Efficient and traceless aptamer-based selection of naïve and early memory CD8 T cells for CAR T cell therapy

Abstract

Chimeric antigen receptor (CAR) T cell therapies have shown clinical success in cancer treatment. However, the compositions of the final products can differ substantially between patients, leading to variable treatment responses. Recent studies suggest that CAR T cells manufactured from defined T cell subsets show greater potency and persistence and improved predictability of therapeutic efficacy. Current clinical-scale selection of T cell subsets relies on antibody-based magnetic activated cell sorting, which is costly and results in suboptimal product purity and yield, presenting a significant challenge for clinical translation. Here, we report a high-affinity CD62L aptamer and a traceless, sequential selection system for high-yield and high-purity isolation of CD62L⁺CD8⁺ T cells without residual selection labels. We demonstrate that multiple aptamer-reversal agent pairs can be integrated into a magnetic platform for multi-parameter and high-throughput cell sorting. CAR T cells manufactured from aptamer-selected CD62L⁺CD8⁺ T cells, encompassing naïve and early memory CD8⁺ T cells, exhibit distinct phenotypic and functional advantages compared to those manufactured from bulk CD8⁺ T cells. This aptamer-based approach has the potential to improve the clinical efficacy of current adoptive T cell therapies by enabling precise and scalable selection of T cell subsets, with broad applications beyond T cell subset selection.

Graphical Abstract

In this work, we developed an aptamer-based system for the high-purity and high-yield selection of CD62L⁺CD8⁺ T cells. We demonstrate that multiple aptamer-reversal agent pairs can be integrated into a magnetic platform for multi-parameter and high-throughput cell sorting at production scale. We further showed that CAR T cells manufactured from aptamer-isolated CD62L⁺CD8⁺ T cells have phenotypic and functional advantages over bulk CD8⁺ T cells.

Introduction

While CAR T cell therapies have shown tremendous clinical success in treating hematological cancers, treatment efficacies vary widely among patients, and high relapse rates remain a concern. Accumulating pre-clinical and clinical evidence suggests that the starting T cell population greatly affects the antitumor efficacy and long-term persistence of CAR T cells, with tumor-redirecting T cells derived from less differentiated naïve (T_N), stem cell-like memory (T_SCM) and central memory (T_CM) T cell populations conferring superior antitumor immunity, greater in vivo expansion, and improved persistence compared to unselected or more differentiated T cell subsets.^[1–9] Retrospective analyses of clinical studies also indicate that the proportion of CD8⁺ T_SCM and T_CM cells in patients undergoing CAR T cell therapies strongly correlates with in vivo T cell expansion and clinical remission, surpassing disease- or patient-related factors in predicting remission outcomes.^[10–11] Initial clinical trials that investigated the efficacy and safety of CAR T cells derived from highly defined cell compositions support these findings but revealed the manufacturing challenges that hinders clinical translation of this promising approach.^[3]

One of the major markers for T_N, T_SCM, and T_CM populations is L-selectin, or CD62L, an adhesion molecule expressed exclusively on hematopoietic cells. CD62L facilitates the extravasation of naïve and early memory T cells from the vasculature and contributes to their lymph node-homing ability.^[12–13] CD62L is shed upon activation and is not expressed on effector memory (T_EM) and terminal effector (T_TE) T cells. Clinical studies evaluating CAR T cells derived from naïve and early memory T cell subsets have used CD62L as a selection marker during CAR T cell manufacturing.^{[5, 14–16]} However, precise subset selection increases manufacturing complexity, time, and cost and reduces product yield.

Clinical manufacturing and evaluation of CAR T cells from defined T cell subsets has been challenging in large part due to the limited technologies available for T cell subset selection at the clinical scale.^{[3, 17]} To avoid excessive expansion of T cells during the manufacturing process, large cell numbers are usually isolated by higher-throughput magnetic activated cell sorting (MACS) instead of by fluorescence-activated cell sorting (FACS). However, multi-parameter cell sorting with conventional, antibody-based MACS is limited due to the reliance on a single sorting parameter (the presence or absence of magnetic labels) and the retention of selection labels (generally antibodies immobilized on magnetic beads) on the cell surface after selection. Consequently, further selection on positively selected cells is strictly precluded. As a result, selection of specific cell subsets requires a combination of positive and negative selection strategies that result in suboptimal purity and yield dependent on sample variability.^{[14–16, 18–19]} The necessity for multiple antibody-based isolation procedures and a complex panel of biologically derived reagents to obtain naïve and memory T cells also adds significant cost to an already costly manufacturing process.

