Alignment of Practices for Data Harmonization across Multi-Center Cell Therapy Trials: A Report from the Consortium for Pediatric Cellular Immunotherapy

Abstract

Immune effector cell (IEC) therapies have revolutionized our approach to relapsed B cell malignancies, and interest in the investigational use of IEC is rapidly expanding into other diseases. Current challenges in the analysis of IEC therapies include small sample sizes, limited access to clinical trials, and a paucity of predictive biomarkers of efficacy and toxicity associated with IEC therapies. Retrospective and prospective multi-center cell therapy trials can assist in overcoming these barriers through harmonization of clinical endpoints and correlative assays for immune monitoring, allowing additional cross trial analysis to identify biomarkers of failure and success. The Consortium for Pediatric Cellular Immunotherapy (CPCI) offers a unique platform to address the challenges described above by delivering cutting-edge cell and gene therapies for children through multi-center clinical trials. Here we discuss some of the important pre-analytic variables, such as biospecimen collection and initial processing procedures, that affect biomarker assays commonly used in IEC trials across participating CPCI sites. We review the recent literature and provide data to support recommendations for alignment and standardization of practices that can affect flow cytometry assays measuring immune effector function and that can affect interpretation of cytokine/chemokine data. We also identify critical gaps that often make parallel comparisons between trials difficult or impossible.

INTRODUCTION

Cellular immunotherapy using autologous or allogeneic T cells genetically modified to express chimeric antigen receptors (CARs) or antigen-specific T cell receptors (TCRs) has proven to be a transformative therapy for some cancer patients. Pivotal Phase I/II pediatric clinical trials have led to approval of a CD19-directed CAR, tisagenlecleucel (Kymriah), as the first “living drug” for children with relapsed or refractory B-cell acute lymphoblastic leukemia (B-ALL) [1,2]. Although these trials provided a proof of concept for IEC efficacy and safety, the widespread clinical utilization of cellular therapies in pediatrics beyond B cell malignancies still faces a number of barriers, including a lack of predictive biomarkers for toxicity and efficacy, small patient cohorts, and uniform access to trials [3–7]. As broadened application of immunotherapies occurs, there is a significant need for in-depth characterization of IECs and the immune landscape, to identify mechanisms of resistance and to help guide development of the next generation of T cell therapies [8–11].

Multi-center research extends increased access to cell therapy trials and provides larger sample sizes for more effective experimental and clinical analyses. However, the ability to compare correlative data across sites, within the same study or across studies, remains a major challenge. In the context of hematopoietic cell transplant (HCT) studies, previously published work has emphasized the need to harmonize trial design, including selection of similar endpoints, immune monitoring strategies, and timepoints for collection of research samples [3,12–17]. Harmonization is critical to improve reproducibility in analyte measurement, reduce variability, and ultimately yield greater confidence in correlative data and results. Robust datasets that accommodate side-by-side comparison in early-phase cell therapy trials are necessary to establish biomarkers that correlate with efficacy and clinical criteria for grading and treatment of immune effector-related toxicities [18,19]. Acquiring such datasets will require consistency in the handling of biologic samples as well as the use of analytic variables that can influence the generalizability of biomarkers across trials.

Evaluation of biologic samples collected from diverse sources, including peripheral blood, bone marrow, cerebrospinal fluid, tumor tissue, and tumor-infiltrating lymphocytes, is key to identifying biomarkers of safety and efficacy in cell therapy trials. Common assays used to evaluate samples are 1) plasma or serum cytokine analysis, 2) phenotypic and functional characterization of immune effector cells by flow cytometry [20–25], and 3) detection of persistence of IEC by Polymerase chain reaction (PCR) [1]. Consistency and reproducibility of these assays can be affected by many factors: pre-analytic (patient and sample related), analytic (assay related) and post-analytic (data related) (Figure 1). Many recently published studies have highlighted the challenges of reducing analytic and post-analytic variability across multiple assay platforms, studies, or sites [26–30]. One way to reduce analytic variability is to perform correlative experiments using batched samples at central sites with specialized instrumentation and expertise. However, pre-analytic variables such as biospecimen collection and processing procedures can also affect downstream results [31,32]. For example, phenotypic and functional markers of cell differentiation, activation, and exhaustion are most sensitive to pre-analytic variables such as the choice of anticoagulant in the blood collection tubes, and the timing from collection to processing and cryopreservation [33–36]. Further, PCR detection assays are inhibited in DNA samples extracted from whole blood collected in heparin anticoagulant tubes [37]. Finally, the detection of many cytokines is affected by the choice of anticoagulant and processing procedures [38–40]. To enable data comparisons of correlative analysis across sites participating in multi-center studies, as well as interstudy analysis, it is imperative that biospecimen collection be consistent, and that guidance is provided to referring institutions where specimens are collected and shipped to participating study sites for testing using validated assays.

Figure 1: Pre-analytic, analytic, and post-analytic variables that impact correlative data harmonization efforts across sites participating in multi-center cell therapy trials.

Pre-analytic (patient and sample related), analytic (assay related), and post-analytic (data related) factors can impact consistency and reproducibility of assays commonly used to evaluate IEC trial-related samples. An in-depth discussion of the variables highlighted in bold are included within the scope of this work, including pre-analytic variables related to biospecimen collection and processing, in addition to providing recommendations for analytes to be measured using cytokine and flow-based assays.

We set out to overcome some of the challenges of multi-center research by developing a committee of experts from four institutions participating in the Consortium for Pediatric Cellular Immunotherapy (CPCI). The CPCI institutions have significant experience with cellular immunotherapy research, conducting several trials testing CAR T cells directed against single and multi-target antigens (e.g. NCT02028455, NCT03500991, NCT03638167, NCT03500991, NCT02311621, NCT03330691), non-gene modified antigen-specific T cells (NCT01956084), and natural killer cells in combination with other immunotherapies (NCT02573896). We compared standard operating procedures (SOPs) at CPCI sites and reviewed published literature related to key pre-analytic variables, including specimen collection, shipping, processing, and analysis of biologic samples in cytokine and flow-based assays. Our objective was to identify commonalities and opportunities for alignment, and provide best practice recommendations to promote consistency in sample handling across multi-center cell therapy trials. Here, we provide an in-depth analysis of some of the pre-analytic variables that affect sample integrity in the context of cytokine analysis and immunophenotyping by flow cytometry (Figure 1). We further provide analytic recommendations for performing basic cytokine and flow-based panels, which, when used uniformly across different cell therapy trials, could provide statistical power to interpret data across sites and between trials.

