Rituximab induced cytokine release with high serum IP-10 (CXCL10) concentrations is associated with infusion reactions

Abstract

Monoclonal antibody induced infusion reactions (IRs) can be serious and even fatal. We used clinical data and blood samples from 37 treatment naïve patients with chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL) initiating therapy for progressive disease with a single 50 mg dose of intravenous (IV) rituximab at 25 mg/h. Twenty-four (65%) patients had IRs at a median of 78 minutes (range 35–128) and rituximab dose of 32 mg (range 15–50). IR risk did not correlate with patient or CLL characteristics, CLL counts or CD20 levels, or serum rituximab or complement concentrations. Thirty-five (95%) patients had cytokine release response with a ≥4-fold increase in serum concentration of ≥1 inflammatory cytokine. IRs were associated with significantly higher post-infusion serum concentrations of gamma interferon induced cytokines IP-10, IL-6 and IL-8. IP-10 concentrations increased ≥4-fold in all patients with an IR and were above the upper limit of detection (40,000 pg/ml) in 17 (71%). In contrast, to only three (23%) patients without an IR had an ≥4-fold increase in serum concentrations of IP-10 (highest 22,013 pg/ml). Our data suggest that cytokine release could be initiated by activation of effector cells responsible for clearance of circulating CLL cells with IRs occurring in those with higher levels of gamma interferon induced cytokines. These novel insights could inform future research to better understand and manage IRs and understand the role of cytokines in the control of cytotoxic immune responses to mAb.

Graphical abstract

Introduction

Addition of CD20 targeting unconjugated monoclonal antibodies (mAb) to therapy for diffuse large B-cell lymphoma, chronic lymphocytic leukemia/small lymphocytic lymphoma (CLL) and follicular lymphoma significantly improved patient outcomes.^1–5 Rituximab, the prototype CD20 targeting mAb, was approved by the FDA in 1997 and continues to be widely used at the initial recommended dose of 375 mg/m² by intravenous (IV) infusion.⁶ Most patients (77%) with lymphoid malignancies initiating therapy with IV rituximab, and other later generation CD20 targeting mAbs, have a constellation of symptoms and clinical effects constituting an infusion reaction (IR) within 1–2h of initiating therapy which requires interruption of the infusion and symptomatic management.^7–10 Patients usually resume and complete their first rituximab infusion after resolution of an IR, but this complication can be more severe and even fatal.¹¹ The high frequency of IRs, and the potential for serious and life-threatening complications, limits the safety and efficiency of use of these highly effective targeted therapies. Better understanding of IRs could improve prevention and management resulting in safer, more effective, accessible and cost-effective mAb therapy. These data could also provide important insights in the activation of immune cytotoxicity by mAb.

The primary mechanism of action of CD20 targeting mAbs is activation of the innate immune system. Murine and human research strongly supports a major role for fixed tissue macrophage antibody-dependent cellular phagocytosis (ADCP), with lesser roles for complement dependent cytotoxicity and antibody-dependent cellular cytotoxicity.^12–21 Previous studies of rituximab IRs suggested an association with increased circulating concentrations of TNF-α and IL-6, and higher numbers of circulating lymphocytes.^9,22 However, the exact relationship between IRs and patient and CLL characteristics, mAb type, dose and infusion rates remain unclear.

We report a correlative study using patient data and blood specimens from a recently published phase II clinical trial.²³ Treatment naïve CLL patients with progressive disease started therapy with a single slow infusion of a small dose of rituximab to test the hypothesis that IV rituximab rapidly induces an increase in concentration of circulating inflammatory cytokines, and that differences in individual cytokine concentrations correlate with IRs. Our data show that 95% of patients had a significant increase in serum concentrations of inflammatory cytokines and that IRs are closely associated with large increases in the serum concentration of IP-10.

