatory function may potentially lead to indirect support of the CD8 + T cell-based anti-tumoral response in HCC, the direct anti-tumoral activity mediated by adaptive NK cells is limited.

Adaptive NK Cells Are Not Specifically Impaired in HCC
Next, we investigated whether the HCC-associated adaptive NK cells exhibit characteristics of specific impairment. First, we analyzed the expression of NKG2D on FcεRIγ-based CD56 dim NK-cell populations, relevant in tumor-associated NK activation. We did not detect differences in NKG2D expression, neither on FcεRIγ − adaptive nor on FcεRIγ + conventional NK cells from HCC patients, compared to control groups ( Figure 4A). NKG2D expression on adaptive NK cells did not differ between matched samples obtained from blood, adjacent non-tumor liver and HCC tumor tissue ( Figure 4B). NKp46 and NKp30 expressions were also comparable on FcεRIγ − adaptive NK cells obtained from HCC patients and the control groups ( Figure S9A,B).

Adaptive NK Cells Are Not Specifically Impaired in HCC
Next, we investigated whether the HCC-associated adaptive NK cells exhibit characteristics of specific impairment. First, we analyzed the expression of NKG2D on FcεRIγ-based CD56 dim NK-cell populations, relevant in tumor-associated NK activation. We did not detect differences in NKG2D expression, neither on FcεRIγadaptive nor on FcεRIγ + conventional NK cells from HCC patients, compared to control groups ( Figure  4A). NKG2D expression on adaptive NK cells did not differ between matched samples obtained from blood, adjacent non-tumor liver and HCC tumor tissue ( Figure 4B). NKp46 and NKp30 expressions were also comparable on FcεRIγadaptive NK cells obtained from HCC patients and the control groups ( Figure S9A,B).  (D, F)) and HCC lesion (n equal to blood) (right). Each dot represents an HCMV + individual with more than 10% adaptive FcεRIγ -CD56 dim NK cells, and the line connects the samples from one patient. Bars indicate median with IQR. Statistical significance was tested by using mixed-effect analysis (A,C,E) and the Kruskal-Wallis test (B,D,F). HCC: patients with hepatocellular carcinoma, HD: healthy donors, cirrhosis: patients with liver cirrhosis. *: p < 0.05; ***: p < 0.001.
We also tested for checkpoint molecule expression. While TIGIT expression on FcεRIγ -CD56 dim NK cells was comparable to FcεRIγ + CD56 dim NK cells ( Figure 4C), a higher fraction of FcεRIγadaptive CD56 dim NK cells expressed PD1 ( Figure 4E). PD1 expression was significantly increased on FcεRIγadaptive CD56 dim NK cells from HCMV + HCC patients, compared to HCMV + HD. Expression of both TIGIT and PD1 did not differ between matched blood, adjacent non-tumor liver tissue and tumor tissue ( Figure 4D,F).  (D, F)) and HCC lesion (n equal to blood) (right). Each dot represents an HCMV + individual with more than 10% adaptive FcεRIγ − CD56 dim NK cells, and the line connects the samples from one patient. Bars indicate median with IQR. Statistical significance was tested by using mixed-effect analysis (A,C,E) and the Kruskal-Wallis test (B,D,F). HCC: patients with hepatocellular carcinoma, HD: healthy donors, cirrhosis: patients with liver cirrhosis. *: p < 0.05; ***: p < 0.001.
We also tested for checkpoint molecule expression. While TIGIT expression on FcεRIγ − CD56 dim NK cells was comparable to FcεRIγ + CD56 dim NK cells ( Figure 4C), a higher fraction of FcεRIγ − adaptive CD56 dim NK cells expressed PD1 ( Figure 4E). PD1 expression was significantly increased on FcεRIγ − adaptive CD56 dim NK cells from HCMV + HCC patients, compared to HCMV + HD. Expression of both TIGIT and PD1 did not differ between matched blood, adjacent non-tumor liver tissue and tumor tissue ( Figure 4D,F). Additionally, no significant differences in NKG2A expression were detected between CD56 dim NK cell subpopulations in HCMV + HCC and control groups ( Figure S9C). By comparing the functional capacities, we did not observe differences in the degranulation ( Figure 5) and cytokine secretion ( Figure S10) of FcεRIγ − adaptive NK cells obtained from HCMV + HCC patients compared to HCMV + control cohorts, neither after stimulation with tumor cells (Figure 5A-C, Figure S10A-F), nor with cytokines ( Figure 5D, Figure S10G,H), nor after CD16 crosslink ( Figure 5E, Figure S10K,L), nor after co-cultivation with activated autologous CD8 + T cells ( Figure 5F, Figure S10I,J). Thus, we did not observe functional impairment of adaptive and conventional CD56 dim NK cells in HCMV + HCC.