Aptamers are single-stranded oligonucleotides that fold into sequence-defined secondary structures capable of binding to their targets with affinities and specificities comparable to antibodies. Typically discovered from a library selection process known as systematic evolution of ligands by exponential enrichment (SELEX), aptamers have been discovered and applied for diverse applications ranging from therapeutics to diagnostics to purification.^[20–21] DNA aptamers and complementary strand displacement with a reversal agent are uniquely positioned to improve T cell subset selection for adoptive T cell manufacturing by enabling high throughput and cost-effective multi-parameter cell sorting via MACS. First, aptamers can be discovered to bind a wide range of targets and complementary strand displacement can be readily achieved by designing the corresponding reversal agents for each aptamer. After each isolation, aptamers can be released from the cells, rendering the cells label-free and available for re-labeling and subsequent selections.^{[20, 22]} Having a label-free product also prevents masking the receptor in downstream analysis and minimizes any undesired impact on cell signaling or proliferation induced by ligand binding.^[23–25] Second, aptamers can be chemically synthesized at large scale with minimal batch-to-batch variation at much lower cost than antibodies, reducing variability and the high cost of materials associated with CAR T cell manufacturing.^[26–30] Finally, the ability to release aptamers from cells after isolation enables their translation to a variety of clinical-scale production platforms. Because aptamer immobilization is not limited to mobile, bead-based supports (e.g. magnetic beads), they have been applied in resin-based chromatography systems for semi-continuous flow cell purification.^[31]

Leveraging the reversible nature of aptamer binding, we herein report an aptamer-based, traceless and serial cell selection system for the positive isolation of T cell subsets for adoptive T cell manufacturing. First, by a modified cell-SELEX method, we discovered a CD62L aptamer that binds with sub-nanomolar affinity to human T cells. We showed that the aptamer binds to both soluble CD62L and membrane-bound CD62L with high affinity and specificity and characterized its binding properties via flow cytometry, biolayer interferometry, siRNA knockdown and competition assays. We then engineered the aptamer and designed a reversal agent to enable specific, reversible CD62L aptamer binding. We combined this aptamer with our previously discovered CD8 aptamer and their respective reversal agents to enable the sequential and traceless isolation of CD62L⁺CD8⁺ T cells. CD62L⁺CD8⁺ T cells isolated using the aptamer-based, serial positive selection approach are phenotypically similar to cells isolated using the traditional antibody-based, negative-then-positive selection approach. However, the aptamer-based approach results in a near 2-fold increase in yield that is critical for clinical manufacturing of CAR T cells derived from naïve and memory T cells. Finally, we showed that CD19-directed CAR T cells manufactured from aptamer-selected CD62L⁺CD8⁺ T cells exhibit distinct functional and phenotypical advantages compared to CAR T cells manufactured from bulk CD8⁺ T cells when challenged against Raji lymphoma cells in vitro. By enabling the precise selection of T cell subsets with high yield and purity, this approach has the potential to improve the clinical efficacy of current adoptive T cell therapies.

Results and Discussion

Discovery and Validation of a CD62L Aptamer by Cell-SELEX

We conducted a modified cell-SELEX experiment to identify T cell-binding aptamers by using primary human CD4⁺ T cells for positive selection and J.RT3-T3.5 cells (CD3⁻ and CD28⁻ mutant derivatives of Jurkat leukemia cells) or T cell-depleted apheresis product as cells for negative selection (Fig. 1A).^[32] We used CD4⁺ T cells for positive selection because the library partitioned strongly towards CD8-binding aptamers when CD8⁺ T cells were present during positive selection.^[20] In the last round of selection, we incorporated a competitive cell-SELEX strategy by incubating the round 7 library with peripheral blood mononuclear cells (PBMCs), followed by isolating untouched T cells and associated aptamers using negative magnetic separation. We monitored selection progress with flow cytometry binding study. After 8 rounds of selection, we observed preferential binding of our library to CD4⁺ T cells compared to T cell-depleted apheresis product (Fig. 1B).

