Prognostic impact of natural killer cell count in ......May 31, 2019 · Robert Marcus,5 6Laurie H Sehn, 7Umberto Vitolo, Alexandra Bazeos,8 Valentin Goede,9 Harald Zeuner, 3 Andrea

1

Prognostic impact of natural killer cell count in follicular lymphoma and

diffuse large B-cell lymphoma patients treated with immunochemotherapy

Magdalena Klanova,1–3* Mikkel Z Oestergaard,3* Marek Trněný,1 Wolfgang Hiddemann,4

Robert Marcus,5 Laurie H Sehn,6 Umberto Vitolo,7 Alexandra Bazeos,8 Valentin Goede,9

Harald Zeuner,3 Andrea Knapp,3 Deniz Sahin,3 Nathalie Danesi,3 Christopher R Bolen,10

Andres Cardona,3 Christian Klein,11 Jeffrey M Venstrom,10 Tina Nielsen,3 and Günter

Fingerle-Rowson3

*equal contribution

Affiliations

1Charles University General Hospital, Prague, Czech Republic; 2Institute of Pathological

Physiology, Charles University, Prague, Czech Republic; 3F. Hoffmann-La Roche Ltd, Basel,

Switzerland; 4University of Munich, Munich, Germany; 5Kings College Hospital, London,

United Kingdom; 6Centre for Lymphoid Cancer, British Columbia Cancer Agency and the

University of British Columbia, Vancouver, Canada; 7A.O.U. Citta' Della Salute e della

Scienza, S.C. Ematologia, Turin, Italy; 8Imperial College London, London, United Kingdom;

9Center of Integrated Oncology Cologne-Bonn, University Hospital Cologne, Cologne,

Germany; 10Genentech Inc., South San Francisco, USA; 11Roche Innovation Center Zurich,

Schlieren, Switzerland

Corresponding Author:

Dr Magdalena Klanova, 1st Department of Medicine, Charles University General Hospital, U

Nemocnice 2, Prague 2, 12808, Czech Republic

Email: [email protected]; Tel: +420 774 097 744

Dr Mikkel Z Oestergaard, Oncology Biomarker Development, F. Hoffman-La Roche,

Grenzacherstrasse 124, 4070 Basel, Switzerland

Email: [email protected]; Tel: +41 61 688 32 15

Cancer Research. on November 10, 2020. © 2019 American Association forclincancerres.aacrjournals.org Downloaded from

Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Author Manuscript Published OnlineFirst on May 3, 2019; DOI: 10.1158/1078-0432.CCR-18-3270

mailto:[email protected]


http://clincancerres.aacrjournals.org/

2

Running title: Prognostic impact of NK cell count in FL and DLBCL

Word count manuscript: 3365

Tables/Figures: 3 Tables/3 Figures

Keywords: natural killer cells, natural killer cell count, follicular lymphoma, diffuse large B-

cell lymphoma, progression-free survival

Conflict-of-interest disclosures

MK, MZO, HZ, AC, AK, ND, and JV, report employment for F. Hoffmann-La Roche Ltd. JV

and CB report employment for Genentech. CB, TN, DZ, CK, and GFR report employment

and equity ownership for F. Hoffmann-La Roche Ltd. LHS reports consultancy and honoraria

for F. Hoffmann-La Roche Ltd, Genentech, Janssen, Amgen, Celgene, AbbVie, and Seattle

Genetics. WH reports research funding, honoraria, and advisory board membership for F.

Hoffmann-La Roche AG, Janssen, and Celgene. RM reports consultancy, honoraria, and

speakers bureau for F. Hoffmann-La Roche Ltd; and support for attending meetings for

Celgene. UV reports honoraria for F. Hoffmann-La Roche Ltd, Celgene, Janssen, Takeda,

and Gilead; advisory board membership for F. Hoffmann-La Roche Ltd, Celgene, and

Janssen; and research funding for F. Hoffmann-La Roche Ltd. VG reports advisory board

membership for F. Hoffmann-La Roche Ltd, Janssen, and Gilead; and honoraria and travel

grants for F. Hoffmann-La Roche Ltd and Janssen. AB reports a year-long academic

collaboration contract with F. Hoffmann-La Roche Ltd. MT reports consultancy and honoraria

for F. Hoffmann-La Roche Ltd, Celgene, Janssen, AbbVie, BMS, Takeda, and Gilead; and

research funding for F. Hoffmann-La Roche Ltd and Celgene.




3

Translational relevance: NK cells are key effector cells for anti-CD20 monoclonal

antibodies, such as obinutuzumab and rituximab. However, it is unclear whether

pretreatment NK cell count (NKCC) affects outcome in patients receiving these therapies.

Here, we report that low peripheral blood NKCC at baseline is independently associated with

shorter PFS in both FL and DLBCL, and shorter OS in FL. We also show that low tumor NK

cell gene expression is associated with shorter PFS in obinutuzumab-treated, but not

rituximab-treated, DLBCL patients, which may reflect the stronger ability of obinutuzumab, a

Fc-glycoengineered antibody, to trigger ADCC compared with rituximab. These findings are

of translational relevance as they indicate the number of NK cells in peripheral blood at

baseline could impact the clinical outcome of FL and DLBCL patients treated with anti-CD20

antibodies. Peripheral blood NKCC might therefore represent a valuable and easily

accessible biomarker in clinical practice.




4

Abstract

Purpose: Natural killer (NK) cells are key effector cells for anti-CD20 monoclonal antibodies,

such as obinutuzumab (G) and rituximab (R). We assessed whether low pretreatment NK

cell count (NKCC) in peripheral blood or tumor tissue was associated with worse outcome in

patients receiving antibody-based therapy.

Methods: Baseline peripheral blood NKCC was assessed by flow cytometry (CD3-CD56+

and/or CD16+ cells) in 1064/1202 follicular lymphoma (FL) patients treated with G or R plus

chemotherapy in the phase III GALLIUM trial (NCT01332968) and 1287/1418 diffuse large

B-cell lymphoma (DLBCL) patients treated with G or R plus cyclophosphamide, doxorubicin,

vincristine, and prednisone (CHOP) in the phase III GOYA trial (NCT01287741). The

prognostic value of tumor NK cell gene expression, as assessed by whole transcriptome

gene expression using Truseq RNA sequencing, was also analyzed. The association of

baseline variables, such as treatment arm, was evaluated using multivariate Cox regression

models using a stepwise approach.