RESULTS

Specimen type and sample collection methods affect downstream assays

Prior to analyzing samples from different sites in a multi-center cell therapy trial or across different trials, it is important that sites align on which biospecimen types to collect and how to process samples of interest. All CPCI sites had well-defined collection and isolation procedures for peripheral blood mononuclear cells (PBMC), plasma or serum, while some sites also had practices for collecting bone marrow, cerebrospinal fluid (CSF) or tumor tissue, depending on the specific clinical trial requirements. Major gaps identified between CPCI site trials related to specimen collection and type were 1) timepoints of sample collection before and after IEC infusion, 2) the type of anticoagulant used in blood collection tubes, and 3) the use of plasma versus serum to assess cytokines (Table 1). Collection timepoints varied based on the nature of the underlying disease, patient population and logistics of collection, and were driven primarily by clinical trial aims. Sample types used for correlative assays and methods of collection and initial processing offered opportunities for alignment, as discussed below.

Table 1:

Gap analysis of general collection, shipment, PBMC isolation, and cryopreservation practices across CPCI sites.

	Site A	Site B	Site C	Site D	Gap Identified
	Blood Collection Tubes
Plasma isolation	EDTA	Sodium heparin	Sodium heparin	Sodium heparin	X
Serum isolation	Red top	-	-	-	X
	Shipment of Correlative Samples
Blood/bone marrow temperature	Ambient	Ambient	Ambient	Ambient
CSF temperature	4°C	4°C	4°C	4°C
Shipment container	Temperature-controlled	Temperature-controlled	Temperature-controlled	Temperature-controlled
Qualified courier	Local courier	Local courier	Local courier	Local courier
	Density Gradient Centrifugation
Whole blood lysis step to remove RBCs	Yes	Yes	Yes	Yes
Tube type	50mL SepMate	50mL SepMate	50mL Conical	50mL Leukosep	X
Speed (x g)	830	1200	400	800	X
Time (min)	20	10	30	15	X
	PBMC Centrifugation
Speed (x g)	250	400	400	500	X
Time (min)	10	5	10	10	X
	Cryopreservation
Cell freezing	Media + 10% DMSO	Media + 10% DMSO	Media + 10% DMSO	Media + 10% DMSO
Avg cell number cryopreserved/vial	5–10×106	5–10×106	5–10×106	5–10×106

Sodium heparin and EDTA are common anticoagulants for most clinical flow cytometry assays. We found that 3 out of 4 CPCI sites isolated plasma instead of serum for cytokine analysis in cell therapy trials. Sodium heparin was the preferred anticoagulant for collection tubes compared to EDTA, especially when prioritizing plasma isolation, because sodium heparin permits plasma and PBMC isolation from a single collection tube, while serum is isolated in a red top tube lacking anticoagulant. Constraints caused by phlebotomy limits (i.e. typically 3mL/kg with a maximum of 20–40 mL per collection) for safety in children favor sodium heparin tubes for plasma collections. Ultimately, the cytokine(s) being measured in downstream assays dictated the site’s choice to isolate plasma versus serum. Interestingly, cytokine levels can be higher in serum compared with plasma (Table 2), possibly related to higher non-specific background found in serum [38,39]. For example, measurement of TGFβ, a key cytokine whose overexpression in solid tumors is associated with immunosuppression and resistance to CAR T cell therapy [41–43], requires centrifuging samples at very high speeds to remove contaminating platelets that can aggregate, release TGFβ, and confound results [44,45]. We suggest isolating plasma instead of serum to detect low-level changes in abundance, especially for cytokines that may exhibit transient systemic elevations in response to immune effector cell infusion.

Table 2: Serum/plasma MFI of cytokines commonly assessed in cell therapy trials.

Ratio of serum/plasma mean fluorescence intensity (MFI) of cytokines measured in cell therapy trials. Representative NCT numbers were sourced from Berranondo et al. [39] and clinicaltrials.gov. Ratio of MFI values referenced from Rosenberg-Hasson et al. [38]

Cytokine	Representative cytokine clinical trials	Ratio of serum/plasma MFI
GM-CSF	NCT02933333 NCT04408092	2.615
	NCT01495637 NCT03769844 NCT02502786
IFNα	NCT04534634	1.206
IFNγ	NCT01965327 NCT02593773 NCT03548818 NCT02797080 NCT03378102 NCT03063632	1.584
IL-1α	N/A	1.491
IL-1β	N/A	2.225
IL-1RA	NCT04169022 NCT02780583	1.226
IL-2	NCT03138889 NCT02983045 NCT03282344 NCT03435640 NCT02983045 NCT02350673 NCT02627274 NCT03386721 NCT03063762 NCT03063762	2.259
IL-4	N/A	2.134
IL-6	N/A	4.058
IL-8	NCT03400332	1.280
IL-10	NCT02923921	1.433
IL-15	NCT02989844 NCT01875601 NCT02465957 NCT01385423 NCT01369888 NCT02689453 NCT02384954	2.868
TNFα	NCT03293784	1.986
TGFβ	NCT02423343 NCT02734160 NCT02581787 NCT03451773 NCT03451773	2.506
VEGF	NCT01984242	3.393

Shipment considerations for biologic samples collected on cell and gene therapy trials

With multi-site trials, local participating institutions are required to ship samples, either fresh or frozen, to a centralized laboratory for advanced analysis (e.g. NCT02028455, NCT03330691, NCT03684889). Consideration should be given to packaging material, shipping temperature, and integrity of the shipping container upon arrival at a central site for processing and analysis. If possible, the recipient laboratory should monitor factors affecting sample integrity, such as the temperature at which the sample was shipped/received, the accuracy of patient identifiers and the collection tube type and expiration, and document any deviations prior to processing. A recommended checklist for biospecimen receipt criteria is shown in Supplemental Table S1.

