Predictive Value of Cytokines and Immune Activation Biomarkers in AIDS-Related Non-Hodgkin Lymphoma Treated with Rituximab plus Infusional EPOCH (AMC-034 trial)

The risk for developing B-cell non-Hodgkin lymphoma (NHL) is significantly and markedly increased in persons living with HIV infection (1–5). The introduction of combination antiretroviral therapy (HAART) has had a significant impact on overall survival of persons living with HIV infection (6–10). The incidence of AIDS-NHL has decreased in the HAART era, but not to the same extent as that of Kaposi sarcoma or other AIDS-defining conditions. In addition, the widespread availability of HAART appears to have had a differential effect on the incidence of AIDS-NHL subtypes, with a marked decrease in the incidence of primary central nervous system lymphoma (PCNSL), while that of other forms of AIDS-NHL, such as Burkitt lymphoma or diffuse large B-cell lymphoma (DLBCL), either has not decreased or has remained unchanged (7, 11). Therefore, lymphoma remains a significant clinical problem in the HAART era. In fact, NHL appears to be the most common AIDS-related cancer in populations with ready access to HAART, and remains a significant cause of morbidity and mortality in HIV⁺ persons in the post-HAART era (12, 13).

B-cell hyperactivation, as well as loss of regulation of Epstein-Barr virus (EBV)-infected B cells, are believed to play important roles in the development of AIDS-NHL (14–17). We and others have shown elevated serum/plasma levels of several B-cell–stimulatory cytokines, including IL6, IL10, and CXCL13, are present over a period of several years before the diagnosis of AIDS-NHL (18–29). Elevated levels of circulating IL6 or IL10 also are seen after AIDS-NHL diagnosis (30–31). AIDS-NHL cell lines also are known to produce cytokines, including IL6 and IL10 (32). In addition, we saw elevated levels of the expression of activation-induced cytidine deaminase (AICDA), a DNA-mutating enzyme, in circulating mononuclear cells, preceding the diagnosis of AIDS-NHL (33). AICDA expression has also been reported to be elevated in B cells and lymphoma cells infected with HCV (34). HIV can directly induce B-cell AICDA expression, as well as their secretion of several cytokines (IL6 and IL10) and surface molecules (CD23), in vitro (17, 35).

In the present study, we evaluated plasma levels of several B-cell activation-associated molecules (sCD23, sCD27, sCD30, and IgE) and B-cell–stimulatory cytokines (IL6, IL10, and CXCL13), in persons who had an AIDS-NHL diagnosis, before and after the initiation of treatment, with the aim of better defining postdiagnosis, pretreatment levels of these molecules in AIDS-NHL, and to determine how levels of these immune system stimulatory molecules are affected by treatment for AIDS-NHL. We found that AIDS-NHL patients had high pretreatment plasma levels of several B-cell activation-associated molecules (IL6, IL10, CXCL13, sCD27, and sCD30). In addition, treatment of NHL resulted in a rapid decrease in plasma levels of most of these molecules, with decreased levels persisting for one year following the completion of treatment. Importantly, pretreatment levels of some of these molecules were associated with response to lymphoma therapy, as well as overall survival.

Of the 106 AIDS NHL patients enrolled in an AIDS Malignancy Consortium (AMC) trial, AMC protocol #034 (AMC-034), which compared infusional combination chemotherapy (EPOCH: etoposide, vincristine, doxorubicin, cyclophosphamide, and prednisone) with concurrent or sequential rituximab (36), plasma specimens were available from 69 patients with intermediate- or high-grade HIV-associated B-cell NHL (50 patients had DLBCL, 17 Burkitt lymphoma, and 2 were classified only as lymphoma). The median age of lymphoma patients was 42.6 ± 8.8 years. Lymphoma patients had a median HIV plasma level of 9908 [inter-quartile rage (IQR) = 492.5–45,660], and a median CD4 number of 187 cells/mm³ (IQR = 82–333). Plasma was collected before the initiation of therapy, at the end of the first cycle (within a week or less of treatment), and at 6 months and one year following the completion of treatment. Clinical responses were defined as described in the report detailing the AMC-034 trial results (36).

Details regarding the treatment protocol can be found in Sparano and colleagues (36). Clinical responses were defined by the International Response Criteria for Non-Hodgkin Lymphoma (which uses anatomic but not functional imaging). Toxicity was graded according to the National Cancer Institute Common Toxicity Criteria (Version 2.0). Response was evaluated after every two cycles of EPOCH therapy (with computerized tomography of the chest, abdomen, and pelvis) and continued for two cycles beyond achieving a CR (for a minimum of four and maximum of six cycles), including after completion of R-EPOCH in the concurrent arm, and after completion of EPOCH alone and by rituximab alone in the sequential arm. No patients received rituximab when not approved as part of the study protocol. All patients were required to have bone marrow biopsy and lumbar puncture for cerebrospinal fluid cytologic examination at baseline. A repeat bone marrow biopsy was required if the original study demonstrated lymphomatous marrow involvement, and if the physical examination and imaging studies were consistent with a complete response (36).

