CD38 antibody (Cell Marque #118R-18) was used to detect CD38 expression on formalin fixed paraffin embedded slides derived from NKTL cell lines and commercially acquired NKTL patient samples (n=68), (Pantomics, USA). The immunohistochemistry (IHC) staining was performed on Leica Bond RX autostainer. Membrane staining was scored as 0, 1+, 2+ or 3+, according to its intensity. H-score was then calculated by the formula (% [1+] x 1) + (% [2+] x 2) + (% [3+] x 3) as described previously.24 25 Cells are categorized as CD38hi with a score ≥250, CD38mid with a score ≥50 and CD38lo with any score <50.

For the multiplex immunofluorescence (MIF), a cohort of patient samples were retrieved from the archives of the Department of Pathology, National University Hospital of patients diagnosed with NKTL between 1992 and 2017. The patient study group and preparation of samples for MIF are further described in online supplemental methods.

Image acquisition and analysis were done with the Vectra V.2 multispectral automated imaging system (PerkinElmer) and in Form V.2.0 image analysis software. Nuance software (PerkinElmer) was employed to build the spectral libraries for the chromogens (Opal 520, Opal 690 and DAPI). These chromogen signature profiles were later used to spectrally unmix and quantitate CD3 and CD38 staining intensity, with appropriate regions for analysis chosen by two pathologists. For each case, four images containing at least 10 000 cells were analyzed. The absolute optical density of each chromogen indicating the intensity of each antibody was obtained for every image and normalized against respective positive cut-offs.

Animal studies were performed in accordance with IACUC policies. Five weeks old C.B-Igh-1b/GbmsTac-Prkdc^scid-Lyst^bg N7 (Beige-Scid) mice were injected subcutaneously with NKS1 resuspended in matrigel (Corning, USA). When the tumor was palpable, the mice were randomized into two treatment groups: IgG 10 mg/kg (control), daratumumab (5 mg/kg and 10 mg/kg). Treatments were administered once a week, intraperitoneal (i.p). Tumor size was measured three times a week using a caliper and derived using the formula: ½×length of tumor × (width of tumor)².

Please refer to online supplemental materials and methods for all further information.

It was first important to understand and characterize the pattern of CD38 expression in NKTL. This analysis was performed in two different cohorts of patients. In the first cohort, commercially available whole tumor tissue (n=68) from Chinese NKTL patients were processed by IHC and stained for CD38 expression. The IHC scores demonstrated that almost all NKTL samples show CD38 positivity. Seventy-eight per cent of the patients presented with a CD38 H-score higher than 100 and the median H-score of CD38 expression in the patient samples was 190 (figure 1A).

A second cohort was evaluated to further characterize CD38 expression specifically in the CD3 +NKTL subpopulation of the tumor sample. This cohort consists of samples (n=50) from patients diagnosed with extranodal NKTL and primary nodal NKTL, the latter of which has been notably classified as EBV +peripheral T-cell lymphoma (PTCL) based on the 2017 WHO recommendations.26 As they are both associated with EBV and characterized by a cytotoxic T or NK-cell infiltrate, many cases of primary nodal NKTL were previously diagnosed as extranodal NKTL in the past. However, we and other groups have shown that nodal NKTL, shows features distinct from extranodal NKTL and often present with advanced stage disease in addition to poorer survival outcomes.27 28 It was therefore of interest to include these cases in our study.

The percentage of double positive CD38+CD3+ cells was calculated from the multispectral imaging analysis (figure 1B) and the summary of these cases represented in table 1. Forty-nine out of 50 patient samples exhibited more than 1% CD38+CD3+ positive staining. Twenty-four per cent (12 cases) demonstrated ≥50% CD38+CD3+ staining, and the highest percentage of CD38+CD3+ staining detected in a patient was 84%. Seventy per cent of the patients (35 out of 50) exhibited greater than 10% CD38+CD3+ expression in the tumor sample (table 1).