A comparison of shipping practices between CPCI sites revealed that all groups ship and receive blood and bone marrow at ambient temperature, preferably in insulated containers that limit temperature fluctuation (e.g. Clinpak©) (Table 1). Some specimen types, such as CSF, required special pre-processing at the clinical trial site prior to shipment to a central laboratory. More specifically, immediately after collection, fresh CSF was transported on ice and centrifuged as soon as possible at 4°C. The supernatant portion intended for cytokine analysis was isolated, aliquoted, and shipped frozen on dry ice to ensure high sample integrity for future batched analysis. CSF cell pellets generated by centrifugation were intended mainly for flow analysis, for example to assess immune cell infiltration or IEC persistence. Cell pellets were diluted in media and shipped at 4°C on ice or in a temperature-controlled shipping container (e.g. NanoCool©). Regardless of specimen type, biologic samples for cell therapy trials should be ideally shipped the same day as collection, to preserve sample integrity: All CPCI sites used a shipping courier with the ability to deliver the next day after pickup.

Incubation time between sample collection and processing must be validated and standardized for follow-up timepoints

As local referring institutions become more involved by collecting and shipping follow-up samples to central sites for analysis, the incubation time between sample collection and processing lengthens. While this workflow aims to increase patient access to clinical studies and eliminates their burden to travel to study sites for long-term follow-up, extended incubation time may compromise sample integrity [46]. Phenotypic and functional cell surface markers can be rapidly downregulated and/or cleaved over time, and fluctuations in cytokine concentrations in blood and CSF can occur after sample collection [36,47]. If shipping fresh samples, the impact of incubation time between collection and processing must be investigated and quantified for each type of specimen (e.g. blood, bone marrow, CSF, etc.) and assay, and the stability of key biomarkers after collection must be defined.

We analyzed validation data generated by a CPCI site to determine the maximum incubation time between blood collection and processing prior to signal loss. To determine post-collection stability standards for T cell and B cell surface markers commonly identified in cellular immunotherapy, peripheral blood (PB) from 4 healthy donors was collected in EDTA tubes and stored at room temperature until processing on Days 1, 2, 3, 4 and 7 post-collection (Day 0=Day of Collection). Quantitative stability analysis was performed by determining the last timepoint at which the observed five cell surface markers were within a +/−20% range of the Day 1 (D1) baseline value or a minimum of 80% of the observed markers were within the inter-assay precision range from Day 1 [48]. We concluded that beyond Day 3, specimen quality was degraded such that benchmark assay levels were unacceptable (Table 3). These data highlight the need for sites to establish acceptability criteria for fresh sample incubation time, consistently applied across multi-center clinical sites.

Table 3: Maximum incubation time (3 days) between blood collection in EDTA tubes and initial processing prior to loss of common T and B cell surface markers.

For baseline analysis (left side), a range of −20% to +20% of the Day 1 (D1) value for all samples and populations was used for quantitative analysis. Subsequent timepoint population values were examined and flagged if out of range. The out-of-range rate was calculated at each timepoint by the number of flagged populations out of the 20 total parameters examined per timepoint (5 populations for each of the 4 samples per timepoint). Total number of flags and out of D1 range rates are shown on the left. For inter-assay precision analysis (right side), the %CV for each population at each timepoint in comparison to the D1 value was determined. Values were flagged if out of range. Total number of flags and out of inter-assay range rates are shown on the right.

	Parameters Out of D1 Baseline Range				Parameters Out of D1 Inter-Assay Precision Range
Post-Collection Timepoint	D2	D3	D4	D7	D2	D3	D4	D7
CD3+	0	0	1	2	0	0	1	2
CD4+	0	0	0	1	0	0	2	2
CD8+	0	1	2	1	0	2	2	2
CD3−	1	0	3	4	0	0	1	3
CD19+CD22+	2	2	3	3	2	2	3	3
Total Number of Flags	3	3	9	14	2	4	9	12
Out-of-Range Rate (%)	15	15	45	55	10	20	45	60

Impact of fresh versus cryopreserved samples on cell therapy trial correlative studies

We first compared standard practices for isolating PBMC across all 4 CPCI sites, regardless of whether freshly isolated or frozen samples were used for analysis. PBMC isolation by ficoll gradient centrifugation of diluted whole blood at room temperature was performed at all 4 sites. While centrifugation lengths, speeds, and tube types used for ficoll purification differed between labs (Table 1), all CPCI sites washed isolated PBMC with a buffered salt solution, and a whole blood lysis step was performed to remove red blood cells (RBC) prior to flow analysis. Isolated PBMC were then used either fresh or frozen in downstream correlative assays. Given the limited amount of whole blood that can be drawn from pediatric patients, an important consideration is to anticipate low abundance of cells of interest (e.g. engineered cells, target cells, etc.). In cases where collection volume is low, we recommend performing ficoll in a smaller tube format (e.g. 15mL tube vs. the typical 50mL tube) and eliminating the RBC lysis step, because the ficoll gradient centrifugation may be sufficient in removing contaminating RBCs from low-volume samples. Additionally, for downstream functional assays using PBMC isolated from low-volume samples, we recommend expanding PBMC ex vivo to achieve a higher frequency of effector cells prior to analysis. Ex vivo PBMC expansion prior to analysis may affect phenotype and functional profile of the expanded cells, therefore, if performed, this expansion should be implemented consistently across samples in the study.

Frozen sample collections facilitate batched sample runs at a central laboratory, which can be performed after all clinical endpoints are satisfied. Batched analysis aims to minimize experimental variability and allows for direct comparison between post-drug product infusion timepoints. While there is no definitive evidence to support using fresh versus frozen samples for analysis in CAR T cell or HCT studies, previously published work assessing T and B cell subsets in human PBMC samples has reported that there are no observed differences between fresh versus frozen samples for some surface and intracellular markers [49,50]. Conversely, other groups have shown that cryopreservation can introduce variables that impact sample integrity, including viable cell recovery and stability of cell populations between freeze/thaw [34,51–53]. We measured recovery of CD34 and CD3 expression in participant-matched fresh versus frozen human PBMC and found that, while there was no difference in CD34 recovery (P=0.0752; paired t test), there was a significant reduction in CD3 detection in frozen samples compared to fresh (Fig. 2A; P=0.0004; paired t test; Supplemental Table S2).