Methods

This IRB approved study used clinical data and blood specimens collected from 37 treatment naïve patients with progressive CLL²⁴ on a phase II clinical trial (ClinicalTrials.gov identifier NCT03788291).²³ Treatment started with a 50 mg of IV rituximab infused at 25 mg/h as monotherapy after standard oral diphenhydramine and acetaminophen premedication. Patients were closely monitored during and after the IV infusion and IRs were graded with Common Terminology Criteria for Adverse Events version 5 (CTCAE). Blood specimens were collected immediately before starting the rituximab infusion, 1h later, at completion of the infusion, and 48h after the initiation of rituximab therapy, and stored as previously described.¹⁹

Blood Cell Analysis

Cells were stained with anti-human CD5 PE-Dazzle 594 (UCHT2, BioLegend), anti-human CD3 APC-Cy7 (UCHT1, BioLegend), anti-human CD19 Brilliant Violet 421 (HIB19, BD Biosciences), rituximab (Genentech), and TO-PRO-3 (Life Technologies), followed by anti-human IgG Alexa Fluor 488 (goat polyclonal, Southern Biotech). CD20 levels were measured after adding a saturating amount of rituximab followed by anti-human IgG Alexa Fluor 488.¹⁹ We used a LSR II flow cytometer and FlowJo 10.5 software (BD Biosciences).²⁵ CLL counts were calculated from the %CD5+/CD3− viable cell events x white blood cell count. CD20 surface molecules/cell calculations used the relative mean fluorescence intensity (RMFI) derived from the geometric MFI (gMFI) of CLL cells/gMFI of T cells and a standard curve generated with Alexa Fluor 488 Molecules of Equivalent Soluble Fluorochrome (MESF) beads (Bangs Laboratories).²⁶

Serum measurements

Complement levels were measured by a reference laboratory (ARUP Laboratories, Salt Lake City, UT), rituximab concentrations using an ELISA kit (Eagle Biosciences, Inc. Nashua, NH)^27–29, and serum cytokine concentrations with a Luminex xMAP technology-based multiplex assay system (Bio-Plex 200, Bio-Rad, Hercules, CA) with Milliplex reagents for detection of 12 human cytokines: IFN-γ, IL-10, IL-18, IL-2, IL-4, IL-6, IL-8, IP-10, MCP-1, MIP-1α, MIP-1β, and TNF-α (MilliporeSigma).

Statistical analysis

Patient characteristics were summarized using medians and ranges for continuous variables and counts and proportions for categorical variables. Peripheral blood parameters were summarized for the entire cohort, and separately based on IR groups, using medians and 95% distribution-free confidence intervals (CI). Changes in peripheral blood parameters for the total cohort were assessed using the paired non-parametric Wilcoxon signed-rank test. Peripheral blood parameters were compared between subjects that experienced an IR versus no IR subjects at each time point using the non-parametric Wilcoxon rank-sum test. Serum cytokine levels were summarized by medians, after imputing low out of range values with 0.5xlower limit of detection and high out of range values with 1+upper limit of detection, and log2-transformation. Cytokine levels were compared between reaction groups at each timepoint using the non-parametric Wilcoxon rank-sum test. To account for the timing of the IR and eliminate potential confounding of subsequent treatment of the reaction infusion, we used Cox models to estimate the association of risk of IR and serum cytokine levels via time-dependent covariates for each cytokine. Hazard ratios for each cytokine were estimated with associated 95% CI. A multivariate Cox model with stepwise selection was used to simultaneous assess all cytokines for independent predictors of risk. All confidence intervals are two-sided, and all hypothesis tests were performed at the two-sided 0.05 level. SAS 9.4 (SAS Institute, Inc. Cary, NC, USA) was used for all analyses.

Results

IRs

Twenty-four of 37 (65%, 95% CI 47.5–79.8) patients had 25 IRs (all CTCAE grade 2) at a median of 78 minutes (range 35–128) after initiation of rituximab infusion (Fig. 1) with a median administered dose of 32 mg of rituximab (range 15–50). After recovery from their IR, rituximab infusion was resumed in all 21 patients who had not yet completed therapy. One patient had a second IR immediately following completion of the rituximab infusion and 80 minutes after the start of the first IR. All patients completed the 50 mg rituximab infusion at a median time from initiation of therapy of 150 minutes (range 119–195).