Subsequently, we studied if there was any association between CD38+CD3+ staining in the tumor sample to the clinical characteristics of the patient (online supplemental table 1). Our analysis showed that the percentage of CD38+CD3+ expression in the tumor showed no significant association with (1) the subset of EBV-positive extranodal NKTL versus EBV-positive PTCL (previously primary nodal NKTL) (figure 1C), (2) patient diagnosis at stages 1–4 (figure 1D) and (3) patients of different genders (figure 1E). Kaplan-Meier analysis comparing the survival curve of patients with tumors exhibiting the top and bottom 25th percentile of CD38+CD3+ expression demonstrated a p value of 0.081, where high CD38+CD3+ expression confers a trend for shorter overall survival of the patients (figure 1F). Previously, we had published the results of a microarray analyses on these same patient tumor samples (GSE90784).27 On study, we established that the percentage of CD38+CD3+ positive cells indeed correlated significantly with the mRNA expression of CD38 in corresponding tumor samples (figure 1G). This implies that relatively meaningful correlations may be made between the CD38+CD3+ population in NKTL and CD38 mRNA expression in prospective patient samples.

As patient NKTL samples are relatively rare, the mechanism of action of Daratumumab in NKTL can be more rigorously evaluated by studying a representative panel of NKTL cell lines. A panel of 8 NKTL cell lines and 2 T-cell Lymphoma cell lines were selected and measurement of mRNA expression by qPCR showed that these cell lines express varying levels of CD38 which can be stratified into CD38hi, CD38mid and CD38lo groups (figure 2A). This was corroborated by fluorescence-activated cell sorting (FACS) analyses whereby the number of molecules of CD38 expressed per cell was determined (figure 2B). Daudi is a Burkitt’s lymphoma cell line which was originally used29 to demonstrate the in vitro efficacy of daratumumab and thus is included here as a positive control. The results show that Daudi, NKS1, and NK92 cells express more than 50 000 molecules per cell and can be categorized as CD38hi. HANK1, NKYS and SNT-8 cells express more than 7000 molecules of CD38 per cell and are designated CD38mid cell lines. The rest of the panel, KHYG, SNK6, SNK1, SNK10 and HUT78 cells express between 1000 and 3000 molecules of CD38 per cell and are considered CD38lo. We also found that healthy primary NK cells also express a low number of CD38 molecules similar to CD38lo cell lines.

Further confirmation of the classification of the panel of NKTL cell lines was determined by IHC (figure 2C) where CD38hi expressing cell lines, NKS1 and NK-92 were indeed highly positive with a H-Score of >250, NKYS and HANK1 are CD38mid with H-score of 100–200, while the CD38lo cell lines showed little CD38 positive staining (online supplemental table 2).

Daratumumab prominently triggers Fc-receptor mediated immune responses including ADCC and complement dependent cytotoxicity (CDC). Therefore, we cultured target NKTL cell lines with immune effector cells in the presence of daratumumab or control IgG. CD38hi-expressing cells displayed the highest induction of ADCC, with 40%–50% of NKS1 targeted to ADCC. 10%–20% lysis was observed in CD38mid-expressing cell lines while CD38lo-expressing cell lines did not undergo ADCC at all (figure 2D). Effector and target only controls were unaffected by Daratumumab treatment (online supplemental figure 1).

Daratumumab had been selected for its ability to trigger potent CDC. The highest induction of CDC was detected in the NK92 cell line which expresses less CD38 molecules than KMS12BM or NKS1 (figure 2E). Additionally, HANK1 but not NKYS was targeted to CDC although both expressed similar levels of CD38 molecules. This implies an additional level of regulation mediating daratumumab-induced CDC.

It was previously reported that out of 14 multiple myeloma (MM) and 19 NHL cell lines, 27 were completely resistant to daratumumab-mediated CDC assays. Susceptibility to CDC seemed to correlate with elevated levels of CD38 and lower levels of complement inhibitory proteins (CIPs) CD59 and CD55.30

Additionally, daratumumab does not show any direct cytotoxic effect on NKTL and requires the recruitment of partner immune factors to mediate its antitumor activity (figure 2F).