Figure 2: Recovery of mononuclear cells, CD34⁺ and CD3⁺ cells in freshly isolated versus cryopreserved samples, and CD3 and CAR marker detection in fresh versus cryopreserved matched pairs.

(A) Comparison of percent recovery of cryopreserved and fresh participant-matched PBMC samples; fresh samples normalized to 100%. Mononuclear cell (MNC) recovery was calculated by comparing MNC cell counts prior to freezing with the live cell counts of the thawed cells for each sample (n=60). Percentages of each antigen expression are derived from Lymphocytes of CD45⁺ events. Similarly, CD3⁺ (n=20) and CD34⁺ (n=46) recovery was calculated by comparing flow cytometry expression between the fresh and thawed samples. Cell viability data (n=62) was calculated by examining 7-AAD expression of the thawed cells. Significance was determined using paired t test. (B) Target population recovery/stability was evaluated for CAR T cells by examining the frequency of lymphocytes expressing markers of interest in participant-matched fresh and cryopreserved cells. Fresh samples (n=26) were obtained from peripheral blood specimens treated with RBC Lysis prior to staining, while frozen samples (n=26) underwent mononuclear cell isolation by ficoll density gradient separation prior to cryopreservation. Gating strategy included viable singlet non-myeloid cell isolation prior to selection of the lymphocyte population of interest. Significance was determined using the Wilcoxon matched-pairs signed rank test.

We further confirmed the reduction in CD3 enumeration in frozen samples in a separate analysis comparing cell marker detection in 26 participant-matched fresh versus frozen PBMC samples from pediatric subjects who received CD19-directed CAR T cells (NCT02028455; Fig. 2B). The percentages of CD3⁺ cells and CD8⁺ cells were significantly decreased in frozen samples compared to fresh (P<0.0001; Wilcoxon matched-pairs test), while the percentage of CD4⁺ cells was not significantly different (P=0.1574). Similarly, we examined the effect of cryopreservation on truncated epidermal growth factor receptor (EGFRt), which is used as a cell surface marker tag to detect transduced CAR T cells [24,54–56]. Interestingly, detection of CAR T cells was also significantly decreased in frozen PBMC samples compared to fresh (CD3/EGFRt⁺, P<0.0001; Wilcoxon matched-pairs test). Together, these data show that cryopreservation can affect detection of cell populations of interest in cell and gene therapy trials. They also highlight the benefit of analyzing freshly isolated samples to accurately monitor the status of cell populations in real-time, in subjects receiving CAR T cell therapy.

Cell and gene therapy products are typically cryopreserved at the end of manufacturing and then thawed at bedside for infusion [57,58]. We next examined the effect of cryopreservation on CAR T cell marker detection, using a dual-transduced, bispecific CAR product (Fig. 3). The product was stained pre-freeze and post-thaw with cetuximab and trastuzumab antibodies to detect the CAR transduction markers EGFRt and Her2tG, respectively. Immediately upon thaw, cells were co-cultured with target cells expressing CAR-specific antigens. Compared to pre-freeze, CAR T cells were not readily detectable post-thaw (day 0) via flow cytometry, and the product appeared predominantly CAR-negative. Ability to detect CAR T cells was rescued by day 2 of culture, and percentages continued to increase between days 2 and 5. It is unclear if the recovery of CAR T cells was due to improved stability or detection of the CAR markers or whether co-culture with antigen-expressing target cells promoted expansion of the CAR T cells over time. Both factors may have affected CAR positivity of the product post-thaw. By day 7 and onward, EGFRt detection declined while Her2tG detection remained relatively consistent over time.

Figure 3: Time course of CAR T cell transduction marker detection upon co-culture with target cells.

Healthy donor T cells underwent transduction with two CAR constructs expressing either the EGFRt or Her2tG reporter molecule. Following the post-transduction expansion period, an aliquot was removed for flow cytometry staining, and the remaining culture was cryopreserved. Cells were later thawed for stimulation and culture expansion via CAR T Rapid Expansion Protocol (REP), with periodic sampling for flow cytometric analysis of the CAR reporter molecules. While detection of EGFRt and Her2tG CAR markers dramatically decreased immediately post-thaw, Her2tG expression and/or detection recovered to pre-freeze levels during the course of the culture.

Together these data show that results from thawed PBMC samples and cell products may not accurately reflect pre-freeze characteristics. They further emphasize the need to consider the stability of phenotypic and transduction markers used to identify cell populations in fresh versus frozen cell therapy samples and products, and their persistence in peripheral blood over time. We recommend fresh analysis when possible, to monitor toxicities in real-time and to eliminate pre-analytic variables introduced by cryopreservation.

Resting effector cells in cytokine-containing media may improve the recovery of cell markers affected by cryopreservation

While analysis of fresh samples is preferred, this workflow may not optimize adherence to follow-up sample collection schedules at referring institutions, especially when local providers may be far from participating study sites. If samples cannot be processed, shipped, and analyzed in real-time, cryopreservation likely will be necessary for expanding patient access and utilization of cellular therapies. A comparison of cryopreservation methods across CPCI sites revealed that all sites were consistent in their practices, using freezing media containing 10% DMSO and initiating freezing in a commercial freezing container (e.g. Nalgene Mr. Frosty), followed by long-term storage in liquid nitrogen (Table 1). Samples were cryopreserved in aliquots to prevent repeat freeze-thaw cycles, and the size of aliquots for plasma and PBMC or bone marrow mononuclear cells was validated by each site per their SOPs.