Figure 1. Rituximab administration and IRs.

Thirty-seven patients received their first treatment with 50 mg of intravenous rituximab at a planned infusion rate of 25 mg/h. Horizontal bars of the swimmers’ plot represent individual patients. Grey shading representing the time from initiation of the infusion to completion of treatment. Twenty-four patients had an IR with time of onset indicated by a red triangle. In 21 patients, a red dot indicates the time of re-starting the rituximab infusion.

Patients with IRs presented with chills/rigors (n=14), nausea (n=10), dizziness (n=8), cutaneous manifestations (rash, itchiness, hyperemia)(n=3), pharyngeal discomfort/dysphagia (n=2), low back pain (n=2), dyspnea/cough/chest tightness (n=2), headache (n=2), and fatigue (n=1). Later onset clinical manifestations were chills/rigors (n=3), hypotension (n=3), nausea (n=2), vomiting (n=2), bradycardia (n=1), syncope (n=1), cutaneous manifestations (n=1), low back (n=1) and hip pain (n=1).

Management of IRs

Rituximab infusion was interrupted at the onset of the IR in the 21 patients with an IR during treatment (Fig. 1). All IRs were treated with rapid IV administration of normal saline and 22 patients were treated with hydrocortisone (100 mg IV). Additional therapies included famotidine (20 mg IV)(n=23), diphenhydramine (50 mg IV)(n=9), meperidine (12.5–25 mg IV)(n=13), oxygen (n=2), albuterol nebulizer (n=1), and ondansetron (n=2).

Patient and CLL parameters

Analysis of patient and CLL characteristics (Table 1) showed no correlations with the risk of IRs.

Table 1:

Patient Description

	Patients n=37
Age at enrollment, median (min, max)	68 (42, 80)
Gender, n (%)
Female	14 (37.8)
Male	23 (62.2)
Race, n (%)
White	36 (97.2)
Black/African American	1 (2.8)
Performance status (ECOG), n (%)
0	18 (48.6)
1	16 (43.2)
2	3 (8.2)
Stage (Rai) at diagnosis, n (%)
0	6 (16.2)
I	10 (27.0)
II	11 (29.7)
III	6 (16.2)
IV	1 (2.7)
Unknown	3 (8.2)
Cytogenetic Defects1, n (%)
13q14 deletion	15 (40.5)
None	5 (13.5)
Trisomy 12	6 (16.2)
11q22.3 deletion	6 (16.2)
17p13 deletion	5 (13.5)
IGHV somatic hypermutation, n (%)
Mutated (>2% vs. germline)	15 (40.5)
Unmutated	22 (59.5)
TP53 mutation, n (%)
Positive	8 (21.6)
Negative	29 (78.4)
TP53 mutation AND 17p13 deletion, n (%)	5 (13.2)
NOTCH1 mutation, n (%)
Positive	9 (24.3)
Negative	28 (75.7)
SF3B1 mutation, n (%)
Positive	4 (10.8)
Negative	33 (89.2)

¹Cytogenetic defects on interphase fluorescent in situ hybridization analysis reported using a hierarchical stratification.⁴⁶

CLL cell counts

Pre-treatment absolute CLL count was not significantly associated with risk of IR (median 77.7 vs. 56.6×10⁹/L, p=0.72). Rituximab infusion caused an 83.7% decrease in the median circulating CLL cell counts at 1h with no subsequent change at the end of the infusion (Fig. 2A). We did not detect a difference in changes in CLL cell counts between patients with or without IRs (1h p=0.62, post p=0.43)(Fig. 2A). Patients with an IR had a decrease in median CLL count of 60×10⁹/L which was not significantly different from a median decrease of 45×10⁹/L in those without an IR (p=0.61). Pre-treatment peripheral blood absolute lymphocyte count (ALC) >50×10⁹/L in CLL patients has been reported to be a risk factor for rituximab induced IRs.⁹ We compared the rates of IRs in patients with pre-treatment ALC ≤50×10⁹/L (n=13) with those with ALC >50×10⁹/L (n=24) and found no significant difference (61.5% vs. 66.7%, p=1.0).