The level of CD38 expression is critical in mediating the overall cytotoxic effects of daratumumab. All-trans retinoic acid (ATRA) has been shown to upregulate CD38 expression and enhance daratumumab-mediated induction of ADCC and CDC in MM through the binding of the retinoic-acid-responsive element within the first intron of the CD38 gene.31 32

To study the singular effect of ATRA in Daratumumab-mediated lysis in NKTL, we selected a non-cytotoxic concentration (100 nM) of ATRA (figure 3A), which can be clinically achieved in patients.33 At this concentration, we observed a higher fold upregulation of CD38 mRNA levels and surface protein expression in CD38mid-lo vs CD38hi cell lines particularly in HuT78 (figure 3B–C). ATRA pretreatment indeed resulted in increased daratumumab-mediated ADCC (figure 3D) and CDC (figure 3E). The importance of CD38 is evident in the HuT78 cell line which was originally CD38lo but transformed to CD38hi with ATRA pretreatment and displayed correspondingly drastic increases in the functional responses of ADCC and CDC to daratumumab.

NKYS, a CD38mid cell line, did not undergo CDC even with ATRA pretreatment unlike HANK1. Again, this suggests that increasing CD38 levels alone is insufficient to overcome inhibitory factors that supress CDC in some NKTL cell lines.

Panobinostat, a pan-histone deacetylase inhibitor, is another small molecule that has demonstrated synergism with Daratumumab through the upregulation of CD38 expression in MM.34 However, Panobinostat treatment only enhanced CD38 expression in one of the three cell lines evaluated, NKYS. Additionally, this was not accompanied by an augmentation in CDC-mediated lysis of NKTL (figure 3F–G).

Cancers cells safeguard from accidental lysis primarily through the expression of membrane bound complementary regulatory proteins CD46, CD55 and CD59.35–37 The CIPs CD46 and CD55 block the complement cascade at the upstream C3 activation stage,38 39 while CD59 interferes with the formation of the membrane attack complex and usually acts additively or synergistically with CD46 and CD55.40 41

We have earlier observed that although the cell lines KMS12BM and NKS1 express higher levels of CD38 in comparison to NK92, NK92 however demonstrates the highest sensitivity to Daratumumab-induced CDC. This suggests the involvement of complement inhibitory factors which may mitigate the effectiveness of daratumumab-induced CDC in CD38hi expressing cells.

To distinguish the role of CIPs CD46, CD55 and CD59 in regulating the efficacy of Daratumumab in NKTL, we first evaluated the level of expression of CIPs in all NKTL cell lines by FACS. Indeed, KMS12BM and NKS1 exhibited significantly higher levels of CD55 and CD59 molecules as compared with NK92 (figure 4A), while CD46 expression showed less difference. This seemingly inverse correlation between expression of CIPs (CD55 and/or CD59) and CDC, is shown in figure 4B, where cell lines which expressed higher levels of CIPs displayed a lower induction of CDC across the CD38hi, mid or low categories. The relative expression levels of CD46, CD55 and CD59 in these cell lines were also confirmed by IHC (online supplemental table 3).