Related to the effect of cryopreservation highlighted above, there is evidence to suggest that resting of cryopreserved samples post thaw can promote recovery of T cell function and restore several phenotypic and functional markers [59,60]. Using a frozen cell therapy product as an example, we sought to determine the impact of thawing, with or without resting, on the ability to detect CAR T cell populations. A dual-transduced, bispecific CAR product was thawed and then stained with cetuximab and trastuzumab antibodies to detect EGFRt-tagged and Her2tG-tagged CAR T cells, respectively. The product was stained immediately post-thaw (Day 0), after 24 hours of rest, and after 48 hours of rest (Fig. 4). Cells were rested with or without addition of IL-2 to the media. We found that the ability to detect surface expression of the CAR T cell markers, especially EGFRt (lower panels), was suboptimal immediately post-thaw compared to detection in fresh samples (similar to Fig. 3). In the absence of IL-2, 24- or 48-hour resting was able to improve detection of the Her2tG tag, but not EGFRt. Interestingly, supplementing media with IL-2 during the 24-hour rest markedly improved detection of EGFRt from 1.96% post-thaw to 24.2% compared to resting in media without IL-2, which only had a slight increase in detection to 4.5%. No further increase in EGFRt was detected after 48 hours of rest compared to 24 hours in IL-2 supplemented media. These data show that stability and/or detection of CAR transduction markers is impacted by freeze/thaw. If cryopreservation is unavoidable, however, these data support incorporation of a resting step with IL-2 upon thaw to promote recovery of CAR T cell markers of interest prior to downstream assays.

Figure 4: CAR T cell transduction marker detection with or without rest in media with or without IL-2.

Enhanced CAR transduction-marker recovery in cryopreserved cells was attempted by the addition of a rest period post-thaw, with or without addition of IL-2. Transduced CAR T cells were examined pre- and post-cryopreservation, as described in Figure 3. Following thaw, cells were resuspended in either R10 media or R10 media supplemented with 50U/mL of rhIL-2 cytokine. Cells were incubated at 37°C and evaluated by flow cytometry staining on day 1 (D1) and day 2 (D2) post-thaw. The addition of IL-2 showed enhanced CAR T cell transduction-marker staining compared to day 0 (D0; immediately post-thaw) and the unsupplemented culture; extending the rest period beyond one day did not result in any further increase in reporter molecule detection in either culture.

Key cytokines to measure in correlative samples for cell therapy trials

After careful consideration of pre-analytic variables that adversely affect correlative data results, we also compared markers used at our respective CPCI sites for cytokine profiling in order to develop a consensus panel for pediatric cell and gene therapy samples. Table 4 shows a list of recommended cytokines to analyze in participant-derived plasma. All of the cytokines included have established roles in immune effector related toxicities associated with IEC therapy, including cytokine release syndrome (CRS) and immune effector cell-associated neurotoxicity syndrome (ICANS). For example, IFNγ, IL-6, IL-10, GM-CSF and CCL2 (also known as monocyte chemoattractant protein-1) have all been implicated in CRS onset and severity, as well as neurotoxicity [21,56,61–64]. Additionally, hemophagocytic lymphohistiocytosis (HLH), a syndrome of cytokine-driven immune activation that results in multisystem organ dysfunction and failure, has largely been associated with high levels of IL-1β, IL-6, IFNγ and TNFα [20,65,66].

Table 4: Recommended cytokines to measure in cell therapy trials.

Cytokines are listed with associated roles in immune effector-related toxicities.

Cytokine	Toxicities	Roles
CSF-1		Regulates monocyte/macrophage differentiation [99]
GM-CSF	CRS, HLH and neurotoxicity	Drives CRS and neuroinflammation [63]
IFNα		MSC production [100] and cytokine delivery, induces expression of tumor suppressor proteins [101]
IFNγ *	CRS and HLH	Contributes to immunotherapy, tumor suppression and the efficacy of immune checkpoint blockade [102]
IL-1 *	CRS	Innate immunity [103]
IL-2 *	CRS, HLH and neurotoxicity	Promotes expansion of T and NK cells
IL-4		Promotes B cell proliferation; mediates inflammation [104]
IL-5		Promotes B cell proliferation [105]
IL-6 *	CRS and HLH	Associated with CRS onset and severity
IL-8	CRS	May be predictive of resistance to ICIs [106]
IL-10	CRS
IL-12		Activation and regulation of macrophages, T and NK cells [107]
IL-13		Promotes B cell proliferation; mediates inflammation [104]
IL-15 *	CRS	Induces proliferation of CD8 memory and NK cells, cytotoxicity, and release of other cytokines (e.g. IFNγ) [108]
IL-17		Pro-inflammatory cytokine [109]
IL-21		Activates STAT3 signaling in T cell and B cell differentiation [110]
TGFβ		Promotes cancer progression [111]
TNFα *	CRS and HLH	Mediates inflammation, anti-tumor responses and infection [112]

^*Indicates cytokine should be included in a basic minimum panel.

The cytokines listed in Table 4 have roles in potentiating the immune response (e.g. IFNα, IL-2, IL-10, IL-12, IL-15, IL-21, GM-CSF) [67,68] or inhibiting immunosuppressive activity (e.g. IL-1, TNFα, TGFβ, CSF-1) [69–72], and are therapeutically targeted in clinical trials [39]. In particular, IL-1 blockers, such as anakinra and canakinumab, are FDA approved and widely used for the treatment of autoimmune and autoinflammatory diseases (e.g. NCT02179853, NCT04656184, NCT02780583) and are currently being tested in preclinical and human clinical trials for cellular immunotherapy (e.g. NCT04148430) [73–75]. Importantly, we include cytokines that may be produced by genetically modified therapeutic cells themselves or by other cell types within the microenvironment (e.g. myeloid cells, tumor cells, etc.). Analyzing cytokines produced by multiple cell types maximizes data generated from pediatric samples, while improving understanding of the immune landscape.

Methods to analyze cytokine biomarkers in biologic samples differed across CPCI sites. Some sites measured cytokines in plasma using standard Luminex and/or ELISA-based assays, while others detected intracellular cytokine expression and mean fluorescence intensity (MFI) via flow cytometry. Given technical limitations of each assay platform, it may not be possible to assay all of the cytokines listed in Table 4 simultaneously. At a minimum, we suggest analyzing IFNγ, IL-1, IL-2, IL-6, IL-15, and TNFα in a basic cytokine panel, due to the potential clinical utility of these cytokines as well as their wide use in current pediatric cellular immunotherapy clinical trials.