Results: In this exploratory analysis, low baseline peripheral blood NKCC was associated

with shorter PFS in both FL (hazard ratio [HR] 1.48, 95% confidence interval [CI], 1.02-2.14,

P = 0.04) and DLBCL (HR, 1.36, 95% CI, 1.01-1.83, P = 0.04), and OS in FL (HR, 2.20, 95%

CI, 1.26-3.86, P = 0.0058). Low tumor NK cell gene expression was associated with shorter

PFS in G-CHOP-treated DLBCL patients (HR, 1.95, 95% CI, 1.22-3.15, P < 0.01).

Conclusion: These findings indicate that the number of NK cells in peripheral blood may

affect the outcome of B-cell non-Hodgkin lymphoma patients receiving anti-CD20-based

immunochemotherapy.




5

Introduction

Natural killer (NK) cells are a key element of the innate immune system, with an

important role in maintaining immune surveillance of virus-infected or transformed cells. NK

cells are part of the hematopoietic system and are derived from CD34+ hematopoietic

progenitor cells (1). In general, NK cells are characterized by their absence of the T-cell

receptor complex (CD3) along with the presence of CD56, a 140 kDa isoform of neural cell

adhesion molecule. The surface expression of CD56 defines two main subsets of NK cells,

CD56bright and CD56dim (2, 3). CD56bright NK cells make up approximately 10% of circulating

blood NK cells, and are characterized by a high surface expression of CD56 and low or

negative levels of Fcγ receptor IIIA (FcγRIIIa or CD16) (3, 4). In contrast, CD56dim NK cells

represent at least 90% of the NK cell population in peripheral blood, and are characterized

by a low surface expression of CD56 and high levels of FcγRIIIa (3, 4).

FcγRIIIa is a low affinity activating receptor that can bind to the constant (Fc) region

of IgG antibodies when immobilized on the cell surface (5). Ligation of FcγRIIIa on NK cells

with the Fc region of antibody-coated tumor cells results in degranulation of NK cells, with

subsequent lysis of tumor cells by perforins and granzymes and secretion of

proinflammatory cytokines; this consequently leads to the activation of other immune cells,

including monocytes/macrophages, dendritic cells, and granulocytes. This process is known

as antibody-dependent cellular cytotoxicity (ADCC), and represents an important mechanism

of action of monoclonal antibodies, such as the anti-CD20 monoclonal antibody, rituximab

(R). Other mechanisms responsible for the antitumor efficacy of R have been described,

including direct cell death and complement-mediated cellular cytotoxicity (6-9).

ADCC enhancement through Fc domain modification has also shown promise in the

development of next generation antibodies (10). Obinutuzumab (GA101; G) is a type II anti-

CD20 monoclonal antibody with enhanced direct cell death activity and a glycoengineered

Fc region that enhances its binding affinity to FcγRIIIa on NK and other effector cells,




6

thereby resulting in stronger ADCC compared with R (11-14). Such high affinity ligation of

FcγRIIIa has been shown to prime NK cells for interferon-gamma (IFN-γ) production (15).

As NK cells are the key mediators of ADCC, we postulated that low pretreatment NK

cell count (NKCC) may be associated with worse outcomes in patients with B-cell non-

Hodgkin lymphoma (B-NHL) treated with anti-CD20-based immunochemotherapy. Here, we

evaluate the prognostic impact of baseline NKCC in follicular lymphoma (FL) patients treated

with G or R plus chemotherapy in the phase III GALLIUM trial and diffuse large B-cell

lymphoma (DLBCL) patients treated with G or R plus cyclophosphamide, doxorubicin,

vincristine, and prednisone (CHOP) in the phase III GOYA trial (16, 17). The prognostic

value of tumor NK cell gene expression was also analyzed in DLBCL patients.

Methods

Patients, treatments, and assessments

The GALLIUM and GOYA study designs are described in full elsewhere (16, 17). In

brief, in the GALLIUM study, eligible patients had previously untreated, histologically

documented, CD20-positive FL (histological grades 1–3a), Eastern Cooperative Oncology

Group (ECOG) performance status 0–2, stage III/IV disease (or stage II with bulky disease,

i.e., largest tumor diameter ≥7 cm), and required treatment according to the Groupe d’Étude

des Lymphomes Folliculaires (GELF) criteria. Patients were treated with G or R plus

chemotherapy (bendamustine; cyclophosphamide, vincristine, and prednisone [CVP]; or

CHOP) for six or eight cycles depending on the chemotherapy. Choice of chemotherapy was

stipulated by site, with all patients at a given site receiving the same chemotherapy

backbone. Patients who achieved complete or partial remission at the end of induction

received maintenance therapy with G or R every two months for two years, or until disease

progression or study withdrawal if earlier. In the GOYA study, eligible patients had previously

untreated, histologically documented, CD20-positive DLBCL, an ECOG performance status

0–2, and an International Prognostic Index (IPI) score ≥2 (or IPI score 1 if aged ≤60 years,




7

with or without bulky disease; or IPI score 0 with bulky disease, i.e., one lesion ≥7.5 cm).

Patients were treated with eight 21-day cycles of G or R plus six-to-eight cycles of CHOP

chemotherapy. In both studies, tumor response was assessed using a modified response

criteria for NHL (18). GALLIUM was conducted in accordance with the Declaration of

Helsinki and the International Conference on Harmonization guidelines for Good Clinical

Practice. GOYA was conducted in accordance to the Declaration of Helsinki. The protocols

for the GALLIUM and GOYA trials were approved by the ethics committees of participating

centers and the trials were registered at ClinicalTrials.gov (NCT01332968 and

NCT01287741, respectively). All patients provided written informed consent.

Flow cytometry analysis

Baseline NKCCs were centrally assessed from EDTA-anticoagulated peripheral

blood by flow cytometry (Quintiles laboratories Ltd, Marietta, Georgia). Briefly, peripheral

blood was collected in Cytochex BCT tubes and shipped to central laboratories located in

the US, Europe, and Asia. Whole blood (50 μL) was stained with the BD Multitest 6-color

TBNK assay in TruCount tubes (BD Biosciences, San Jose, California). BD lyse solution was

used for red blood cell lysis. The number of NK cells, which were defined as CD3-CD56+

and/or CD16+ cells, was measured by flow cytometry (FACSCanto II; BD Biosciences, San

Jose, California; normal range [determined by central laboratory]: 95–640 cells/µL).