Plasma levels of B-cell stimulatory cytokines and molecules associated with immune system activation were assessed by ELISA. IL6 was measured using an ultrasensitive assay (Biosource/Invitrogen), with color development time extended to 40 minutes to ensure consistent low-level detection (detection limit = 0.2 pg/mL). IL10 was measured using a human IL10-specific assay (Biosource/Invitrogen) that does not cross-react with EBV viral IL10 (21), modified to increase sensitivity by extending the standard curve (detection limit = 2 pg/mL), increasing sample incubation time to 3 hours, and performing all incubations on a microtiter plate rotator (500 rpm). CXCL13/BCA-1 was measured using the R&D Systems ELISA kit according to the manufacturer's protocol, with a 1:2 dilution (detection limit = 7.8 pg/mL). sCD27 was determined using the PeliKine-compact ELISA kit and Toolset according to the manufacturer's protocol (CLB/Sanquin, the Netherlands), with 1:20 dilutions on all HIV+ samples (detection limit = 32 U/mL, taking dilution into account). Assays for sCD23 (detection limit = 13 U/mL) and sCD30 (detection limit = 6 U/mL) were performed according to the manufacturer's protocols (Bender MedSystems USA). Total plasma IgE was determined utilizing the CIA-7.12 and CIA-4.15 monoclonal antibodies (37) as previously described (38), with the following modifications: IgE ELISA plates were blocked with 10% fetal bovine plasma in PBS-Tween buffer, and all plasma samples were diluted 1:10 using PBS-Tween buffer before addition to the ELISA plate (19). Diluted sera and all subsequent reagents were added at 50 μL per well, and all incubations were performed on a microtiter plate rotator (500 rpm). The IgE standard was pooled normal plasma (generously provided by Drs. Andrew Saxon and Ke Zhang); when referenced to the WHO IgE standard NIBSC 75/502 (which is also pooled human sera), the mean conversion factor was 0.67 ng per IU. The concentration of the lowest IgE standard was 0.8 ng/mL; taking the dilution into account, the lower limit of detection in plasma samples was 8 ng/mL. Plasma samples used for the measurement of IL6, IL10, sCD27, sCD30, and IgE were frozen and thawed once. Samples used for the measurement of CXCL13 and sCD23 were frozen and thawed twice.

The Wilcoxon rank-sum test was used to compare biomarker levels of complete responders with patients who did not achieve a complete response with treatment for lymphoma. To assess their association with outcome measures (complete response, overall survival and progression-free survival; PFS), levels for CXCL13, CD23, CD27, CD30, IL6, and LDH were dichotomized at their median value. Fisher exact test was used to compare those who achieved complete response and those who did not with respect to international prognostic index (IPI) score (age-adjusted) and each biomarker. Those biomarkers that were associated with complete response at the 0.05 significance level were incorporated into a stepwise logistic regression model.

Proportional hazards models were used to evaluate the association of each individual biomarker with overall survival and PFS. Those factors associated with overall survival and PFS at the 0.05 level were incorporated into a stepwise proportional hazards model.

Plasma levels of B-cell–stimulatory cytokines (IL6, IL10, CXCL13), and of molecules associated with B-cell activation (sCD23, sCD27, sCD30, IgE), were measured by ELISA in specimens collected, pre- and posttreatment, from persons enrolled in an AIDS Malignancy Consortium trial comparing infusional combination chemotherapy (EPOCH: etoposide, vincristine, doxorubicin, cyclophosphamide, and prednisone) with concurrent or sequential rituximab (AMC protocol #034). There was a negative correlation between CD4 counts and CXCL13 levels (P = 0.042); correlations of CD4 with the other biomarkers were not significant.

A central aim of this study was to determine whether there is a change in the levels of these cytokines and immune activation molecules following treatment for AIDS-NHL. In order to study this, we measured plasma levels of these molecules before (Pre-Rx), during lymphoma treatment (Post-Rx), as well as 6 months (6mo FU) and one year (1 year FU) after lymphoma treatment. Treatment was seen to result in a marked decrease in plasma CXCL13 levels following the initiation of treatment (P = 0.005, Wilcoxon signed rank test; Fig. 1); this decrease was a consequence of treatment and not of survival. This decrease was maintained over time, with decreased plasma CXCL13 levels seen at 6 months and 1 year after the completion of treatment (P = 0.016 and 0.031, respectively, Wilcoxon signed rank test).

In contrast, plasma levels of IL6 or IL10 were not seen to decrease significantly following treatment (P = 0.421 and 0.492, respectively), nor at 6 months (P = 0.688 and 0.625, respectively) or 12 months posttreatment (P = 1.00 and not evaluable, respectively; Fig. 1).