Evaluation of function and exhaustion of immune effector cells (IECs) by flow cytometry

We next compared basic flow staining practices, gating strategies and panels across participating CPCI sites. We found that flow analysis of biologic samples for cell therapy trials mainly aimed to monitor immune effector cell engraftment, persistence, effector function, and the immune landscape of pediatric malignancies. All CPCI sites used basic immunophenotyping and memory/differentiation markers to distinguish T and B cell subsets, NK cells and monocytes, as previously described in the context of cell-based therapies in HCT [3]. However, pediatric cell and gene therapy trials require a more in-depth analysis of genetically modified immune cells to discover new biomarkers of functional or dysfunctional responses. Activation, exhaustion and functional status of adoptively transferred cells is important to correlate with toxicity data and critical endpoints such as loss of persistence, relapse or clinical outcome.

We recommend the markers shown in Table 5 to assess activation, exhaustion and functional status of immune effector cells. All CPCI sites analyze these markers as a part of the exploratory objectives of clinical trials, with the goal of linking them to other correlative findings including cytokine/chemokine data. All of these markers are commercially available with staining kits, making them even more accessible for sites to explore as a part of multi-center research.

Table 5: Recommended markers to evaluate IEC activation or exhaustion status.

A list of suggested markers to evaluate immune effector cell (IEC) activation or exhaustion status as a part of cell therapy trials.

Marker	Role
CD3	Lineage
CD4	Lineage
CD8	Lineage
CD25	Activation
TIM-3 (CD366)	Exhaustion
CTLA-4 (CD152)	Exhaustion
LAG-3 (CD223)	Exhaustion
PD-1 (CD279)	Exhaustion
Perforin or Granzyme B*	Function
IFNγ*	Function
TNFα*	Function
IL-2*	Function

^*Indicates intracellular staining required.

DISCUSSION

Correlative studies contribute significantly to our understanding of the in vivo activity, safety and performance of adoptive cellular therapies. Most cell and gene therapy pediatric clinical trials include collection of biologic samples for short- and long-term monitoring during and after treatment. These correlates are necessary to detect and study inherent risks related to factors such as site of vector integration, persistence of the gene product, replication competence of the vector, and associated immunogenicity-related reactions. CAR T cell clinical trials have also shown that severe CRS and ICANS/neurotoxicity are associated with early increases (< 3 days post T cell infusion) in predictive biomarkers such as IFNγ, IL-2, IL-6, IL-10, GM-CSF, MCP-1, and VEGF [21,76–78]. Furthermore, the initial expansion and long-term persistence of CAR T cells after infusion is one of the key correlates of long-lasting clinical remission [79–81]. We report consensus recommendations from the CPCI sites leading cellular immunotherapy trials for pediatric disease, to align practices and improve reliability and reproducibility of correlative assays for key biomarkers in cellular immunotherapy.

We identified the choice of blood collection tubes, the incubation time between sample collection and processing, and the choice of plasma or serum for cytokine measurement as important variables of correlative studies relevant to cell therapy trials. The anticoagulant used in collection tubes can impact the recovery of IECs and their phenotypic and functional attributes, and can also impact the feasibility of obtaining plasma prior to isolation of PBMC or bone marrow mononuclear cells (BMMC) [47,82]. Plasma is preferred over serum for cytokine analysis due to the ease of collection in a single sodium heparin tube, especially for low level blood draws from children, and the ability to detect low-level changes in cytokine abundance. Our review also assessed the impact of cryopreservation on markers used to identify and characterize IECs and cellular products in correlative assays. Assays that require detection of specific cell markers that are altered by cryopreservation should be performed on freshly isolated samples if at all possible, although a combination of fresh and frozen samples may be required to promote broader study participation. We recommend validating and using the same sample preparation within individual trials for consistency and reproducibility.

While our results highlight the effects of cryopreservation and thawing on CAR T cell products and markers expressed on PBMC, our findings can be extended to other sample types. Flow-based detection of IECs or rare immune cell subsets in cryopreserved-then-thawed samples may be even more challenging compared to cellular products that are enriched for immune effector cells. For example, phlebotomy volume limits to ensure safety in pediatric patients can also impact the ability to detect CAR T cells post-infusion in bone marrow or CSF samples in real-time. This limitation will be exacerbated by relying on cryopreservation of leftover sample in these patients. While CAR transduction marker tags (e.g. EGFRt or Her2tG) can be useful for identifying engineered cells in fresh products or samples, we recommend avoiding use of these markers for flow-based detection of cryopreserved and banked samples. Instead, we suggest using direct scFv fluorochrome conjugates, protein L, and/or other strategies for flow-based CAR detection [25]. Alternatively, where cryopreservation or biobanking of leftover samples for future research is preferred, molecular-based approaches such as qPCR or digital droplet PCR (ddPCR) to detect CAR T cell DNA may be more sensitive than flow-based strategies to detect cells. PCR or other molecular-based approaches may also be advantageous in cases when there are delays in shipping and/or receiving follow-up samples from referring institutions to participating study sites.

In addition to peripheral blood, bone marrow and CSF samples, in recent years, more studies are using information from biofluid samples, such as dialysate and urine, as non-invasive tests to evaluate treatment response and resistance to cancer immunotherapy [83]. As utilization of biofluids shows promise in toxicity assessments of IEC therapy, future work is needed to standardize collection and processing methods for pediatric patients.