Cell-of-origin analysis

Cell-of-origin (COO) classification was based on gene expression profiling using the

NanoString Lymphoma Subtyping Research-Use-Only assay (NanoString Technologies,

Inc., Seattle, US) on RNA extracted from formalin-fixed paraffin-embedded (FFPE) tumor

tissue.

Whole transcriptome RNA sequencing

RNA was extracted from FFPE tissues using the RNeasy FFPE kit (Qiagen, Hilden,

Germany). Whole transcriptome gene expression was analyzed using TruSeq RNA




8

sequencing. A 57-gene signature designed to reflect NK cell tumor infiltration was

subsequently applied to RNA sequencing data, as previously described (19), and a signature

score was calculated as the first principal component of expression in those 57 genes.

Median score was used to define low/high subgroups for tumor NK cell gene expression.

Statistical analysis

This exploratory post-hoc analysis was carried out in FL (GALLIUM) and DLBCL

(GOYA) patients with an evaluable peripheral blood NKCC at baseline. The optimal cut-off

for defining low and normal/high NKCC subpopulations was analyzed in FL and DLBCL

patients using a SAS macro that implements the maximal Chi-square statistic (20, 21); this

cut-off, which led to the greatest difference in PFS between low and normal/high NKCC

subgroups, was 99 cells/μL for FL and 102 cells/μL for DLBCL. Low NKCC was thus defined

as <100 cells/µL and normal/high NKCC was defined as ≥100 cells/μL, as previously

described by He et al (22).

The association of baseline variables (sex, geographic region, treatment arm,

chemotherapy backbone/number of planned chemotherapy cycles, Follicular Lymphoma

International Prognostic Index [FLIPI]/IPI, extranodal/bone marrow involvement, sum of

products of diameter [SPD], and Ann Arbor stage) with peripheral blood NKCC was

evaluated using multivariate Cox regression models using a stepwise approach. These

statistical tests were two-sided with no adjustment for multiplicity. For the tumor NK cell gene

expression signature, hazard ratios (HRs) and 95% confidence intervals (CIs) were

estimated using a multivariate Cox proportional hazard, with treatment arm, number of

planned chemotherapy cycles, IPI, geographic region (GOYA) or treatment arm, FLIPI, and

chemotherapy backbone (GALLIUM) included as covariates. A Cox multivariate model was

used to evaluate the potential predictive value of NKCC as a biomarker factor for

progression-free survival (PFS) and overall survival (OS), and of tumor NK cell gene

expression signature as a biomarker factor for PFS, in patients treated with R-chemo and G-

chemo (treatment-by-NKCC interaction).




9

Qualified researchers may request access to individual patient level data through the

clinical study data request platform (www.clinicalstudydatarequest.com). Further details on

Roche's criteria for eligible studies are available here

(https://clinicalstudydatarequest.com/Study-Sponsors/Study-Sponsors-Roche.aspx). For

further details on Roche's Global Policy on the Sharing of Clinical Information and how to

request access to related clinical study documents, see here

(https://www.roche.com/research_and_development/who_we_are_how_we_work/clinical_tri

als/our_commitment_to_data_sharing.htm).

Results

Patients and baseline peripheral blood NKCCs

The analysis population consisted of patients with an evaluable NKCC at baseline

(1064/1202 FL patients [88.5% of the GALLIUM ITT population] and 1287/1418 DLBCL

patients [90.8% of the GOYA ITT population]. Baseline patient and disease characteristics

were comparable between the analysis population and the respective ITT populations

(Supplemental Table 1). Median (range) baseline peripheral blood NKCCs were 222 cells/μL

(13–3327) in FL and 196 cells/μL (5–1930) in DLBCL patients. Overall, 108/1064 (10.2%) FL

patients and 255/1287 (19.8%) DLBCL patients had a low baseline peripheral blood NKCC

(<100 cells/μL) (Supplementary Fig. 1A).

COO and baseline NKCC were available in 857/1418 (60.4%) DLBCL patients. By

COO subtype, median (range) baseline peripheral blood NKCCs were 186 cells/μL (6–

1659), 167 cells/μL (7–1715), and 200 cells/μL (8–1930) in germinal center B-cell-like

(GCB), unclassified, and activated B-cell-like (ABC) DLBCL, respectively. Low baseline

peripheral blood NKCC was detected in 83/485 (17.1%) GCB, 37/140 (26.4%) unclassified,

and 54/232 (23.3%) ABC DLBCL patients (P = 0.022 representing a significant imbalance

between COO subtypes) (Supplementary Fig. 1B).



https://www.roche.com/research_and_development/who_we_are_how_we_work/clinical_trials/our_commitment_to_data_sharing.htm

https://www.roche.com/research_and_development/who_we_are_how_we_work/clinical_trials/our_commitment_to_data_sharing.htm


10

Baseline patient and disease characteristics of FL and DLBCL patients with low vs.

normal/high baseline peripheral blood NKCCs are shown in Table 1. FL patients with low

baseline peripheral blood NKCC more frequently had extranodal involvement, elevated

lactate dehydrogenase (LDH) levels, and high FLIPI score compared with patients with

normal/high NKCC. Similarly, DLBCL patients with low baseline peripheral blood NKCC

more frequently had higher Ann Arbor stage, ECOG performance status, and IPI score,

extranodal involvement, and elevated LDH levels compared with patients with normal/high

NKCC. SPD was higher in patients with low baseline peripheral blood NKCC compared with

patients with normal/high NKCC in FL and DLBCL. No difference in the frequency of bone

marrow involvement between patients with low vs. normal/high baseline peripheral blood

NKCC was observed in either FL or DLBCL.