Figure 1.

A. Schematic of cell-SELEX. An N52 DNA library underwent 7 rounds of cell-SELEX using primary human T cells for positive selection and J.RT3-T3.5 cells or T cell-depleted apheresis product for negative selection. In round 8, the library was incubated with primary human PBMCs, and T cells. The bound sequences were isolated by negative magnetic selection. B. Flow cytometry analysis of binding of 250 nM round 8 pool or a random aptamer control (RAN) to CD4⁺ T cells (left) and T cell-depleted PBMCs (right). C. Binding curve of A1 to CD4⁺ T cells thawed after cryopreservation. Data points and error bars, and K_D values, represent the mean ± s.d.; n = 3 independent experiments. D. Binding curve of A1 to CD4⁺ T cells cultured overnight in RPMI 10% FBS with 20 ng/mL IL-2. Data points and error bars, and K_D values, represent the mean ± s.d.; n = 3 independent experiments.

We sequenced DNA pools by next generation sequencing (NGS) to identify individual aptamer sequences and analyzed the abundance and enrichment of unique sequences between rounds with the FASTAptamer tool kit (Table S3)^[33], focusing on the top 50 most prevalent sequences in the round 8 pool for motif prediction and phylogenic tree generation with the MEME toolkit and Simple Phylogeny, respectively.^[34–35] We selected five highly abundant and enriched sequences from the round 8 pool, A1, A3, A6, A29, and A40, each representing different branches of the phylogenetic tree, for evaluation (Fig. S1). Only A1 (with motif 2) showed binding to most, albeit not all, CD4⁺ T cells cultured in media (Fig. S2). Recognizing the broad potential applications of T cell-binding aptamers in cell isolation, immunomodulation and targeted delivery, we proceeded with further characterization of the A1 sequence.

We first evaluated the binding affinity of A1 to its target cells by flow cytometry. A1 binds with sub-nanomolar affinity to both cryopreserved CD4⁺ T cells used immediately after thawing (Fig. 1C) and cryopreserved CD4⁺ T cells cultured overnight in media (apparent K_D = 0.31±0.09 nM; Fig. 1D). Intriguingly, the saturated A1 binding to cells directly after thawing was approximately a third of the saturated binding to cells cultured overnight, suggesting upregulation of the target receptor during overnight incubation. A time-course experiment showed that A1 binds only a minor fraction of CD4⁺ T cells immediately after thawing; however, the bound fraction increased over time in culture, likely reflecting progressive expression of the target receptor (Fig. S3). A1 binding to CD4⁺ T cells increases approximately 4 hours after incubating the cells in RPMI 10% FBS, reaching maximum levels after around 15 hours. In addition to CD4⁺ T cells, A1 shows binding to a subpopulation of CD8⁺ T cells, B cells and NK cells in PBMCs (Fig. S4). Despite being used in initial selection rounds as a negative selection cell source, A1 also shows high levels of binding to J.RT3-T3.5 cells (Fig. S5A) with sub-nanomolar affinity (Fig S5B), potentially due to insufficient negative selection stringency in rounds 2 and 3.

To determine the target receptor of A1, we adapted an aptamer-based pulldown assay.^[36–38] We used biotinylated A1 aptamer immobilized on streptavidin-coupled Dynabeads to pull down target proteins from membrane protein extract of J.RT3-T3.5 cells – a cell line we selected for their ease of culture and rapid expansion. We analyzed captured proteins after elution by liquid chromatography-tandem mass spectrometry (LC-MS/MS) and by SDS-PAGE with colloidal blue staining for visualization (Fig. S6). A1 enriched one protein band at about 70 kDa relative to the control. In contrast to most previously reported methods which analyze only excised protein bands after SDS-PAGE, we analyzed all eluted proteins by LC-MS/MS to ensure the identification of poorly resolved or low-abundance proteins even after enrichment, an approach informed by our extensive experience with prior challenges in aptamer-based pulldown assays.

