CD20-targeted measles virus shows high oncolytic specificity in clinical samples from lymphoma patients independent of prior rituximab therapy

Abstract

New therapeutic modalities for B-cell non-Hodgkin’s lymphomas (B-NHL) are needed, especially for relapsing and aggressive subtypes. Toward this end, we previously generated a fully CD20-targeted and armed measles virus, and tested its efficacy in a xenograft model of mantle cell lymphoma (MCL). Here, we quantify its spread in peripheral blood mononuclear cells and/or tissue of patients with different histological subtypes of B-NHL, including splenic marginal zone lymphoma (SMZL). CD20-targeted MV efficiently infects lymphoma cells from SMZL and MCL while sparing most cells in the CD20-negative population, in contrast to the parental vaccine-lineage MV, which infects CD20-positive and CD20-negative cells equally. Rituximab therapy (4–8 months before relapse) did not interfere with the infectivity and specificity of MV^greenH^blindantiCD20 in patient lymphoma samples. Thus, CD20-targeted oncolytic virotherapy is likely to be effective after previous antiCD20 therapy.

INTRODUCTION

Non-Hodgkin’s lymphoma classifies common hematological neoplasms accounting for an estimated 19 500 US deaths in 2009,¹ with approximately 80–90% of cases involving B lymphocytes. Current therapeutic options include chemotherapy, radiotherapy and monoclonal antibody immunotherapy, and although these therapies have proven clinical utility, new treatment modalities would be valuable, especially for aggressive and relapsing non-Hodgkin’s lymphoma.²

Oncolytic virotherapy exploits the preferential replication of certain viruses in transformed tissues to achieve therapeutic cancer responses, and shares little to no cross-resistance with other therapeutic modalities. Many viruses are currently being tested pre-clinically, and multiple viruses have progressed to phase I–III clinical trials.^3–6 One promising platform is measles virus (MV), a non-segmented, negative-strand RNA Paramyxovirus.⁷ The wild-type and vaccine strains of MV have a natural lymphotropism, using the signaling lymphocytic activation molecule (CD150) as a receptor,⁸ and spontaneous regressions of hematological malignancies have been observed following natural infections.⁹ Based on a reverse genetics system,¹⁰ MV vaccine-lineage strain (Edmonston B) has been developed as a cancer therapeutic. Tumor-cell specificity has been engineered by appending single-chain antibodies to the attachment protein hemagglutinin^11–13 whereas vaccine strain tropism, which includes the ubiquitously expressed CD46,¹⁴ has been restricted by mutating hemagglutinin residues sustaining receptor interactions.¹⁵

A hallmark of B-cell non-Hodgkin’s lymphoma (B-NHL) is CD20 expression, a molecule targeted by rituximab, a therapeutic monoclonal antibody.¹⁶ Previously, CD20-targeted MV has shown efficacy in a lymphoma xenograph model in immunocompromised mice,^13,17 but limited data exists for the infectivity of this virus in primary patient samples. Here, we quantify the infectivity of a CD20-targeted MV in patient samples from multiple histological subtypes of B-NHL, including splenic marginal zone lymphoma (SMZL), mantle cell lymphoma (MCL), Burkitt’s lymphoma, diffuse large B-cell lymphoma and B-cell chronic lymphocytic/small lymphocytic lymphoma. By comparison to the isogenic vaccine-lineage vector, we show lymphoma-cell specificity of the CD20-targeted virus, and that previous antiCD20 therapy does not compromise infectivity or CD20 specificity in patient samples. These findings support continued development of CD20-targeted MV toward clinical trials.