Impact of low baseline peripheral blood NKCC on clinical outcome

Multivariate analyses with baseline patient and disease characteristics, treatment

arm, and chemotherapy backbone (in the case of GALLIUM) as covariates showed that low

baseline peripheral blood NKCC was independently associated with shorter PFS in FL (HR,

1.48, 95% CI, 1.02-2.14, P = 0.04; 3-year PFS rate, 71.6% vs. 80.1% in patients with

normal/high baseline peripheral blood NKCC) and DLBCL (HR, 1.36, 95% CI, 1.01-1.83, P =

0.04; 3-year PFS rate, 62.8% vs. 70.0%) (Table 2; Fig. 1A), and OS in FL (HR, 2.20, 95%

CI, 1.26-3.86, P = 0.0058; 3-year OS rate, 87.6% vs. 94.3%); (Table 3; Fig. 1B). The DLBCL

PFS result appeared to be driven by COO subtype, with the highest estimated HR in the

GCB subtype (HR 1.58, 95% CI, 1.00-2.50, P = 0.05; 3-year PFS rate, 70.0% vs. 77.4% in

patients with normal/high baseline peripheral blood NKCC); no effect was observed in the

unclassified and ABC subtypes (Fig. 2A). No significant impact of low baseline peripheral

blood NKCC on OS was observed in the GCB, unclassified, or ABC subtypes (Fig. 2B).

Low baseline peripheral blood NKCC was associated with shorter PFS irrespective of the

type of anti-CD20 antibody used ([univariate analysis] FL: R-chemo, HR, 1.19, 95% CI, 0.71-




11

1.99; G-chemo, HR, 2.06, 95% CI, 1.24-3.41; DLBCL: R-CHOP, HR, 1.47, 95% CI, 1.06-

2.06; G-CHOP, HR, 1.26, 95% CI, 0.90-1.77).

Multivariate analysis with treatment-by-NKCC interaction showed the treatment effect

of G-chemo/CHOP and R-chemo/CHOP on PFS and OS persisted regardless of the NKCC

of patients.

Impact of low tumor NK cell gene expression signature on clinical outcome

Gene expression by whole transcriptome RNA sequencing was assessed in tumor

tissue of 236/1202 (19.6%) FL patients in GALLIUM and 552/1418 (38.9%) DLBCL patients

in GOYA. Low CD56 mRNA expression alone (as a continuous variable) correlated with

shorter PFS (P = 0.043) in DLBCL, but not in FL (P = 0.447). Baseline peripheral blood

NKCC and tumor NK cell gene expression were evaluable in 201/1202 (16.7%) and

508/1418 (35.8%) FL and DLBCL patients, respectively. There was no correlation between

baseline peripheral blood NKCC and tumor NK cell gene expression among biomarker-

evaluable FL (r = 0.053, P = 0.46) and DLBCL (r = 0.05, P = 0.25) patients (Supplemental

Fig. 2A/B). Multivariate analysis showed no significant difference in PFS between low vs.

high tumor NK cell gene expression in patients with FL (HR, 0.84, 95% CI, 0.50-1.4; P = 0.5)

or DLBCL (HR, 1.31, 95% CI, 0.95-1.81; P = 0.11) (Fig. 3A and 3D). In a subgroup

(multivariate) analysis of tumor NK cell gene expression and outcome, no effect of low tumor

NK cell gene expression on PFS was observed in FL patients treated with R-chemo (HR,

0.56, 95% CI, 0.29-1.10; P = 0.09) or G-chemo (HR, 1.68, 95% CI, 0.72-3.94; P = 0.23) (Fig.

3B and 3C), or in DLBCL patients treated with R-CHOP (HR, 0.92, 95% CI, 0.57-1.48, P =

0.72) (Fig. 3E); by contrast, low tumor NK cell gene expression was associated with shorter

PFS in DLBCL patients treated with G-CHOP (HR, 1.95, 95% CI, 1.22-3.15, P < 0.01) (Fig.

3F). Results of an interaction test showed that the prognostic value of the NK cell signature

was significantly stronger among G-CHOP-treated patients compared with R-CHOP-treated

patients (P = 0.016).




12

Baseline NKCC and risk or severity of infusion-related reactions and infections

The frequency of grade 3–5 and serious infusion-related reactions (IRRs) were

comparable in patients with low and normal/high baseline peripheral blood NKCC in both FL

and DLBCL (Supplemental Table 2). No grade 5 IRRs were observed in either GOYA or

GALLIUM. The frequency of grade 3–5 and serious infections were comparable in DLBCL

patients with low and normal/high baseline peripheral blood NKCC. In contrast, there

appeared to be more grade 3–5 and serious infections among FL patients with low NKCC,

driven by more infections in the G-chemo arm (Supplemental Table 3).

Discussion

In the current study, we evaluated the prognostic impact of baseline peripheral blood

NKCC in patients with FL and DLBCL who were treated with the anti-CD20 monoclonal

antibodies, G or R, in combination with chemotherapy in the first-line, phase III GALLIUM

(FL) and GOYA (DLBCL) trials. In this exploratory, post-hoc analysis carried out on the

largest prospective collection of NKCCs to date, we found that a subset of patients with FL

and DLBCL had a low NKCC at baseline, which was independently associated with shorter

PFS in both FL and DLBCL, and shorter OS in FL (but not DLBCL). The prognostic impact of

baseline peripheral blood NKCC in these lymphoma subtypes is consistent with the results

of previous smaller retrospective studies (22-26) and is in accordance with findings from in

vitro studies of ADCC, which unequivocally demonstrate that the of level tumor-cell killing is

decreased with lower effector-to-target ratios (27, 28).

We showed evidence for a prognostic role of NK cells in germinal center-derived

lymphomas, such as FL and GCB DLBCL, while no significant effect was observed in the

unclassified or ABC DLBCL subtypes. The reason for this observation remains unclear,

although it could be explained by differences in their microenvironment and underlying tumor

genetics given that particular COO subtypes exhibit distinct mutational profiles and are

driven by different oncogenic pathways (29-31). For example, the GCB DLBCL subtype,




13

which has a more favorable prognosis than other DLBCL subtypes, is molecularly similar to

FL, with frequent alterations in epigenetic modifiers, such as EZH2, or translocation of the

BCL2 gene resulting in deregulation of apoptosis (31-33). In contrast, the hallmarks of ABC

DLBCL include constitutive activation of the NFkB pathway and cell cycle deregulation due

to loss of CDKN2A (30, 34). It may be the case that low NKCC loses its importance in the

presence of the latter genetic abnormalities and an unfavorable microenvironment, both of

which are associated with the poor prognosis of ABC DLBCL. The number of patients with

unclassified (n = 140) and ABC (n = 232) DLBCL was fewer than those with GCB DLBCL (n

= 485), and thus limited information was available on NKCC in patients with these DLBCL

subtypes.