To identify the target from the LC-MS/MS results, we screened for proteins that are 1) uniquely identified in the aptamer sample but absent in the control sample and 2) localized to the cell membrane. Among the potential protein receptors identified by LC-MS/MS, CD62L/L-selectin emerged as a strong candidate. CD62L expression is reduced in cryopreserved PBMCs, with recovery occurring during post-thaw culture,^{[23, 39–41]} which aligns with our observation that A1 binding to thawed CD4⁺ T cells increases with time in culture. Additionally, our observed binding of A1 to various leukocyte populations is consistent with CD62L’s broad expression on circulating lymphocytes, B cells, and polymorphonuclear granulocytes.^[42] Finally, the enriched band at approximately 70 kDa from SDS-PAGE analysis is also consistent with the reported molecular weight of CD62L with glycosylation.^[13]

To assess A1 binding to CD62L, we first co-stained J.RT3-T3.5 cells with A1 and a monoclonal CD62L antibody (clone DREG-56). The strong, positive correlation between A1 and the antibody suggests the aptamer likely binds to CD62L (Fig. 2A). We next incubated J.RT3-T3.5 cells with a fixed concentration of DREG-56 and varying concentrations of A1 to evaluate their competition for binding to J.RT3-T3.5 cells (Fig. 2B). The binding of DREG-56 to J.RT3-T3.5 cells is reduced to 30% when A1 was added at an equimolar concentration and further decreased to 10% when A1 was added at 4-fold molar excess. These results suggest that A1 and the DREG-56 antibody likely share a binding site on CD62L and that A1 exhibits a higher binding affinity for CD62L than DREG-56.

Figure 2.

Characterization of A1 binding to CD62L and truncation designs. A. Overlaid flow cytometry plots of unstained, Cy5-labeled A1 stained, Alexa Fluor (AF) 488-labeled CD62L antibody (clone DREG-56) stained, and Cy5-labeled A1 and AF488-labeled CD62L antibody (clone DREG-56) co-stained J.RT3-T3.5 cells. B. Competitive binding of varying concentrations of A1 with 50 nM CD62L antibody (clone DREG-56) to J.RT3-T3.5 cells. Data points and error bars represent the mean ± s.d.; n = 3 independent experiments. ns > 0.5, ****P < 0.0001 (ordinary two-way ANOVA with Sidak correction). C. Aptamer truncation from CD62LApt.88 to CD62LApt.52 based on NUPACK 3-predicted minimum free energy secondary structures (temperature = 4 °C; Na⁺ = 137 mM; Mg²⁺ = 5.5 mM). D. Flow cytometry binding curves of CD62LApt.88 and CD62LApt.52 to J.RT3-T3.5 cells. Data points and error bars, and K_D values, represent the mean ± s.d.; n = 3 independent experiments. E. Binding kinetics of 50 nM CD62LApt.52 immobilized on streptavidin sensors to varying concentrations of recombinant human CD62L by bio-layer interferometry (BLI). The association and dissociation phases are separated by the vertical dotted line. K_D values were calculated by performing a global fit of the kinetic data at different concentrations of CD62L protein to a 1:1 binding model. K_D values represent mean ± standard deviation; n = 3 individual concentrations.

To further validate specific A1 binding to CD62L, we knocked down CD62L expression in J.RT3-T3.5 cells with short interfering RNA (siRNA) duplexes targeting the SELL gene, which encodes CD62L, and compared aptamer and antibody binding to J.RT3-T3.5 cells transfected with the SELL siRNA versus a control siRNA. We confirmed using antibody staining that CD62L expression was reduced to 17.00% by siRNA (Fig S7A). The relative aptamer binding to these cells was remarkably similarly reduced to 16.92% (Fig S7A), demonstrating that A1 specifically binds to CD62L on the cell surface. We next interrogated the direct binding kinetics of A1 to recombinant human CD62L protein by biolayer interferometry (BLI) using biotinylated A1 immobilized on streptavidin sensors and calculated the dissociation constant (K_D) for A1 binding to recombinant human CD62L protein to be 4.91±0.20 nM (Fig. S7B). Collectively, the antibody co-staining and competition experiments, siRNA knockdown assays, and BLI results rigorously conclude that A1 binds to CD62L/L-Selectin with high affinity and specificity. A1 was therefore renamed to CD62LApt.88, where 88 denotes the length of the aptamer in nucleotides.