Compared with pretreatment, sCD30 plasma levels were seen to decrease significantly (P < 0.004, Wilcoxon signed rank test; Fig. 2), and appeared to remain at lower levels at 6 months and 1 year after treatment completion, although this was not statistically significant (P = 0.297 and 0.063, respectively).

Similarly, plasma sCD27 levels were significantly decreased following the initiation of treatment (P < 0.001; Fig. 2), and remained lower at 6 months (P = 0.016) and 1 year after the completion of lymphoma treatment (P = 0.063), although this decrease was not statistically significant at one year posttreatment.

Plasma levels of sCD23 were seen to decrease significantly (P < 0.001, Wilcoxon signed rank test) after the initiation of lymphoma treatment and at 6 months posttreatment (P = 0.16), but not at 1 year after the completion of treatment (P = 0.219; Fig. 2).

Plasma levels of IgE appeared unchanged after lymphoma treatment (Fig. 2).

In this study, patients were treated with EPOCH with concurrent rituximab, or administration of rituximab after EPOCH treatment was complete. There were no significant differences seen between patients receiving concurrent or sequential rituximab, in terms of plasma levels of these biomarkers (not shown).

Lower levels of CXCL13, IL10, IL6, and LDH, as well as IPI scores, were significantly associated with complete response to therapy, in univariate analyses (Table 1). Similarly, when we assessed the association of these biomarkers with any clinical response (complete and partial responders vs. nonresponders), CXCL13, LDH, and IPI remained significant predictors of a clinical response and detectable IL10 was associated with nonresponders (not significant; Table 2). In addition, when a stepwise logistic regression was used to assess the relative contribution of these factors to complete response, only CXCL13 was seen to be associated with complete response (P = 0.003; OR = 5.5, 1.81–16.68).

We also looked at the association of these biomarkers with overall survival or PFS. In univariate analyses, the following factors were significantly associated with overall survival: IPI, CXCL13, IL6, IL10, and LDH (Table 3). However, when they were incorporated into a stepwise proportional hazard model for overall survival, only IL6 (P = 0.010, OR = 0.30, 0.12–0.75) was associated with overall survival.

The following factors were significantly associated with PFS in univariate analyses: IPI, CXCL13, IL6, IL10, and LDH (Table 3). However, when these markers were incorporated into a stepwise proportional hazard model, only IL10 (P = 0.024; OR = 2.86, 1.15–7.15) was significantly associated with PFS. Furthermore, patients with CXCL13, IL6, or IL10 levels higher than the median levels for the NHL group as a whole had lower overall survival and PFS, in Kaplan–Meier analyses (Fig. 3).

We also assessed the association of cytokine/activation markers with response to treatment by NHL pathologic types (DLBCL and Burkitt lymphoma). With respect to OS and PFS, the direction of the HRs was the same for Burkitt lymphoma and DBLCL, but frequently achieved statistical significance only when all patients are included (not shown). Therefore, while there was insufficient statistical power to definitively determine whether there was a difference in the results observed for these two NHL subgroups, it does not appear that there were marked differences seen between DLBCL and Burkitt lymphoma.

In this study, we found that plasma levels of several molecules that are associated with immune activation and inflammation are detectable in those who have untreated AIDS-NHL, and that plasma levels of some of these molecules (sCD23, sCD27, sCD30, CXCL13) showed marked reductions after EPOCH and rituximab treatment. We did not see any differences in the levels of these biomarkers between the concurrent or sequential EPOCH and Rituximab treatment groups.

Most notably, we saw that pretreatment levels of CXCL13, IL6, IL10, and LDH, as well as IPI, were significantly lower in those patients who went on to have complete responses to treatment. After conducting a stepwise logistic regression analysis, CXCL13 was the only factor that significantly correlated with subsequent treatment response. Similarly, multivariate analyses showed that only pretreatment IL6 levels were associated with overall survival, and IL10 levels with PFS.

We also assessed the association of cytokine/activation markers with response to treatment by the major NHL pathologic types (DLBCL and Burkitt lymphoma) included in the AMC-034 study. Although there was insufficient statistical power to definitively determine whether there was a difference in the results observed for these two NHL subgroups, it did not appear that there were marked differences seen between DLBCL and Burkitt lymphoma.

Overall, these results suggest that plasma levels of CXCL13, IL6, and IL10 have significant potential as prognostic biomarkers for AIDS-NHL, and may add additional information over LDH or IPI, which are commonly used to assess prognosis (39). Certainly, the prognostic value of these cytokines needs to be confirmed in larger studies. In addition, it is important to determine whether the levels of these cytokines reflect tumor burden, or alternatively, if they are markers for the inability of patients to respond to therapy for other reasons relating to their poor health. Also, further studies are needed to define the prognostic value of measuring these cytokines when measured after the initiation of treatment for AIDS-NHL. However, these molecules show promise as new tools for the assessment of prognosis, and potentially, for the selection of treatment regimens for AIDS-NHL.