Figure 2. Peripheral blood parameters.

CLL cells and serum collected from patients at baseline (BL), 1h after initiation of rituximab infusion (1h) and at completion of the rituximab infusion (Post) were examined. Plots represent the median values with error bars showing 95% confidence intervals either for all patients (black line, n=37) or the subsets of patients with (red, n=24) or without (blue, n=13) rituximab IRs. A: Circulating CLL counts measured as a percentage of baseline. B: CLL cell membrane CD20 levels (x10³ molecules/cell). C: Estimated number of CLL cell membrane CD20 molecules (x 10¹²/L) calculated as the product of the absolute CLL cell count (x 10⁹/L) and CLL cell CD20 levels (x 10³ molecules/cell). D: Serum rituximab levels measured in mg/ml. E: Total complement levels (CH50) measured in Units/ml (normal range 38.7–89.9).

CD20 Antigen

Median CLL cell membrane CD20 levels (x10³ molecules/cell) decreased throughout the rituximab infusion from 7.30 at baseline to 4.97 at 1h (p=0.002) and then to 2.72 at completion of the infusion (p<0.0001 versus baseline and 1h)(Fig. 2B). There were no significant differences between patients with or without an IR (baseline 7.22 vs 7.39; p=0.97: 1h 4.40 vs 5.70; p=0.54: post 2.72 vs 3.02; p=1.0)(Fig. 2B).

The estimated number of circulating CLL cell membrane CD20 molecules (x10¹² molecules/L) was calculated as the product of the circulating CLL count and CD20 molecules/cell.³⁰ Rituximab therapy significantly decreased the median from 496.4 at baseline to 55.4 at 1h (p<0.0001) and then to 25.0 post infusion (p<0.0001 compared to baseline, not significantly different from 1h p=0.09)(Fig. 2C). There was no significant difference in patients with or without an IR in the median estimated number of CLL cell membrane CD20 molecules at baseline (559.8 vs 485.3 p=0.54) or the median decrease from baseline after treatment (1h 345.5 vs 403.9; p=0.87: post 536.1 vs 421.3; p=0.50)(Fig. 2C).

Serum rituximab and complement concentrations

The median serum rituximab concentration increased throughout the infusion (Fig. 2D). There were no significant differences in rituximab concentration in patients who had an IR compared to those without an IR (p>0.4 for all comparisons)(Fig. 2D). Rituximab infusion was associated with a progressive decrease in the median serum complement level measured with the CH50 assay (73.4 to 65.1 units/ml, p<0.0001)(Fig. 2E). Changes in CH50 levels were not significantly different in patients with an IR compared to those without an IR (p>0.4 at each time point)(Fig. 2E).

Serum cytokine concentrations

Twelve cytokines and chemokines were chosen for study based on data from previous studies in CLL patients^9,31,32 and analysis of serum samples from a patient treated on protocol who was excluded because of a reclassification of his diagnosis as a non-CLL CD5+ B-cell lymphoproliferative disorder. Measurable serum cytokine concentrations (pg/mL) in all 37 patients are shown in Fig S1 and summarized in Fig. 3A. Thirty-five (95%) patients had a ≥4-fold increase in concentration of at least one cytokine compared to pre-treatment levels (Fig. 3B). IRs did not occur in either of the two patients without a ≥4-fold increase in at least one serum cytokine concentration. Rituximab infusion had minimal observed impact on serum concentrations of IL-4 and IL-18, and IL-2 concentrations were below the range of detection at all time points (Fig. 3A&B). Median cytokine concentrations returned to near baseline within 48h of initiation of rituximab therapy (Fig. 3A).

Figure 3. Serum Cytokine Concentrations.