Aptamer Truncations and Comparison with Previously Reported CD62L Aptamers

To reduce the downstream production cost, streamline design control and potentially improve aptamer affinity and specificity, we designed aptamer truncations based on NUPACK-predicted secondary structures.^[43] Guided by the predicted secondary structure, we successively truncated CD62LApt.88 at the 5’ and 3’ ends to 71, 52, and 41 nucleotides in length, forming CD62LApt.71, CD62LApt.52, and CD62LApt.42 (Fig. S8A). We tested binding of FAM-labeled CD62LApt.71, CD62LApt.52, and CD62LApt.42 to JRT.3-T3.5 cells by flow cytometry (Fig. S8B). CD62LApt.71 and CD62LApt.52, but not CD62LApt.42, retain binding to JRT.3-T3.5 cells. We selected CD62LApt.52 for further characterization due to its smaller size, which provides improved design control (Fig. 2C). CD62LApt.52 retains high-affinity binding to JRT.3-T3.5 cells with an apparent K_D of 0.19±0.01 nM and to recombinant human CD62L with a K_D=5.30±0.04 nM (Fig. 2D and 2E), confirming that the constant regions do not actively participate in aptamer binding to CD62L. Interestingly, in BLI experiments, the curve fitting for CD62LApt.88 binding to recombinant human CD62L showed a high χ² value of 2.03 and a low R² value of 0.90, indicating the binding between CD62LApt.88 and CD62L does not closely conform to a 1:1 binding model (Table S6). In contrast, a low χ² value of 0.06 and a high R² value approaching 1 were calculated for CD62LApt.52 binding to recombinant human CD62L, with minimal changes in the calculated K_D. Additionally, whereas there are rapid onsets of association and dissociation of the CD62L protein to and from CD62LApt.88 during BLI, CD62L associates to and dissociates from CD62LApt.52 in a steadier manner that closely conforms to a 1:1 binding model. These results suggest that truncation of the CD62L aptamer to CD62LApt.52 potentially provided an improvement in binding specificity by removing aptamer regions that may contribute to non-specific interactions, steric hinderance, and/or conformational heterogeneity. We also tested the binding of CD62LApt.52 to recombinant human CD62E and CD62P proteins, which share conserved epitopes with CD62L in the extracellular region.^{[12, 44]} No noticeable binding of CD62LApt.52 to either CD62E or CD62P at up to 1000 nM of protein was detected by BLI, demonstrating high specificity of CD62LApt.52 towards CD62L (Fig. S9).

Several DREG-56-competing aptamers (LD174, LD196, LD201 and sgc3) have been previously reported.^[45–47] Although these aptamers do not share sequence homology or predicted secondary structure with CD62LApt.52 (Fig. S10A), we compared our truncated aptamer with truncated versions of three of the previously reported CD62L aptamers, LD174t1, LD201t1 and sgc3b. We verified by BLI that LD174t1, LD201t1 and sgc3b all bind to recombinant human CD62L, with K_D=6.53 nM, K_D=4.99 nM and K_D=3.71 nM, respectively, in our binding conditions (Fig. S10B). In flow cytometry assays, LD174t1, LD201t1 and sgc3b all bind to JRT.3-T3.5 cells with single-digit or sub- nanomolar affinity (Fig. S10C). They also compete with CD62LApt.52 for binding to JRT.3-T3.5 cells, although with slightly lower affinities in our binding conditions (Fig. S10D). Thus, although LD174t1, LD201t1, sgc3b, and CD62LApt.52 have different sequences and predicted secondary structures, the four aptamers share a binding site on CD62L.