RESULTS AND DISCUSSION

Specificity of infection of MV^greenH^blindantiCD20 in patient samples

To identify those histological subtypes of B-NHL most sensitive to infection by MV^greenH^blindantiCD20,^13,17 we infected fresh and frozen lymphoma samples from 30 patients at a multiplicity of infection of one. Figure 1 shows a representative analysis of infection specificity for both MV^greenH^blindantiCD20 and MV^greenNSe. We noted high infectivity of SMZL patient samples, and therefore obtained additional samples for analysis. Infectivity in each subtype was compared with MCL samples, which has been previously identified as a subtype supporting CD20-targeted infection.¹³ At multiplicity of infection of one, MV^greenH^blindantiCD20 infected about 20% of CD20-positive cells from SMZL and MCL samples, whereas MV^greenNSe infected almost 40% of CD20-positive cells in the same samples (Figure 2, compare left and right values in each panel). CD20-positive cells of diffuse large B cell, B-cell chronic lymphocytic small lymphocytic lymphoma and Burkitt’s lymphoma patient samples were also efficiently infected by MV^greenH^blindantiCD20 with means of 13.3, 14.3 and 30.2% infection, respectively (Figure 2, third column: fine dash, solid, coarse dash horizontal lines, respectively). The average infectivity of MV^greenH^blindantiCD20 in SMZL and MCL histological subtypes increased two- to threefold with a multiplicity of infection of five (data not shown). CD20-negative T-cell large granular lymphocytic leukemia samples were poorly infected by MV^greenH^blindantiCD20 but more efficiently infected by MV^greenNSe (Figure 2, fourth column).

Figure 1.

Analysis of CD20 specificity of MV^greenH^blindantiCD20 in a representative B-NHL patient sample (top left). CD45+ lymphocytes were gated into CD20-positive and CD20-negative (bottom left) populations. Samples were left uninfected (two panels, center-left) or infected with the indicated virus. GFP (horizontal axis) was measured for each population to determine the percentage of infected cells. The specificity ratio was calculated by dividing the percentages of CD20+GFP+ and CD20−GFP+ cells.

Figure 2.

CD20-specific infection in patients with different subtypes of B-NHL. The percentage of infected cells (y axis) was determined by fluorescence-activated cell sorting analysis of GFP expression. For SMZL, MCL, T-cell large granular lymphocytic leukemia (T-LGL) and healthy donors, different individuals are represented by unique shapes. Filled symbols represent PBMC samples, open symbols represent tissue samples (spleen, lymph node and tonsil). B-cell chronic lymphocytic (B-CLL) small lymphocytic lymphoma (SLL), Burkitt’s and diffuse large B-cell (DLC) lymphoma samples are combined with each subtype represented by a different shape. Horizontal lines represent means. Data points for one Burkitt’s sample and two DLC samples, and the mean line for the ratio panel in column three, are outside of the presented y axes; the actual value of these points is indicated below them. Wilcoxon signed rank test, one-sided P values are reported. Horizontal lines represent means.

To assess the specificity conferred by CD20 targeting in each patient sample, we calculated the ratio of CD20-positive to CD20-negative cells infected by MV^greenH^blindantiCD20 and MV^greenNSe. As expected, MV^greenNSe infected CD20-positive and CD20-negative cells at similar levels, yielding ratios close to 1 for SMZL, MCL, and all other tested subtypes and control healthy donors (Figure 2, bottom row). In contrast, MV^greenH^blindantiCD20 infected CD20-positive cells 6.5 and 3.3 times more efficiently than CD20-negative cells in SMZL and MCL, respectively, and also showed varying degrees of CD20 specificity in Burkitt’s lymphoma and B-cell chronic lymphocytic small lymphocytic lymphoma (Figure 2, bottom row). In control healthy donors, MV^greenH^blindantiCD20 also spared the CD20-negative population (Figure 2, bottom row, fifth column), achieving ratios 10 times higher than MV^greenNSe. The difference between MV^greenH^blindantiCD20 and MV^greenNSe specificity ratios was significant for both SMZL (P=0.019) and MCL (P=0.016) patient samples (Figure 2, bottom row). We note that green fluorescent protein (GFP) expression by these viruses requires not only cell entry, but also significant replication. This implies that the infection efficiencies documented here are influenced not only by receptor specificity at cell entry, but also by post-entry effects.