CXCL13 is a chemokine that directs the normal trafficking of B cells (40). It is expressed by T-follicular helper cells, dendritic cells, and stromal cells in secondary lymphoid tissue (41). Plasma levels of the chemokine, CXCL13, are elevated during HIV infection (42), and decrease with antiretroviral drug treatment (43). Other reports indicate that there are abnormalities in the CXCR5/CXCL13 system during HIV infection, including loss of expression of CXCR5 on mature B cells (44), and expression of CXCL13 by recirculating B cells (45). Together, these observations raise the possibility that the CXCR5/CXCL13 system may contribute to the abnormalities that are seen in the B-cell compartment during HIV infection, and thus could be involved in the genesis of AIDS-NHL. CXCL13 has been shown to be associated with Sjögren disease, in which CXCL13 contributes to the organization of ectopic reactive lymphoid tissue (46). In addition, elevated serum levels of CXCL13 have been seen before the diagnosis of non-AIDS NHL (47, 48), and CXCL13 and/or CXCR5 have been shown to be associated with several subtypes of non-AIDS–related B-cell lymphomas (49–50).

IL6 and IL10 are inflammation-associated cytokines that are secreted by monocytes, lymphocytes, and other cell types, and can enhance B-cell proliferation, survival, and antibody production. IL6 and IL10 are known to be elevated before lymphoma diagnosis, and their elevated levels are associated with risk for the development of NHL in HIV+ persons (19–29). It is unclear how IL6 and IL10 are contributing to the development of lymphoma. They may be directly promoting the growth and/or viability of cancer cells and/or they may affect other immune cells, presumably creating a favorable environment for the development or growth of lymphoma cells. In addition, they may be secreted by the tumor cells. Thus, the reduction of these cytokines seen following treatment may be due to loss of tumor cells, as well as to the loss of tumor-reactive cells. Alternatively, the loss of cytokines maybe due to survival time bias, as partial responders and nonresponders are more likely not to survive.

In prior work, we reported that serum/plasma levels of sCD23 were significantly elevated some years before the development of AIDS-NHL, but sCD23 levels did not differ between AIDS-NHL cases and controls when measurements were made closer (<1 year) to the time of lymphoma diagnosis (19). Therefore, this molecule seems to be elevated several years before lymphoma diagnosis, but then drops as NHL diagnosis is approached. In this sense, the observed levels of sCD23 seen in NHL patients in this study are consistent with a progressive decrease in plasma sCD23, going from elevated several years before NHL diagnosis to decreased postdiagnosis. It is possible that sCD23 plays an etiologic role in early events in lymphomagenesis, but not in supporting the progressive growth of these cancers. These results are consistent with the known role of sCD23 in promoting IgH class switch recombination (CSR), a molecular event thought to contribute to the genesis of lymphomagenic chromosomal translocations that lead to Burkitt lymphoma (17).

As mentioned above, we and others have reported that these cytokines and molecules are elevated preceding AIDS-NHL (19–29). The results presented here extend, and are generally in agreement with, those prior studies, and support the notion that a B-cell stimulatory environment is associated with the development and progression of AIDS-NHL. In addition to this, some of these AIDS-NHL–associated plasma molecules, especially CXCL13, IL6, and IL-10, appear to have potential value as indicators of subsequent response to NHL treatment and survival.

No potential conflicts of interest were disclosed.

Conception and design: M. Epeldegui, J.Y. Lee, D.P. Widney, D. Regidor, R. Mitsuyasu, R.F. Ambinder, O. Martínez-Maza

Development of methodology: M. Epeldegui, D.P. Widney, O. Martínez-Maza

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Epeldegui, L.I. Mapantay, R. Mitsuyasu, J.A. Sparano, O. Martínez-Maza

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Epeldegui, J.Y. Lee, D. Regidor, J.A. Sparano, R.F. Ambinder, O. Martínez-Maza

Writing, review, and/or revision of the manuscript: M. Epeldegui, J.Y. Lee, A.C. Martínez, R. Mitsuyasu, J.A. Sparano, R.F. Ambinder, O. Martínez-Maza

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Epeldegui, J.Y. Lee, A.C. Martínez, R. Mitsuyasu, J.A. Sparano, O. Martínez-Maza

Study supervision: M. Epeldegui, J.A. Sparano, O. Martínez-Maza

This work was supported by grants from the NIH (U01-CA-121947, R01-CA-121195, and R01-CA-168482). This work was carried out in the facilities of the UCLA AIDS Institute, which are supported, in part, by funds from the NIH grant AI-028697, UCLA Center for AIDS Research (CFAR).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.