A subset of patients with FL and DLBCL had reduced peripheral blood NKCC at

baseline; of note, low NKCC was almost twice as prevalent in DLBCL patients compared

with FL patients (19.8% vs. 10.2%). The precise mechanism explaining the observed

incidence of reduced NKCC is unknown. NK cells develop from bone marrow hematopoietic

stem cells and mature in secondary lymphoid tissue; thus, bone marrow infiltration by

lymphoma cells could suppress normal hematopoiesis, thereby reducing the absolute

number of circulating NK cells. However, consistent with findings from other studies, bone

marrow involvement at baseline was similar here for low versus normal/high baseline

peripheral blood NKCC in FL and DLBCL, suggesting that other mechanisms are involved.

It has been shown that lymphoma cells may block maturation of NK cells in the bone

marrow by interrupting pathways necessary for their development or via mechanisms

leading to direct inhibition or killing of immune cells, including NK cells (35, 36). These

mechanisms operate differently in FL and DLBCL, as well as in particular COO subtypes

(36), and thus may contribute to different levels of host immune suppression in B-NHL

subtypes. In line with this, we observed that FL and DLBCL patients with low baseline

peripheral blood NKCCs had adverse clinical characteristics compared with patients with

normal/high NKCCs. The association of low NKCC or impaired NK cell function with high-risk




14

or advanced-stage disease has also been shown in solid tumors and in hematologic

malignancies (37). These data suggest that low levels of NK cells or defective NK cells may

contribute to inadequate immunity against tumor cells, leading to more aggressive disease

manifestation and vice versa.

A low number of circulating NK cells was associated with worse outcome in

lymphoma patients treated with R or G in combination with chemotherapy. It is not well

known how representative the peripheral blood is for the overall number of NK cells in the

body. The tumor compartment may better reflect the functional interaction between

lymphoma and NK cells. However, the technical difficulties associated with the acquisition of

tumor tissue and the limited amount of tissue available may limit the value of a tissue-based

approach. Interestingly, we found no correlation between peripheral blood NKCC and tumor

NK cell gene expression signature (representative of the number of NK cells in tumor tissue).

The extent of NK cell recruitment to tumor tissue depends not only on their availability (i.e.

on their release from bone marrow to peripheral blood), but also on the intensity of local

chemokine signaling, which attracts NK cells into the tumor tissue and enables penetration

and local activation (38). Lack of such chemoattractant signaling would likely result in

insufficient NK cell infiltration even in the presence of a normal/high peripheral blood NKCC.

We observed that G-treated DLBCL patients with low expression of a tumor NK cell gene

expression signature had significantly shorter PFS compared with those with high

expression of the tumor NK cell signature. This effect was not observed in R-treated

patients, which may reflect the stronger ability of the Fc-glycoengineered G antibody to

trigger ADCC compared with R (12). Although this hypothesis was not confirmed in FL, it

should be noted that the number of FL patients with an evaluable NK cell gene expression

signature was much lower than that for DLBCL patients. Future studies are needed to

provide further insight.

Our results provide a rationale for investigating anti-CD20 antibodies in combination

with agents that enhance NK cell function. New therapeutic IL-2-based molecules have been




15

developed and are currently under clinical evaluation in various types of malignancies. In

addition to stimulating cytokines, such as IL-2 and IL-15, other agents have shown the ability

to activate NK cells. Immunomodulatory agents, such as thalidomide or lenalidomide, can

increase the number of circulating NK cells and augment their direct cytotoxic activity against

tumor cells (39). These data also encourage the investigation of combining anti-CD20

antibodies with adoptive T cell therapy using cytokine-induced killer cells (40). Further

studies are certainly needed to explore if such treatments would be more effective than

standard immunochemotherapy.

In conclusion, the results from the current exploratory analysis of the GALLIUM and

GOYA trials indicate that the number of NK cells in peripheral blood may impact the clinical

outcome of patients with B-NHL treated with anti-CD20 antibodies. Specifically, we found

that low peripheral blood NKCC impacted PFS in both FL and DLBCL, and OS in FL. Our

findings suggest that peripheral blood NKCC could represent a valuable biomarker in clinical

practice, and could pave the way to the development of novel combination treatment

approaches that aim to enhance the activity of anti-CD20 antibodies by providing them with

a sufficient number of functional effector cells in B-NHL.




16

Author’s Contributions

Conception and design: MK, MZO, TN, GFR, HZ, DS, ND, CK, and AK; CB and JV

conceived and designed the sequencing data analysis

Acquisition of data (provided animals, acquired and managed patients, provided

facilities, etc): MT, WH, RM, LHS, UV, AB, and VG provided study materials or patients

Writing of the manuscript: MK

Analysis and interpretation of the data (e.g., statistical analysis, biostatics,

computational analysis): all authors

Critical review and approval of manuscript: all authors

Acknowledgments

The authors wish to thank the study investigators, coordinators, nurses, and patients of the

GOYA and GALLIUM trials, and Susan Robson for her invaluable contribution to the data

analysis and interpretation. GOYA and GALLIUM were supported by F. Hoffmann-La Roche

Ltd. Editorial support was provided by Janis Noonan, PhD (Gardiner-Caldwell

Communications Ltd, Macclesfield, UK), and funded by F. Hoffmann-La Roche Ltd.




17

References

1. Galy A, Travis M, Cen D, Chen B. Human T, B, natural killer, and dendritic cells arise

from a common bone marrow progenitor cell subset. Immunity. 1995;3(4):459-73.

2. Lanier LL, Testi R, Bindl J, Phillips JH. Identity of Leu-19 (CD56) leukocyte

differentiation antigen and neural cell adhesion molecule. J Exp Med. 1989;169(6):2233-8.

3. Caligiuri MA. Human natural killer cells. Blood. 2008;112(3):461-9.

4. Cooper MA, Fehniger TA, Caligiuri MA. The biology of human natural killer-cell

subsets. Trends in immunology. 2001;22(11):633-40.

5. Nimmerjahn F, Ravetch JV. Fcgamma receptors as regulators of immune responses.

Nat Rev Immunol. 2008;8(1):34-47.