Cells 2021, 10 10 of 16 Additionally, no significant differences in NKG2A expression were detected between CD56 dim NK cell subpopulations in HCMV + HCC and control groups ( Figure S9C). By comparing the functional capacities, we did not observe differences in the degranulation ( Figure 5) and cytokine secretion ( Figure S10) of FcεRIγadaptive NK cells obtained from HCMV + HCC patients compared to HCMV + control cohorts, neither after stimulation with tumor cells (Figure 5A-C, Figure S10 A-F), nor with cytokines ( Figure 5D, Figure S10G,H), nor after CD16 crosslink ( Figure 5E, Figure S10K,L), nor after co-cultivation with activated autologous CD8 + T cells ( Figure 5F, Figure S10I,J). Thus, we did not observe functional impairment of adaptive and conventional CD56 dim NK cells in HCMV + HCC.

Adaptive NK Cells Are Expanded in HCMV + HBV-Associated HCC Patients
Lastly, we assessed whether the frequencies of FcεRIγ − CD56 dim NK cells are associated with clinical parameters. We did not detect different frequencies of FcεRIγ − CD56 dim NK cells dependent on the Child-Pugh score, reflecting the severity of cirrhosis and liver function ( Figure 6A), and the BCLC score ( Figure 6B) used to stage HCC. Similarly, no significant correlation of FcεRIγ − CD56 dim NK-cell frequencies with the tumor marker AFP could be observed ( Figure S11, Table S1). We also did not observe changes in the frequencies of FcεRIγ + or FcεRIγ − CD56 dim NK cells induced by ablative therapy ( Figure 6C). Cells 2021, 10 11 of 16 the frequencies of FcεRIγ + or FcεRIγ -CD56 dim NK cells induced by ablative therapy ( Figure  6C). Next, we investigated the impact of the underlying HCC etiology on the presence of adaptive NK cells in HCMV + HCC patients. Adaptive NK-cell frequencies were significantly increased in HBV-associated HCMV + HCC patients compared to HCMV + HD but not in HCMV + HCV-or ASH-associated HCC ( Figure 6D). Thus, HBV infection also alters the NK-cell repertoire in HCMV + HCC patients. Notably, Syk ( Figure 6E), PLZF ( Figure 6F) and Helios ( Figure 6G) expression, cytokine stimulation (Figure S12A), and ADCC ( Figure S12B) were similar in FcεRIγadaptive NK cells from patients with chronic HBV infection (non-cirrhotic) and with HBV-associated HCC (Table S1 and Table S4). Hence, the etiology is a relevant determinant of the NK-cell repertoire in HCC patients.

Discussion
Our comparative analysis of the NK cell repertoire in HCC and liver cirrhosis patients and in HD revealed that the presence of adaptive CD56 dim NK cells, e.g., defined by FcεRIγ -, in HCC and pre-malignant liver cirrhosis was clearly associated with HCMV infection, as previously reported in other patient cohorts including cohorts affected by viral infections or leukemia [22,[31][32][33][34]36,[38][39][40]. Since HCMV infection is frequent [41,42], as also reflected by our cohorts of patients suffering from HCC and cirrhosis, the HCMVassociated emergence of adaptive CD56 dim NK cells is an important parameter to consider in NK cell-mediated cancer immunosurveillance with potential clinical impact. However, Next, we investigated the impact of the underlying HCC etiology on the presence of adaptive NK cells in HCMV + HCC patients. Adaptive NK-cell frequencies were significantly increased in HBV-associated HCMV + HCC patients compared to HCMV + HD but not in HCMV + HCV-or ASH-associated HCC ( Figure 6D). Thus, HBV infection also alters the NK-cell repertoire in HCMV + HCC patients. Notably, Syk ( Figure 6E), PLZF ( Figure 6F) and Helios ( Figure 6G) expression, cytokine stimulation (Figure S12A), and ADCC ( Figure S12B) were similar in FcεRIγ − adaptive NK cells from patients with chronic HBV infection (non-cirrhotic) and with HBV-associated HCC (Tables S1 and S4). Hence, the etiology is a relevant determinant of the NK-cell repertoire in HCC patients.