Only a single CPCI site had guidance related to solid tissue collection and processing. Tumor biopsies or resection tissues from children are particularly rare, and future work will be necessary to develop best practices for collecting, processing and biobanking solid tissue as a part of multi-center trials. Clinical trials evaluating cell and gene therapies targeting the most common types of pediatric solid tumors, such as brain tumors, neuroblastoma, and osteosarcoma, are becoming increasingly prevalent (e.g. NCT03294954, NCT02311621, NCT03618381, NCT04483778, NCT03500991, NCT03638167, NCT04185038, NCT02789228, NCT02573896, NCT01326104), but there are numerous challenges to using cellular therapies in the context of solid tumors [84]. Correlative analysis of fresh or frozen tissue will be essential to gain insight into mechanisms of action of cell therapies within the context of solid tumor microenvironments. Guidance for developing correlative studies for pediatric solid tumor trials is severely lacking and will be an important focus of future CPCI work.

Comprehensive cytokine profiling post-infusion of IECs informs mechanisms of action of cell therapies and immune related effector toxicities, highlighting the importance of establishing a base panel of cytokine analyses to be incorporated across trials. Many of the cytokines in Table 4 are often measured in peripheral blood and bone marrow samples. However, cytokines of interest to evaluate in CSF for cell therapy trials are not well established. Potential biomarkers of neuroinflammation and neurotoxicity associated with CD19-directed CAR T cell infusion in patients with B-ALL have been explored [19,54,76], however tumor heterogeneity and the complexity of the central nervous system (CNS) environment contribute to the variability and inconsistencies in many previous studies. For example, a study evaluating the clinical utility of CCL2/MCP-1, CXCL8/IL-8, CXCL10, CXCL13, and IL-6 found that all of these cytokines were detectable in the CSF of symptomatic patients – but compared to the combination of biomarkers commonly assessed in CSF, the majority of these cytokines had decreased sensitivity and specificity for confirming neuroinflammation (e.g. white blood cell [WBC] counts, oligoclonal bands, total protein levels, CSF/serum albumin ratios) [85]. Future work is needed to align sites on which cytokines to explore in CSF as part of correlative studies. The chosen cytokines will likely vary depending on trial objectives and tumor type, for example between studies assessing CSF biomarkers of neurotoxicity in leukemia patients versus biomarkers of inflammatory response in pediatric patients with CNS tumors.

There is increasing evidence that exhaustion and senescence of engineered cells affect their in vivo proliferative capacity, long-term persistence and anti-tumor function, which in turn impacts the efficacy of CAR responses and ability to achieve durable remissions [86]. We recommend the inclusion of activation, exhaustion and functional markers to basic immunophenotyping flow panels that have been previously described [3]. Finney et al. has recently shown that including additional exhaustion and functional markers in correlative studies has the potential benefit of identifying cellular products and/or patients that may be capable of achieving sustained remissions or be at risk of therapeutic failure [55]. It is also important to note that expression of a single inhibitory receptor alone, such as PD-1, TIM-3 (CD366) or LAG-3 does not always indicate exhaustion. For example, TIM-3 is associated with both co-stimulatory and inhibitory functions, depending on the cell type, so it is important to measure TIM-3 expression in combination with other markers of T cell dysfunction [87,88]. Terminally exhausted T cells exhibit a loss of effector cytokine production and co-express multiple inhibitory receptors, with the number of co-expressed receptors directly correlating with the severity of exhaustion [89,90]. We therefore recommend measuring co-expression of at least three inhibitory receptors, in addition to other functional markers. Significantly, the insight gained from the evaluation of exhaustion markers immediately following IEC infusion could provide guidance for the utilization of checkpoint blockade to enhance the function and persistence of the infused product [91,92]. This insight can also guide development of next-generation cell therapies using gene editing approaches to knock out checkpoints that can improvise the engineered product [93–97], or the manufacture of ‘armored’ CAR T cells, especially in the setting of solid tumors [98].

Future discovery of new biomarkers, and the study of next generation cell and gene therapy products in multi-center trials, will require alignment of sample collection timepoints so data can be compared across sites. In addition to the type of sample and methodologies to evaluate biomarkers, the timing of correlative sample collections can vary greatly depending on clinical trial design, resources available at each site, and whether the participants are undergoing inpatient or outpatient care. For instance, the timing of evaluating pre-infusion bone marrow disease burden for pediatric B-ALL differs not only between trials but also in clinical practice. Some sites collect bone marrow prior to lymphodepletion (LD), before IEC infusion, while other sites collect after LD, and thus a subject may already be considered in a minimum residual disease (MRD)-negative status prior to infusion. This variation could lead to difficulty in interpreting the optimal disease burden required for adequate antigen stimulation or identifying the pre-existing or emerging resistant clones. Similarly, for optimal detection of early in vivo changes, such as engraftment of IECs, cytokine levels or toxicity monitoring, most clinical trials collect 2 or 3 timepoints within the first week of treatment, followed with weekly draws for the first 30–35 days, and monthly draws for the first 6 months [21,76]. Although these considerations are not included in our current review, future research may need to assess the impact of timing of other correlative sample collections on study conclusions.

Increasing patient access to cellular immunotherapy warrants a focus on basic assays that can be performed at multiple participating sites. However, it is important to note that the cell and gene therapy field offers several advanced molecular assay platforms to evaluate correlative samples. For example, next generation sequencing (NGS), single-cell profiling of RNA/DNA, and T cell receptor immuno-sequencing can provide a comprehensive look at the TCR repertoire of IECs infused into patients. Given the complexity of these assays and their cost, it will be challenging to implement them widely across centers, necessitating analysis at a centralized site. Recommendations for sample collection, processing, and shipping for these assays will be an important future direction.

Key considerations need to be taken into account when designing correlative studies across multi-center cell therapy trials (Table 6). Increased standardization of current-state practices will improve sensitivity and specificity of data generated in early-phase IEC trials and of comparisons of data between different trials. Standardization will have the greatest impact in the study of pediatric diseases where correlative studies are limited by small subject numbers and sample collection volumes of biospecimens. When standardizing practices across sites, consultation, and participation with external quality assessment programs is highly encouraged. There are numerous organizations that work toward establishing best practices globally. The scope of assessment can range from pathology testing [for example, the College of American Pathologists (CAP) and UK National External Quality Assessment Scheme (NEQAS)] to technologies that are relevant to IEC trials, such as flow cytometry [for example, the International Society for Advancement of Cytometry (ISAC)]. Accreditation from an external party can also ease harmonization efforts as participating sites will already have met similar standards.