Retargeted viruses offer a potential safety advantage over un-modified virus capable of infecting normal tissues, and also allow the elimination of tumors lacking natural viral receptors.¹⁸ In patients with B-NHL, all normal immune cells express CD46, potentially supporting infection that further compromises immune function. CD20 is a cell-specific molecule with established therapeutic relevance, making it an ideal target. The efficacy of targeted MV therapeutics can also be enhanced by engineering MV to express transgenes, such as purine nucleoside phosphorylase and the sodium iodide symporter that induce a bystander effect by activating chemotherapeutics and concentrating radioiodine, respectively.^13,19 Future work will focus on testing combined retargeting/arming strategies in primary patient samples to independently maximize CD20 specificity and oncolytic efficacy.

Previous pharmacological course and response to targeted OV

As candidate patients for a clinical trial of CD20-targeted MV in B-NHL will have likely undergone previous treatment with rituximab, we sought to correlate infection levels with therapies received by patients before donating samples. Four SMZL and six MCL patients had undergone previous rituximab treatment, with an average duration from last rituximab-containing therapy to sample acquisition of 205 days for SMZL and 138 days for MCL. Previous antiCD20 therapy in SMZL samples decreased the CD20-positive cell population by about half (P=0.014; Figure 3a), consistent with a more indolent clinical course and slower disease relapse, following therapy. Despite a smaller CD20-positive cell population, the specificity ratio in these samples remained unchanged (P=0.5; Figure 3b). The percentage of CD20-positive cells in post-therapy MCL samples was not statistically different for treated and therapy-naïve patient samples (P=0.331; Figure 3a), indicating MCL patients had re-established a CD20-positive cell population. Remarkably, CD20-positive cells from post-therapy MCL samples were infected as well as therapy-naïve samples, and CD20-negative infection rates trended lower, improving the specificity ratio compared with therapy-naïve samples, although not with statistical significance (P=0.089; Figure 3b). These data indicate that previous antiCD20 therapy does not interfere with infection efficiency or CD20 specificity of subsequent therapy with CD20-targeted oncolytic MV, especially in aggressive subtypes that rapidly re-establish CD20-positive disease.

Figure 3.

Effect of previous rituximab therapy on (a) the percentage of CD20-positive cells and (b) on the ratio of CD20-positive to CD20-negative cell infection. Average number of days between the last rituximab-containing therapy and sample acquisition is indicated below the horizontal axis in (a). (b) Ratios from Figure 2 (bottom row) grouped by presence or absence of rituximab-containing therapy before sample donation. Mann–Whitney test, one-sided P values are reported. Horizontal lines represent means.

Our observation that antiCD20 therapy does not interfere with infectivity or CD20 specificity of MV^greenH^blindantiCD20 in lymphoma peripheral blood mononuclear cell (PBMC) and tissue samples suggests that, rituximab treatment is not permanently selecting tumor cells with decreased or aberrant CD20 expression, at least not to a degree negatively affecting subsequent virus infectivity. In fact, the specificity of the CD20-targeted virus trended higher in patient samples that had received rituximab therapy compared with therapy-naïve samples, although this difference is not statistically significant. It is conceivable that previous antiCD20 therapy may purge normal cells with low-level CD20 expression, increasing the likelihood that cells expressing CD20 at levels supporting virus entry will be target lymphoma cells.

On the basis of the present data and proven efficacy in a preclinical murine model,¹⁷ we are developing a clinical trial for pretreated patients with relapsed, refractory MCL and SMZL that combines CD20-targeted MV expressing purine nucleoside phosphorylase with cyclophosphamide and fludarabine, analogous to the clinical FCR regimen, but with rituximab replaced by the targeted virus. This combination exploits the immunosuppressive qualities of cyclophosphamide to maximize virus replication, and enhances traditional chemotherapy by expressing purine nucleoside phosphorylase within the tumor microenvironment, tightly localizing activation of fludarabine to its active drug metabolite 2-fluoroadenine. This novel regimen, using a replication competent virus in place of an antibody, has the potential to achieve improved response rates in relapsed, refractory lymphoma patients through synergistic oncolysis.