A: Heat map showing median serum concentrations (pg/ml) for twelve cytokines at baseline (BL), 1h after initiation of rituximab infusion (1h), at the end of the infusion (Post) and 48h after initiation of the infusion (48h) in 37 patients. B: Number of patients with ≥4-fold increases from baseline for each cytokine at 1 h, post and at either time point (Total).

We then compared cytokine responses to rituximab therapy in patients with or without IRs (Fig. 4A). Analysis of all patients with IRs irrespective of time of onset and therapy of the reaction showed significantly higher median concentrations of IP-10 and IL-6 post-infusion. Assessment of the association between the risk of IRs and preceding cytokine concentrations using time-varying covariates in Cox models (Fig. 4B) showed that increased serum concentrations of IL-6, IL-8 and IP-10 were associated with increased risk of IRs. Multivariate modelling indicated that these cytokines are not independent predictors of IRs. Serum IP-10 concentration trajectories in patients with and without an IR appear different on a heatmap analysis (Fig. 5). All 17 patients with IP-10 serum concentrations above the limit of detection (40,000 pg/ml) post infusion had an IR and all 24 patients with an IR had a ≥4-fold increase in serum IP-10 concentrations. Of the 13 patients without an IR, the highest measured serum IP-10 concentration was 22,013 pg/ml and only 3 (23%) of these patients had ≥4-fold increases in serum IP-10 concentrations.

Figure 4. Serum Cytokine Concentrations and Infusion Reactions.

A: Median concentration of each serum cytokine (pg/ml) plotted separately for patients with IR (n=24, red) and without an IR (n=13, black) with error bars showing 95% distribution free confidence intervals. * indicates cytokines significantly associated with risk of IRs as determined by time to reaction Cox model analysis with detailed results shown in B.

Figure 5. Serum IP-10 Concentrations and Infusion Reactions.

Heat map showing serum concentrations (pg/ml) of IP-10 for 37 patients (rows) at BL, 1h, post, and 48h after initiation of the infusion. Within reaction groups (black no reaction, red IR), patients are sorted by increasing post IP-10 serum concentrations.

Discussion

Initial treatment of CLL patients with slow infusion (50% of standard initial rate) of low dose rituximab (~7% of standard 375 mg/m²) did not decrease the rate of IRs. Although 95% of patients had an inflammatory cytokine response, only 65% had a IR. The risk of IR was only significantly associated with increased of serum concentrations of IP-10, IL8 and IL6.

Our data suggest that the risk of IRs is not associated with the pre-treatment circulating CLL tumor burden, rituximab mediated CLL cell clearance, or circulating CD20 antigen load. This finding is similar to our previously reported data on CLL patients treated with ofatumumab.³⁰ IR risk was not associated with administered dose of rituximab or serum rituximab concentrations and IRs were only observed after at least 35 minutes of rituximab infusion. These data suggest a threshold amount of rituximab required for activation of IRs, or alternatively that the mechanism responsible for an IR requires at least 35 minutes for activation. Decreases in serum complement concentrations, used as a marker of immune activation, were not associated with the risk of IRs. This suggests a limited role for complement in inducing IRs as previously reported for alemtuzumab.³³

Rituximab caused a ≥4-fold increase in median concentrations of nine of twelve measured inflammatory cytokines in most patients (Fig. 3B). The sources of these cytokines and how rituximab increases their serum concentrations are not known. CLL cells can be a source of serum MIP-1α, MIP1β and IL-8³⁴ and rituximab opsonized CLL cells are highly effective at activating innate immune effector cell cytokine secretion via Fc receptor signaling.^35–39 The minimal changes in serum concentrations of IL2 and IL4 in our study suggest a lesser role for T cell derived cytokines in the generation of the cytokine response to rituximab infusion.³⁵ These data, and the important role of macrophage ADCP as a mechanism of rituximab mediated cytotoxicity, suggests that fixed tissue macrophages in the liver and spleen could be an important source of serum cytokines after administration of rituximab.