Discussion
Our comparative analysis of the NK cell repertoire in HCC and liver cirrhosis patients and in HD revealed that the presence of adaptive CD56 dim NK cells, e.g., defined by FcεRIγ − , in HCC and pre-malignant liver cirrhosis was clearly associated with HCMV infection, as previously reported in other patient cohorts including cohorts affected by viral infections or leukemia [22,[31][32][33][34]36,[38][39][40]. Since HCMV infection is frequent [41,42], as also reflected by our cohorts of patients suffering from HCC and cirrhosis, the HCMVassociated emergence of adaptive CD56 dim NK cells is an important parameter to consider in NK cell-mediated cancer immunosurveillance with potential clinical impact. However, a sound correlation analysis between HCMV serostatus and patient outcome was not possible in our study, due to the cohort size.
Besides this link to HCMV, we also detected increased frequencies of adaptive CD56 dim NK cells in HBV-associated HCC compared to the other analyzed etiologies, such as HCV and ASH. Furthermore, no other tested clinical parameters, including liver function or tumor stage, correlated with the frequencies of adaptive CD56 dim NK cells in our cohort, arguing against a general HCMV reactivation in HCC lesions accompanied by an expansion of adaptive CD56 dim NK cells. The underlying mechanisms of the pronounced expansion of adaptive CD56 dim NK cells seem rather to be HBV-related. This is in line with a recent report from our group, showing increased frequencies of FcεRIγ − adaptive CD56 dim NK cells in patients chronically infected with HBV, compared to HCV and HD [36]. However, further studies are required to uncover these HBV-related mechanisms. Possible mechanisms may include the presence of anti-HCMV/anti-HBV antibodies, and selection for specific HCMV strains, cytokines and also genetic factors [36]. Interestingly, Wijaya et al. also demonstrated expansion of KLRG1 + NK cells in the context of chronic HCV and HBV infection without an association with HCMV infection [43][44][45]. However, the lack of FcεRIγ − adaptive NK cells in nearly all HCMV negative patients in our cohort argues against the induction of this adaptive NK cell subset by HBV alone.
FcεRIγ − adaptive CD56 dim NK cells are themselves also a heterogeneous population, as exemplified by previously described FcεRIγ − /NKG2C or FcεRIγ − /Helios co-expression analyses [36,46]. Our results show that the phenotypic and functional characteristics of FcεRIγ − adaptive CD56 dim NK cells are largely conserved in HCC and control cohorts. Clearly, further high-dimensional analyses, e.g., single-cell RNA sequencing, are required to fully elucidate the entire heterogeneity of HCC-associated adaptive CD56 dim NK cells, especially with respect to different etiologies. A high degree of conserved characteristics and similar frequencies of adaptive NK cells were also detectable within tumor and nontumor liver tissue and blood. These observations indicate that adaptive CD56 dim NK cells whose presence in the liver has also previously been described [46], are, rather, circulating lymphocytes instead of tissue-resident subsets. This assumption is also in line with our findings that adaptive CD56 dim NK cells exhibited only a low expression of tissue-resident markers [20,28,29,37] like CXCR6, CD69 and CD49a. Hence, adaptive CD56 dim NK cells found within HCC tissue likely represent circulating/tumor-infiltrating lymphocytes that can be monitored and targeted via the blood. This observation may have important relevance for translational applications, such as for immunotherapeutic interventions. However, additional studies are required to further clarify the circulating/infiltrating nature of adaptive CD56 dim NK cells within the liver, e.g., by the analysis of the KIR repertoire, to exclude that adaptive NK cells are locally expanded.
Compared to the conventional FcεRIγ + CD56 dim NK-cell subset, adaptive FcεRIγ − CD56 dim NK cells displayed reduced direct anti-tumoral activity. This limited anti-tumoral activity was not only evident against the leukemia cell lines K562 or Raji, as reported previously in naturally occurring adaptive NK cells and FcεRIγ − NK cells produced in vitro by gene editing [35,46,47], but also as unraveled by this study against liver cancer cells reflecting solid tumors. This observation may indicate an inherently reduced adaptive NK cell response triggered, e.g., by low MHCI expression. This limited response of adaptive CD56 dim NK cells toward MHCI low target cells may be detrimental in cancer immunosurveillance, reflecting a possible successful cancer evasion strategy. However, it also potentially represents a mechanism of tissue protection, especially in the liver, since hepatocytes express low levels of MHCI [48]. In line with this, Oh et al. reported that the presence of adaptive NK cells was associated with less severe liver damage [49]. A general underlying mechanism is further supported by the observation that we could not detect differences in the anti-tumoral activity of patients with different etiologies of HCC and that adaptive CD56 dim NK cells from HCC patients and HD exhibited a similarly limited anti-tumoral activity. Thus, it is probably not a specific impairment in HCC, but rather an intrinsic and stable characteristic of adaptive CD56 dim NK cells. This also fits our data regarding an unaltered expression of NKG2D, TIGIT and NKG2A on adaptive CD56 dim NK cells obtained from HCC patients compared to HD, which have all been reported to be relevant in mediating the anti-tumoral NK cell response [11,12,50,51]. In contrast to adaptive CD56 dim NK cells, the HCC-associated dysfunction observed within the liver-resident CD56 bright NK-cell population is at least in part directly induced by HCC in a contact-dependent manner [23].