6. Abulayha A, Bredan A, El Enshasy H, Daniels I. Rituximab: modes of action,

remaining dispute and future perspective. Future oncology (London, England).

2014;10(15):2481-92.

7. Maloney DG, Smith B, Rose A. Rituximab: Mechanism of action and resistance.

Seminars in oncology. 2002;29(1s2):2-9.

8. Weiner GJ. Rituximab: mechanism of action. Seminars in hematology.

2010;47(2):115-23.

9. Boross P, Leusen JH. Mechanisms of action of CD20 antibodies. American journal of

cancer research. 2012;2(6):676-90.

10. Kellner C, Zhukovsky EA, Potzke A, Bruggemann M, Schrauder A, Schrappe M, et

al. The Fc-engineered CD19 antibody MOR208 (XmAb5574) induces natural killer cell-

mediated lysis of acute lymphoblastic leukemia cells from pediatric and adult patients.

Leukemia. 2013;27(7):1595-8.

11. Niederfellner G, Lammens A, Mundigl O, Georges GJ, Schaefer W, Schwaiger M, et

al. Epitope characterization and crystal structure of GA101 provide insights into the

molecular basis for type I/II distinction of CD20 antibodies. Blood. 2011;118(2):358-67.

12. Mossner E, Brunker P, Moser S, Puntener U, Schmidt C, Herter S, et al. Increasing

the efficacy of CD20 antibody therapy through the engineering of a new type II anti-CD20

antibody with enhanced direct and immune effector cell-mediated B-cell cytotoxicity. Blood.

2010;115(22):4393-402.

13. Bologna L, Gotti E, Manganini M, Rambaldi A, Intermesoli T, Introna M, et al.

Mechanism of action of type II, glycoengineered, anti-CD20 monoclonal antibody GA101 in

B-chronic lymphocytic leukemia whole blood assays in comparison with rituximab and

alemtuzumab. J Immunol. 2011;186(6):3762-9.




18

14. Klein C, Lammens A, Schafer W, Georges G, Schwaiger M, Mossner E, et al.

Epitope interactions of monoclonal antibodies targeting CD20 and their relationship to

functional properties. MAbs. 2013;5(1):22-33.

15. Capuano C, Pighi C, Molfetta R, Paolini R, Battella S, Palmieri G, et al.

Obinutuzumab-mediated high-affinity ligation of FcgammaRIIIA/CD16 primes NK cells for

IFNgamma production. Oncoimmunology. 2017;6(3):e1290037.

16. Marcus R, Davies A, Ando K, Klapper W, Opat S, Owen C, et al. Obinutuzumab for

the First-Line Treatment of Follicular Lymphoma. N Engl J Med. 2017;377(14):1331-44.

17. Vitolo U, Trneny M, Belada D, Burke JM, Carella AM, Chua N, et al. Obinutuzumab

or Rituximab Plus Cyclophosphamide, Doxorubicin, Vincristine, and Prednisone in

Previously Untreated Diffuse Large B-Cell Lymphoma. J Clin Oncol. 2017:JCO2017733402.

18. Cheson BD, Pfistner B, Juweid ME, Gascoyne RD, Specht L, Horning SJ, et al.

Revised response criteria for malignant lymphoma. J Clin Oncol. 2007;25(5):579-86.

19. Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, et al. Robust

enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015;12(5):453-7.

20. D MRaS. Maximally selected chi square statistics. Biometrics. 1982;38(4).

21. B ZQaD. A SAS® Macro for biomarker analysis using maximally selected Chi square

statistics with application in oncology. PharmaSUG. 2012:8.

22. He L, Zhu HY, Qin SC, Li Y, Miao Y, Liang JH, et al. Low natural killer (NK) cell

counts in peripheral blood adversely affect clinical outcome of patients with follicular

lymphoma. Blood Cancer J. 2016;6(8):e457.

23. Plonquet A, Haioun C, Jais JP, Debard AL, Salles G, Bene MC, et al. Peripheral

blood natural killer cell count is associated with clinical outcome in patients with aaIPI 2-3

diffuse large B-cell lymphoma. Ann Oncol. 2007;18(7):1209-15.

24. Shafer D, Smith MR, Borghaei H, Millenson MM, Li T, Litwin S, et al. Low NK cell

counts in peripheral blood are associated with inferior overall survival in patients with

follicular lymphoma. Leukemia research. 2013;37(10):1213-5.

25. Du J, Lopez-Verges S, Pitcher BN, Johnson J, Jung SH, Zhou L, et al. CALGB

150905 (Alliance): rituximab broadens the antilymphoma response by activating unlicensed

NK cells. Cancer immunology research. 2014;2(9):878-89.

26. Jurczak W, Zinzani PL, Gaidano G, Goy A, Provencio M, Nagy Z, et al. Phase IIa

study of the CD19 antibody MOR208 in patients with relapsed or refractory B-cell non-

Hodgkin's lymphoma. Ann Oncol. 2018;29(5):1266-72.

27. Palazzo A Herter S Grosmaire L JR, Frey CR, Limani F, et al. The PI3Kδ-selective

inhibitor idelalisib minimally interferes with immune effector function mediated by rituximab or

obinutuzumab and significantly augments B cell depletion in vivo. J Immunol. 2018;200

(7):2304-12.




19

28. Wang SY Racila E TR, Weiner GJ. NK-cell activation and antibody-dependent

cellular cytotoxicity induced by rituximab-coated target cells is inhibited by the C3b

component of complement. Blood. 2008;111(3):1456-63.

29. Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A, et al. Distinct

types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature.

2000;403(6769):503-11.

30. Lenz G, Wright GW, Emre NC, Kohlhammer H, Dave SS, Davis RE, et al. Molecular

subtypes of diffuse large B-cell lymphoma arise by distinct genetic pathways. Proc Natl Acad

Sci U S A. 2008;105(36):13520-5.

31. Oestergaard M, Bolen C, Mattiello F, et al. Superiority of obinutuzumab over

rituximab in a new molecular follicular lymphoma-like subgroup of DLBCL: results from an

exploratory analysis of the phase 3 Goya trial. Blood. 2017;130(1543).