In order to examine the specific role each CIP may contribute to CDC, a single knockdown was performed in CD38 positive (CD38hi-mid) NKTL cell lines. As observed in figure 4A, these cell lines express different levels of each of the CIP and thus were selected for the knockdown of CD46/CD55/CD59 based on levels of CIP being expressed. NK92 and HANK1 express significantly high levels of CD46 and almost negligible levels of CD55 or CD59, thus making these cell lines excellent candidates to study the role of CD46 in daratumumab-mediated CDC. We observed that suppressing the levels of CD46 in these cell lines does not have any significant effect on the percentage of CDC induced (figure 4C). NKYS and NKS1 cell lines express highest levels of CD55 among the CD38hi-mid NKTL panel. Here, silencing CD55 in CD38mid NKYS was able to further sensitize the cells to daratumumab-induced CDC. Additionally, downregulation of CD55 was also able to significantly enhance complement-mediated lysis of NKS1 cells similar to levels of that stimulated in NK92 in figure 2E (figure 4D). Except NKS1, all the CD38hi-mid NKTL cell lines express relatively low levels of CD59. Thus, in addition to NKS1, KMS12BM was selected for CD59 knockdown due high expression levels of CD59 and CD38. It is noteworthy that the suppression of CD59 resulted in only NKS1 but not KMS12BM cells exhibiting an enhancement in CDC (figure 4E). And interestingly, while each single knockdown of the CIP CD55 or CD59 in NKS1 results in an augmentation of the CDC activity of daratumumab compared with non-targeting control, the simultaneous double knockdown of CD55 and CD59 does no result in any further additive or synergistic lysis of the cells. Additionally, despite the double knockdown of the CIPs, KMS12BM is remains resistant to daratumumab-mediated CDC (figure 4E). This led us to hypothesize that the ratio of CD38 to CIP surface expression is an overriding factor regulating the efficacy of Daratumumab activity over the absolute number of molecules of CD38 or CIP CD55 or CD59 expressed by the NKTL. A closer analysis demonstrates that although the individual ratios of CD38:CD59 expression increases by approximately 2-fold after CD59 knockdown in both cell lines, the values of the ratio are vastly different, 83.8 in NKS1 and 1.82 in KMS12BM (figure 4F). Previously, only total levels of expression of either CD38, CD55 or CD59 were studied in patients, little has been done to evaluate the ratio of these levels. Our observations led us to hypothesize that the CD38: CIP ratio may instead constitute a more accurate and critical factor determining the susceptibility of a tumor to daratumumab as opposed to the absolute expression of each regulatory protein.

As the efficacy of monoclonal antibodies such as daratumumab in NKTL rely considerably on the expression levels of CIPs, we sought to investigate the relationship between the gene expression levels of CD59 and CD55 (from GSE90784) and clinical characteristics of our patients. To our knowledge, this has not been described in recent reports in NKTL. We did not observe any significant correlation between the gene expression levels of CD55 or CD59 with the gender, versus EBV-positive extranodal NKTL versus EBV-positive PTCL (primary nodal NKTL), CD38+CD3+ expression of the patient or overall survival (online supplemental figure 2A–D). However, one interesting observation was the significant elevation of the expression of CIP CD55 and CD59 from stage 2 to late stage 4 NKTL (figure 4G,H). This may have considerable impact in decreasing the NKTL tumor CD38:CD55/CD59 ratio and corresponding effectiveness of Daratumumab therapy as the disease progresses.

ATRA did not significantly affect the expression of CD55 and CD59 while it could drastically increase the expression of CD38 (figure 5A). We had earlier hypothesized that increasing the ratio of CD38: CIPs may particularly sensitize CD38mid-lo cells to further CDC. To test this hypothesis, we increased the expression ratio of CD38: CIP by downregulating the expression of CIPs with siRNA followed by upregulating CD38 expression with ATRA.

The CD38mid-expressing cell line NKYS displays increase in CDC with single CD55 knockdown. Subsequent treatment with ATRA after CD55 silencing results in the doubling of amount of CDC stimulated (figure 5B). We observed a similar effect in NKS1 where silencing of CD55 and CD59, followed by the addition of ATRA induced almost a total lysis of all NKS1 cells (figure 5C). Again, an identical trend was observed in a CD38lo and CD55hiCD59hi cell line Hut78. The dual silencing of CD55 and CD59 in HuT78 cells was insufficient to trigger CDC. However, preincubation with ATRA overcame the inhibitory effect of the CIPs through a dramatic increase in the CD38:CIP ratio which then subsequently triggered a massive amount of Daratumumab-mediated CDC even higher than ATRA treatment alone too (figure 5D).