Expression of PD1, a checkpoint inhibitor molecule, was also described on CD56 dim NK cells of cancer patients and healthy controls [52,53]. In particular, Liu et al. reported elevated PD1 expression on NK cells obtained from HCC patients [53]. Furthermore, in another study analyzing HCMV seropositive individuals, PD1 + NK cells exhibited a higher expression of CD57 [52], a maturation marker that is also highly expressed on adaptive CD56 dim NK cells [31]. In line with these reports, we detected an increased expression of the checkpoint inhibitor PD1 on adaptive cells compared to conventional CD56 dim NK cells that was significantly higher in HCC patients compared to HD. However, the overall frequency of PD1 expressing CD56 dim NK cells was extremely low in the blood, adjacent non-tumor, and tumor liver tissue of HCC patients. Of note, tumor-infiltrating NK cells in mice and colon cancer patients also showed low expression rates of PD1 [50], questioning a general direct role of PD1 as a checkpoint inhibitor of NK cells. Hence, we assume a minor role for this checkpoint inhibitor molecule in adaptive CD56 dim NK cell-mediated immunosurveillance in HCC [54].
In addition, as in liver cirrhosis patients and HD, adaptive CD56 dim NK cells obtained from HCC patients displayed reduced cytokine responsiveness, as described previously in other settings [31][32][33][34][35][36], and decreased immunoregulatory function toward activated CD8 + T cells. The decreased immunoregulatory function of adaptive CD56 dim NK cells may result from the low NCR expression. This is in line with previous reports in the LCMV mouse system [55][56][57], showing, for example, that an altered regulation of CD8 + T cells by NK cells in FcεRIγ -/mice resulted in better survival. However, it remains unclear whether the FcεRIγ -/-NK cells in this mouse model are comparable to their human counterparts since, in the murine system, adaptive/memory-like NK cells are defined by other features such as Ly49H expression. In summary, in the context of cancer immunosurveillance, the conserved functional features of FcεRIγ − adaptive NK cells may favor tumor-specific immunity by T cells at the expense of a broad, less specific NK cell-mediated anti-tumoral defense induced by altered-self, missing-self and bystander/cytokine-mediated mechanisms, also in HCC. The observation that adaptive CD56 dim NK cells are more frequent in HBV-associated HCC therefore suggests that HCC etiology is linked to differences in cancer immunosurveillance, a fact that needs to be considered in the design and application of immunotherapies in HCC.
Supplementary Materials: The supplementary information includes supplementary methods and figures, as well as detailed study cohort characteristics. The following are available online at https://www.mdpi.com/article/10.3390/cells10061369/s1, Supplemental Figure S1. Increased frequency of FcεRIγ-adaptive NK cells in HCMV+ HCC patients. Figure S2. Adaptive NK-cell profile is comparable between HCC patients and control cohorts. Figure S3. Adaptive NK-cell repertoire in blood and tissue of HCC patients. Figure S4. CD56 dim NK cells do not express tissue residency markers in matched tumor/non-tumor HCC tissue. Figure S5. CD56 dim NK cells do not express tissue residency markers in liver tissue. Figure S6. MIP-1ß production by FcεRIγadaptive and FcεRIγ+ conventional NK cells in HCC patients. Figure S7. IFNγ production by FcεRIγ-adaptive and FcεRIγ+ conventional NK cells in HCC patients. Figure S8. Negative controls for functional NK-cell studies. Figure S9. NCR and NKG2A expression of adaptive NK cells in HCC patients. Figure S10. Conserved cytokine production of CD56dim NK-cell subpopulations. Figure S11. Frequency of FcεRIγ-adaptive NK cells do not correlate with AFP values. Figure S12. Conserved functional profile of FcεRIγ-adaptive CD56dim NK cells between HCMV+ HBV and HCMV+ HBV-associated HCC patients. Table S1. Study cohort of HCC patients (PBMCs). Table S2. Study cohort of healthy donors (HD). Table S3. Study cohort of patients with liver cirrhosis. Table S4. Study cohort of patients with chronic HBV infection (HBV). Table S5. Study cohort of liver and tumor samples from the hepatologic and gastroenterologic biobank of the University hospital Freiburg. Table S6. Study cohort of liver samples obtained from the Karolinska Institute, Sweden.