32. Morin RD, Mendez-Lago M, Mungall AJ, Goya R, Mungall KL, Corbett RD, et al.

Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature.

2011;476(7360):298-303.

33. Pasqualucci L, Trifonov V, Fabbri G, Ma J, Rossi D, Chiarenza A, et al. Analysis of

the coding genome of diffuse large B-cell lymphoma. Nat Genet. 2011;43(9):830-7.

34. Davis RE, Ngo VN, Lenz G, Tolar P, Young RM, Romesser PB, et al. Chronic active

B-cell-receptor signalling in diffuse large B-cell lymphoma. Nature. 2010;463(7277):88-92.

35. Richards JO, Chang X, Blaser BW, Caligiuri MA, Zheng P, Liu Y. Tumor growth

impedes natural-killer-cell maturation in the bone marrow. Blood. 2006;108(1):246-52.

36. de Charette M, Houot R. Hide or defend, the two strategies of lymphoma immune

evasion: potential implications for immunotherapy. Haematologica. 2018;103(8):1256-68.

37. Hejazi M, Manser AR, Frobel J, Kundgen A, Zhao X, Schonberg K, et al. Impaired

cytotoxicity associated with defective natural killer cell differentiation in myelodysplastic

syndromes. Haematologica. 2015;100(5):643-52.

38. Przewoznik M, Homberg N, Naujoks M, Potzl J, Munchmeier N, Brenner CD, et al.

Recruitment of natural killer cells in advanced stages of endogenously arising B-cell

lymphoma: implications for therapeutic cell transfer. J Immunother. 2012;35(3):217-22.

39. Davies FE, Raje N, Hideshima T, Lentzsch S, Young G, Tai YT, et al. Thalidomide

and immunomodulatory derivatives augment natural killer cell cytotoxicity in multiple

myeloma. Blood. 2001;98(1):210-6.

40. Pievani A, Belussi C, Klein C, Rambaldi A, Golay J, Introna M. Enhanced killing of

human B-cell lymphoma targets by combined use of cytokine-induced killer cell (CIK)

cultures and anti-CD20 antibodies. Blood. 2011;117(2):510-8.




20

Table 1. Baseline patient demographics and disease characteristics.

FL (GALLIUM)

n = 1064 DLBCL (GOYA)

n = 1287

NKCC <100

cells/μL n = 108

NKCC ≥100

cells/μL n = 956

NKCC <100

cells/μL n = 255

NKCC ≥100

cells/μL n = 1032

Age, years ≥65 31 (28.7) 303 (31.7) 86 (33.7) 423 (41.0)

ECOG PS ≥2 5 (4.6) 30 (3.1)a 45 (17.6) 125 (12.1)f

Ann Arbor stage

III–IV 98 (90.7) 872 (91.9)b 220 (86.3) 758 (73.4)

Extranodal involvement

Yes 77 (71.3) 625 (65.4) 185 (72.5) 683 (66.2)

Bone marrow involvement

Yes 55 (51.4)c 491 (51.8)d 32 (12.7)g 103 (10.1)d

Bulky disease Yes 48 (44.4) 425 (44.5)e 110 (43.1) 356 (34.6)g

FLIPI High 54 (50.0) 393 (41.1) – –

IPI High – – 50 (19.6) 153 (14.8)

SPD, mm2 Mean (SD) 7762

(8961) 7053

(6520) 8624

(14302) 6631

(16850)

LDH Elevated 42 (38.9) 276 (29.0) 175 (68.6) 568 (55.2)

Treatment group

R-chemo/CHOP 56 (51.9) 456 (47.7) 116 (45.5) 532 (51.6)

G-chemo/CHOP 52 (48.1) 500 (52.3) 139 (54.5) 500 (48.4)

Data are n (%) unless otherwise specified. aData missing in 3 patients; bdata missing in 7 patients; cdata missing in 1 patient; ddata missing in 9 patients; edata missing in 2 patients; fdata missing in 1 patient; gdata missing in 3 patients.

Chemo, chemotherapy; CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; DLBCL, diffuse large B-cell lymphoma; ECOG PS, Eastern Cooperative Oncology Group performance status; FL, follicular lymphoma; FLIPI, Follicular Lymphoma International Prognostic Index; G, obinutuzumab; IPI, International Prognostic Index; LDH, lactate dehydrogenase; NKCC, natural killer cell count; R, rituximab; SD, standard deviation, SPD, sum of products of diameter of up to 6 target lesions.




21

Table 2. Multivariate analysis of the association between baseline variables and PFS in FL

(A) and DLBCL (B).a

A

FL, n = 1057 HR 95% CIc P value

NKCC <100 cells/μL, n = 108 1.48 1.02-2.14 0.0377

Treatment arm R-chemo, n = 508 1.31 1.02-1.70 0.0369

Chemotherapy backbone

CHOP, n = 349 1.09 0.82-1.45 0.5590

CVP, n = 108 1.73 1.18-2.54 0.0047

FLIPI Low, n = 219 0.67 0.46-0.98 0.0389

Intermediate, n = 393 0.79 0.59-1.06 0.1167

Sex Male, n = 497 1.60 1.24-2.08 0.0004


Yes, n = 698 1.46 1.09-1.96 0.0102

SPD, mm2b 1.00 1.00-1.00 0.0010

B

DLBCL, n = 850 HR 95% CIc P value

NKCC <100 cells/μL, n = 173 1.36 1.01-1.83 0.0416

Geographic region

Western Europe, n = 314 0.64 0.47-0.88 0.0060

Eastern Europe, n = 145 0.79 0.54-1.14 0.2051

North America, n = 143 0.63 0.42-0.96 0.0301

Other, n = 30 0.53 0.23-1.25 0.1455

IPI

Low, n = 165 0.37 0.24-0.56 <0.0001

Low-intermediate, n = 294 0.47 0.33-0.66 <0.0001

High-intermediate n = 255 0.62 0.44-0.86 0.0046

COO subtype GCB, n = 479 0.63 0.47-0.84 0.0017

Unclassified, n = 140 0.91 0.64-1.30 0.6113

SPD, mm2b 1.00 1.00-1.00 0.0057




22

aThe reference variables used included: (GALLIUM) NKCC, ≥100 cells/μL; treatment arm, G-chemo; chemotherapy backbone, bendamustine; FLIPI, high; sex, female; extranodal involvement, no; (GOYA) NKCC, ≥100 cells/μL; geographic region, Asia; IPI, high; COO subtype, ABC; banalyzed as

continuous variable; cWald confidence interval.