These experiments showed that by increasing the CD38:CIP ratio, we could strongly sensitize cells to further lysis via CDC. To statistically study this, a Spearman’s rank correlation study was performed in two ways (1) Between the absolute number of molecules of CD38, CD46, CD55 and CD59 to CDC and (2) Between the ratios of the number of molecules of CD38: CIP (CD46/CD55/CD59) to CDC. The data points are gathered based on the results obtained from the previous CDC experiments performed (figures 2–5).

Our analysis shows that, the number of molecules of CD38 positively correlates with CDC and that of CD55 inversely correlates with CDC while CD46 and CD59 do not show any significant correlation (figure 5E).

However, when we studied the correlation between the ratios of CD38:CD46, CD38:CD55 or CD38:CD59 and the extent of Daratumumab-mediated CDC, we observe that the CD38: CIP ratio consistently exhibits a significant and positive correlation with the extent of CDC lysis. In addition, the correlation coefficient values of CD38:CD55 or CD38:CD59 show robust associations closer to 1, thereby implying that the CD38: CIP ratio may more accurately predict the efficacy of daratumumab than absolute expression level of the molecules alone (figure 5F).

NKS1 is a cell line that was developed from a patient via serial transplantation in a mouse xenograft.42 As myeloma patients show evidence of a depletion in primary NK cells after daratumumab treatment, SCID Beige mice were selected to capture this clinical scenario leaving only a functional complement system to exert the antitumor effect of the antibody. Daratumumab treatment significantly inhibited the growth of NKTL in the NKS1 xenograft model and in some mice almost completely reversed tumor formation (figure 6A). This suggests that the complement system and potentially other non-immune modulatory effects of Daratumumab are sufficient to induce the potent antitumor in NKTL during the period of study. Survival studies also show that mice treated with 10 mg/kg of daratumumab has longer survival rates as compared with control mice (figure 6B).

To gain insight as to the mechanisms of resistance against daratumumab treatment, we selected three mice which developed exponential tumor growth after 32 days of last dose of 10 mg/kg daratumumab treatment (daratumumab 10 mg/kg R—average tumor size 1646 mm³ ±172) and three mice which maintained a good response (daratumumab 10 mg/kg S - average tumor size 348 mm³ ±302), and compared the mRNA expression of these tumors to three untreated (IgG 10 mg/kg – 2442 mm³ ±471) mice which were harvested on day 15. The qPCR analysis showed that there was a slight decrease in CD38 mRNA expression regardless of tumor sensitivity or resistance to daratumumab. However, there was a significant upregulation of CD55 and CD59 in ‘resistant’ tumor tissues compared with ‘sensitive’ tumor tissues and a downregulation in CD46 mRNA levels (figure 6C). This interesting response implies that CD55 and CD59 may contribute to mechanisms of resistance which arise from the tumor in response to daratumumab treatment.

L- asparaginase-based regimens feature increasingly as first-line treatment for newly diagnosed or advanced stage NKTL.43 L-asparaginase, unlike anthracyclines, cannot be effluxed by the multidrug resistant P-glycoprotein overexpressed in NKTL.44 It breaks down and removes Asparagine, a non-essential amino acid that cannot be synthesized by the cancer cell. The combination of daratumumab and L-asparaginase in vitro is able to further suppress the survival of NKTL over L-asparaginase treatment alone (figure 6D). As drugs such as panobinostat and ATRA are able to increase CD38 expression, we studied the effect of L-Asparaginase on CD38, CD55 and CD59 surface expression levels. There was no significant change in CD38 or CD59 expression, however, a 18%±1.6% and 48%±6.3 decrease in CD55 expression was detected at IC25 and IC50 concentration of L-asparaginase, respectively (figure 6E). Subsequently, the combination treatment of daratumumab and L-asparaginase was investigated in vivo. We found that the addition of daratumumab at 2 mg/kg to L-asparaginase was able to significantly further inhibit tumor growth over L-asparaginase alone, congruent with what was observed in vitro (figure 6F). Interestingly, daratumumab treatment alone seemed more effective than the combination of L-asparaginase and daratumumab (data not shown).