Table 6:

Alignment considerations for sites participating in multi-center cell therapy trials

Parameter	Key Considerations
Experimental Design	• What type of correlative data is necessary for the clinical trial?
	• What sample types and collection timepoints are required?
	• What analytes are to be measured in biologic samples?
	• What instruments and detection parameters are required?
	• Will samples be cryopreserved or banked for future research?
Site Evaluation	• What instruments are available at each participating site?
	• What assays can be performed at each participating site?
Assignment vs. Centralization	• Will any assays be performed locally or will samples be shipped to a central lab?
	• Has assay validation been performed at sites using fresh or frozen samples?
	• Consider determining acceptable ranges for measurements of key analytes across sites?
	• Partner with an external quality assurance organization or program to broaden validation and harmonization efforts?
	• Are centralized SOPs created to facilitate staff training on correlative sample collection, processing, and analysis?

Once SOPs have been established and standardized across sites participating in multi-center cell therapy trials, initial validation testing should be conducted to compare instrument performance and processing standards across sites. During initial testing, baseline and threshold parameters should be established for each procedure and acceptable ranges should be determined. To ensure continuous harmonization, periodic reviews should be performed to ensure uniformity and reproducibility. In one example study, standardization procedures between instruments were repeated every 3–6 months [27]. The timepoints between harmonization efforts may depend on the length of a particular study and the types of correlative studies. After initial validation testing, sites should regularly monitor instrument performance and quality control to establish guidelines for periodic review. Proficiency samples with known performance ranges should be tested regularly for each analyte of interest at participating sites to confirm continued performance within range across sites. Maintaining SOPs and providing training to study site staff related to correlative studies and validation testing are key to ensuring process and data harmonization.

Open discussion of correlative studies strategies during clinical trial development, consideration of quality control criteria and boundaries for acceptable measurements of key analytes, and consistency in pre-analytic sample handling are key steps to ensure reliable and high-quality correlative data across sites.

MATERIALS AND METHODS

Refer to Supplemental Table S3 for a comprehensive listing of all reagents used, vendor information and catalog numbers.

Consortium for Pediatric Cellular Immunotherapy (CPCI) correlative working group and gap analysis

The CPCI working group consisted of representatives from each participating site and included experts in clinical trial design, laboratory scientists, and physician-scientists. The group met regularly to review and align practices across sites. This effort involved independent review of SOPs at each site, identification of similarities and gaps between SOPs, systematic review of the literature, and analysis of relevant cell therapy data.

Cell thawing

Cryopreserved cell vials were thawed quickly in a 37°C water bath before slowly diluting the cells into R10 media (RPMI (Gibco) with 10% FBS (Atlas) and 1% L-glutamine (Gibco)). Cells were washed with R10 media before allocation to downstream assays. Cells were either stained immediately post-thaw or incubated in the relevant conditions for other analyses.

Whole specimen PB staining

A maximum of 2 mL of peripheral blood (PB) specimen collected in EDTA was lysed with RBC Lysis Buffer (Invitrogen) before counting the recovered number of white blood cells. For cell yields below the maximum allowable number of cells for staining assays, the entire recovered cell product was stained.

Recovery of mononuclear cells (MNC), CD34 and CD3 in cryopreserved versus freshly isolated samples

Percent recovery of cryopreserved and fresh PBMC samples were compared; fresh samples were normalized to 100% (see Figure 2A and Supplemental Table S2). MNC recovery was calculated by comparing MNC cell counts prior to freezing with the live cell counts of the thawed cells for each sample. Similarly, CD3 and CD34 recovery was calculated by comparing flow cytometry expression between the fresh samples prior to cryopreservation and the subsequently thawed samples. Cell viability data was calculated by examining 7-AAD expression of the thawed cells. Non-viable cells that have lost membrane integrity are identified by uptake of 7-AAD (i.e. 7-AAD positive cells by flow cytometry) and were excluded from the analysis.

Flow staining

Samples were treated with viability dye before incubating with Fc Receptor Block (Miltenyi) to inhibit non-specific antibody binding. A cocktail of CD3 V450, CD4 BV605, CD8 PerCP-Cy5.5, CD36 FITC (BD), and EGFRt APC and Her2tG-biotin (BD, custom conjugates) diluted in Brilliant Stain Buffer (BD) was used for surface staining. Antibody staining was performed at room temperature in the dark for 20 minutes. Cells were washed twice with PBS (Gibco) between all staining and fixation steps. Secondary staining with Streptavidin BUV395 (BD) was performed in Brilliant Stain Buffer in the dark for 20 minutes at room temperature. Cells were fixed with 2% PFA before acquisition on an LSRFortessa (BD) flow cytometer. Data was analyzed using FlowJo software (FlowJo).

CAR T Rapid Expansion Protocol (REP)

Cryopreserved T cells previously transduced to express CAR constructs were thawed and mixed at a standard ratio with irradiated target cells bearing the target antigen of interest and co-stimulatory molecules. Cells were cultured in R10 media supplemented with recombinant human IL-15 and IL-2 cytokines at 37°C for up to 21 days post-thaw. Cultures were split on a biweekly basis to allow for cell expansion and to replenish the cytokine supplements. Samples were removed from culture on a regular basis for immediate use in cell staining assays.

IL-2 resting post-thaw

Cryopreserved cells were thawed into R10 media. Following removal of an aliquot for staining, the remaining cells were divided and pelleted by centrifugation. Cell pellets were either resuspended in R10 media or R10 media supplemented with recombinant human IL-2 cytokine at 50U/mL. Cells were incubated at 37°C in cell culture flasks for two days. Cells were removed from culture on Day 1 and at the termination of culture on Day 2 for immediate use in cell staining assays.

Highlights.

• Multicenter trials increase access to cellular therapies in pediatrics.

• Correlative studies yield immune effector cell safety and efficacy data.

• Harmonization is key for sample collection, processing, cryopreservation and assays.

• We provide data for alignment of best practices across sites and between trials.

• We provide recommendations for basic immunophenotyping and cytokine panels.