ABC, activated B-cell like; Chemo, chemotherapy; CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; CI, confidence interval; COO, cell-of-origin; CVP, cyclophosphamide, vincristine, and prednisone; DLBCL, diffuse large B-cell lymphoma; FL, follicular lymphoma; FLIPI, Follicular Lymphoma International Prognostic Index; G, obinutuzumab; GCB, germinal center B-cell like; HR, hazard ratio; IPI, International Prognostic Index; NKCC, natural killer cell count; PFS, progression-free survival; R, rituximab; SPD, sum of products of diameter of up to six target lesions.




23

Table 3. Multivariate analysis of the association between baseline variables and OS in FL

(A) and DLBCL (B).a

A

FL, n = 1057 HR 95% CIc P value

NKCC <100 cells/μL, n = 108 2.20 1.26-3.86 0.0058

Chemotherapy backbone

CHOP, n = 349 0.41 0.23-0.73 0.0024

CVP, n = 108 0.68 0.29-1.58 0.3725

FLIPI Low, n = 219 0.31 0.14-0.69 0.0043

Intermediate, n = 393 0.56 0.33-0.93 0.0260

Sex Male, n = 497 2.13 1.34-3.40 0.0014


Yes, n = 698 1.69 0.98-2.91 0.0587

SPD, mm2b 1.00 1.00-1.00 <0.0001

B

DLBCL, n = 850 HR 95% CIc P value

IPI

Low, n = 165 0.25 0.14-0.45 <0.0001

Low-intermediate, n = 294 0.41 0.27-0.62 <0.0001

High-intermediate n = 255 0.54 0.36-0.80 0.0024

Number of planned CHOP cycles

Eight, n = 235 1.81 1.32-2.49 0.0003

COO subtype GCB, n = 479 0.64 0.45-0.90 0.0115

Unclassified, n = 140 0.91 0.58-1.43 0.6904

SPD, mm2c 1.00 1.00-1.00 0.0031

aThe reference variables used included: (GALLIUM) NKCC, ≥100 cells/μL; chemotherapy backbone, bendamustine; FLIPI, high; sex, female; extranodal involvement, no; (GOYA) IPI, high; number of planned CHOP cycles, six; COO subtype, ABC; banalyzed as continuous variable; cWald confidence interval. ABC, activated B-cell like; CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; CI, confidence interval; COO, cell-of-origin; CVP, cyclophosphamide, vincristine, and prednisone; DLBCL, diffuse large B-cell lymphoma; FL, follicular lymphoma; FLIPI, Follicular Lymphoma International Prognostic Index; GCB, germinal center B-cell like; HR, hazard ratio; IPI, International Prognostic Index; NKCC, natural killer cell count; OS, overall survival; SPD, sum of

products of diameter of up to six target lesions.




24

Figure 1. Association of baseline peripheral blood NKCC with PFS (A) and OS (B) in FL and

DLBCL.

aMultivariate analysis adjusted for treatment arm, chemotherapy backbone, FLIPI-1, geographic region, sex, Ann Arbor stage, extranodal involvement, and SPD; bmultivariate analysis adjusted for treatment arm, number of planned CHOP cycles, IPI, geographic region, sex, bone marrow involvement, SPD, and COO. CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; CI, confidence interval; COO, cell-of-origin; DLBCL, diffuse large B-cell lymphoma; FL, follicular lymphoma; FLIPI, Follicular Lymphoma International Prognostic Index; HR, hazard ratio; IPI, International Prognostic Index; NKCC, natural killer cell count; OS, overall survival; PFS, progression-free survival; SPD, sum of products of diameter of up to six target lesions.

Figure 2. Association of baseline peripheral blood NKCC with PFS (A) and OS (B) by COO.

aMultivariate analysis adjusted for treatment arm, number of planned CHOP cycles, IPI, geographic region, sex, bone marrow involvement, COO, and SPD. ABC, activated B-cell-like; CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; CI, confidence interval; COO, cell-of-origin; GCB, germinal center B-cell-like; HR, hazard ratio; IPI, International Prognostic Index; NKCC, natural killer cell count; OS, overall survival; PFS, progression-free survival; SPD, sum of products of diameter of up to six target lesions.

Figure 3. Association of tumor NK cell gene expression and PFS in FL patients treated with

R-chemo and G-chemo (pooled data) (A), R-chemo (B), and G-chemo (C) and DLBCL

patients treated with R-CHOP and G-CHOP (pooled data) (D), R-CHOP (E), and G-CHOP

(F).

Median score was used to define high/low subgroups; amultivariate analysis adjusted for number of planned chemotherapy cycles, IPI, and geographic region. CI, confidence interval; chemo, chemotherapy; CHOP, cyclophosphamide, doxorubicin, vincristine, and prednisone; DLBCL, diffuse large B-cell lymphoma; FL, follicular lymphoma; HR, hazard ratio; NK, natural killer; PFS, progression-free survival.













Published OnlineFirst May 3, 2019.Clin Cancer Res Magdalena Klanova, Mikkel Z Oestergaard, Marek Trnený, et al. with immunochemotherapylymphoma and diffuse large B-cell lymphoma patients treated Prognostic impact of natural killer cell count in follicular

Updated version

10.1158/1078-0432.CCR-18-3270doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2019/07/04/1078-0432.CCR-18-3270.DC1

Access the most recent supplemental material at:

Manuscript

Authorbeen edited. Author manuscripts have been peer reviewed and accepted for publication but have not yet

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/early/2019/05/31/1078-0432.CCR-18-3270To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-18-3270

http://clincancerres.aacrjournals.org/content/suppl/2019/07/04/1078-0432.CCR-18-3270.DC1

http://clincancerres.aacrjournals.org/cgi/alerts


http://clincancerres.aacrjournals.org/content/early/2019/05/31/1078-0432.CCR-18-3270


Documents

Prognostic impact of natural killer cell count in ......May 31, 2019 · Robert Marcus,5 6Laurie H Sehn, 7Umberto Vitolo, Alexandra Bazeos,8 Valentin Goede,9 Harald Zeuner, 3 Andrea