MATERIALS AND METHODS

Patient samples

Fresh circulating lymphoma cells in the PBMCs and archived frozen tissue samples (between 2003 and 2009, stored in 10% dimethyl sulfoxide and 20% fetal calf serum) were obtained from patients diagnosed with various histological subtypes of B-NHL and T-cell large granular lymphocytic leukemia according to the World Health Organization classification.²⁰ This study was approved by the Institutional Review Board and all patients gave written informed consent. Lymphoma subtype details are as follows: SMZL (frozen spleen, n=6; frozen PBMC, n=1; fresh PBMC, n=1), MCL (frozen spleen, n=6; frozen tonsil, n=1, frozen PBMC, n=1; fresh PBMC, n=3), diffuse large B-cell lymphoma (frozen spleen, n=3; frozen submandibular, n=1), B-cell chronic lymphocytic small lymphocytic lymphoma (frozen spleen, n=2; frozen mesenteric lymph node, n=1; frozen left axilla, n=1), T-cell large granular lymphocytic leukemia (frozen spleen, n=2; frozen left axilla, n=1; frozen right axilla, n=1) and Burkitt’s lymphoma (frozen lymph node, n=2).

Virus strains and infection

We used the CD20-targeted MV expressing the GFP named MV^greenH^blindanti-CD20 previously generated in our laboratory.^12,13 Briefly, the viral hemagglutinin was mutated to abrogate CD46 and signaling lymphocytic activation molecule recognition,¹⁵ and a single-chain antibody specific for CD20 was appended to its C-terminus. GFP was placed in an additional transcription unit upstream of the N gene. For comparison purposes, the parental MV^greenNSe also encoding GFP upstream of N was included. For infection of fresh PBMCs, whole blood was collected in heparin-EDTA and subjected to Ficoll-Paque density gradient centrifugation (GE Healthcare Biosciences, Pittsburgh, PA, USA). The PBMC layer was collected and seeded (5×10⁴ per well) on 24-well plates in RPMI 1640 (Mediatech, Herndon, VA, USA) supplemented with 10% fetal calf serum and 1% penicillin (10 000 IU/ml)/streptomycin (10 000 μg/ml) (Mediatech, Manassas, VA, USA). For infection of archived samples, cells were thawed, washed twice in 10% RPMI and seeded 5×10⁴ cells per well on 24-well plates. Cells were infected with 5×10⁴ plaque-forming unit of MV^greenH^blind antiCD20 or MV^greenNSe in 1 ml of growth media.

Flow cytometry analysis

Infection and analysis of patient samples were performed as previously described.¹³ Briefly, 48 h after infection, cells were harvested and washed twice with washing buffer (5% fetal calf serum, 0.05% sodium azide in phosphate-buffered saline) and labeled for 1.5 h in the dark at 4 °C using PE CD20 (cat: 555623), PerCP CD45 (cat: 340665) and APC CD3 (cat: 557597; all antibodies from BD Biosciences, San Jose, CA, USA). Respective isotype controls were included for all experiments. Following incubation, cells were washed twice with phosphate-buffered saline, fixed in 1% paraformaldehyde in phosphate-buffered saline and 10 000 events were obtained and analyzed using fluorescence-activated cell sorting (FACSCalibur, BD Biosciences) and FACSDiva software (BD Biosciences). CD45+ lymphocytes were gated for CD20-positive and CD20-negative populations, and GFP signal within each population determined infectivity of viruses in each subpopulation. Specificity for CD20-positive cells was assessed via the ratio of CD20+GFP+/CD20−GFP+ using percentages of infected cells (Figure 1). Non-parametric statistics (Wilcoxon signed rank test, Mann–Whitney test) were calculated using GraphPad Prism (GraphPad Software, Inc., La Jolla, CA, USA) and reported as one-tailed P values.