There is limited data on the mediators and mechanisms of rituximab induced IRs. Our study suggest that increased serum concentrations of IP-10 are the best predictors of IR. IRs were also associated with significantly increased serum concentrations of IL-6 and IL-8 as previously reported^9,40. Serum concentrations of TNFα and INFγ were increased in most patients following infusion of rituximab and were not significantly different in patients with IRs (Fig. 3C). It is biologically plausible that the significantly increased levels of IL-8, IL-6 and IP-10 are mechanistically interlinked (e.g. induced by γinterferon) and drive the inflammation that causes rituximab induced IRs. This hypothesis is supported by our finding on multivariate modeling that these cytokines are not independent predictors of IRs.

Cytokine release syndromes (“cytokine storms”) have previously been described with activation of inflammation by therapeutic interventions (chimeric antigen receptor T cells), viral infections (SARS-CoV2), bacterial sepsis, and immune disorders (primary hemophagocytic lymphohistiocytosis, macrophage activation syndrome).^36,41 In this study, rituximab induced IRs were distinct from previously described cytokine release syndromes with shorter clinical duration, lower risk of organ damage, no increases in IL-18 concentrations, and short duration of increases in other serum cytokine concentrations.

Current recommendations on prevention and management of IRs in consensus and manufacturers guidelines include use of corticosteroid pre-medications for patients with higher pre-treatment absolute lymphocyte counts and the routine use of prophylactic antihistamine and acetaminophen.^10,42 These recommendations are based on limited data and have not been rigorously tested for efficacy. Our data suggests that the risk of IR is not increased by higher ALCs and is not decreased by use of smaller doses and slower infusion of rituximab, a finding compatible with a previous report.¹⁰ Identification of the specific molecular mediators of IRs and a better understanding of the role of inflammatory cytokines in the mechanism of action of mAb cytotoxicity could facilitate the development of targeted interventions to prevent IRs. This is especially important because IR rates and severity are higher than those observed for rituximab for next generation anti-CD20 mAb therapies such as ofatumumab and obinutuzumab.^43,44 Determining if IP-10 has a role in causing IRs would be especially important because pathway inhibitors including a clinically tested CXCR3 antagonist are available for evaluation.⁴⁵

This study design provided a unique opportunity to prospectively obtain high quality clinical and laboratory data on IRs in patients with CLL receiving their first dose of rituximab as initial monotherapy. All patients were treatment naïve, had uniform pre-medication which specifically excluded corticosteroids, and rituximab monotherapy was infused with infusion pumps at a standard rate with careful continuous monitoring and precise measurement of time of onset and recovery from IRs. Our data are constrained by limitations inherent to clinical research on the number and size of blood samples collected which precluded repeat cytokine measurements in patients with levels of IP-10 and MIP-1β above the upper limit of the measurement range. IV corticosteroid treatment of patients with IRs could have altered their cytokine serum concentrations. This study used a lower and slower infusion of rituximab than standard of care and our data could underestimate the rate of IR occurring in usual practice. We could not study cytokine responses to a second IV dose of rituximab because the clinical study design required that all subsequent treatment be administered subcutaneously. Our data is also limited to rituximab and could not be fully representative of the often more severe reactions to defucosylated anti-CD20 mAbs such as obinutuzumab.

We conclude that that rituximab IRs are likely caused by production of inflammatory cytokines by cytotoxic immune effector cells activated by rituximab independent of the pre-treatment amount of circulating CLL cells or CD20 molecules, or efficacy of rituximab in reducing the circulating CLL cell count. IRs occurred with a delay of at least 35 minutes from initiation of therapy concurrent with a rapid decrease in circulating CLL cells. This observation suggests that cytokine release could be a result of activation of immune effector cells responsible for CLL cell elimination. Our data provides novel insights into the potential mechanism of IRs and could form the basis of further research to determine methods to control this clinically important and therapy limiting phenomenon.

Highlights.

• Small doses of IV rituximab rapidly decrease circulating CLL cell counts

• IV rituximab increases serum inflammatory cytokine concentrations in CLL patients

• Rituximab infusion reactions associated with high serum concentrations of IP-10