Reversal Agent Design for Rapid Aptamer Release

A major advantage of aptamers as capture agents in cell selection is the ability to rapidly and specifically reverse aptamer binding to cells after separation, rendering the selected cells label-free and available for re-labeling and further selection. We adopted the toehold-mediated strand displacement method for reversing aptamer binding in consideration of the specificity, low cost and gentle mechanics afforded by this method.^{[20, 22, 48]} We hypothesized that the multi-nucleotide sequence at the 3’ end of CD62LApt.52 can serve as a natural toehold, allowing the reversal agent direct access to the aptamer to disrupt its secondary structures. We designed a 28-nucleotide reversal agent predicted to extensively disrupt CD62LApt.52’s secondary structure from the 3’ end (Fig. S11A). Indeed, addition of this reversal agent at >10-fold molar equivalent to CD62LApt.52 rapidly reverses aptamer binding to J.RT3-T3.5 cells by >90% after a 10-min incubation at room temperature in static well experiments (Fig. S11B).

Optimization of Capture and Release Conditions for High-Yield CD62L Selection

Towards our goal of traceless CD62L⁺CD8⁺ T cell selection, we optimized a CD62LApt.52-based, traceless magnetic selection system. We first optimized the CD62L aptamer for magnetic capture of CD62L⁺ cells in a model mixture composed of CD62L⁺ J.RT3-T3.5 cells and CD62L⁻ K562 cells. Using biotinylated CD62LApt.52 to label target cells followed by capture with Anti-Biotin Microbeads (Miltenyi Biotec), we achieved high CD62L purity (>98%) but low CD62L⁺ cell yield (~11%; data not shown), with high numbers of CD62L⁺ cells present in the flow through (non-captured) fraction (Fig. S12A). We hypothesized that, despite including a hexa-ethylene glycol spacer, the proximity of the 5’ biotin label to the truncated aptamer structure imposes steric hindrance that restricts access of the Anti-Biotin Microbeads to the biotinylated aptamer/protein complex. We designed CD62LApt.52.s20 by introducing a 20-nucleotide spacer at the 5’ end to extend the biotin moiety away from the aptamer structure (Fig. S12B). The spacer arm did not affect the binding affinity of CD62LApt.52.s20 but its introduction improved the capture yield to 98% (Fig. S12C and S12D). We then demonstrated that CD62LApt.52.s20 immobilized on microbeads at down to 10 nM retained capture efficiency and yield of the biotinylated aptamer (Fig. S13A). Lastly, we applied the designed reversal agent to tracelessly elute CD62L⁺ cells. Despite our efficient aptamer release in static well experiments (Fig S11B), we observed low cell release (<20%) from magnetic columns under the same experimental conditions (Fig. S13B), possibly owing to insufficient reversal agent contact with the aptamers in the densely packed column matrix. We therefore designed a 2-step elution strategy, where additional reversal agent is passed into the column after the first elution and allowed to displace any remaining aptamers. The 2-step elution strategy markedly increased cell release and led to ~76% CD62L⁺ cell yield that is suitable for cell selection applications, likely resulting from improved diffusion dynamics of the reversal agent in the matrix (Fig. S13B).

Serial, Traceless Selection of CD62L⁺CD8⁺ T Cells from PBMCs

Towards multi-parameter cell sorting at the clinical scale, we developed a traceless method to select CD62L⁺CD8⁺ T cells from PBMCs by using our CD62LApt.52.s20 and CD8 aptamer (CD8Apt) with their respective reversal agents (Fig. 3A).^[20] We benchmarked the aptamer-based, serial positive selection strategy against the established antibody-based, negative selection-then-positive selection approach in terms of yield, purity and cell phenotype using cryopreserved PBMCs from 3 donors.^[14] For aptamer-based selection, we tracelessly isolated CD8⁺ cells using CD8Apt followed by reversal agent elution and then enriched for the CD62L⁺CD8⁺ cell population using CD62LApt.52.s20 selection. We compared our approach with antibody-based isolation following two commercial protocols (Miltenyi Biotec). We isolated CD62L⁺CD8⁺ T cells first by negative selection using a commercial antibody cocktail to deplete non-CD8 T cells, followed by positive selection on untouched CD8⁺ T cells with CD62L Microbeads (Miltenyi Biotec). For all selections steps, we collected column flow through (FT, unselected cells), flush (FL, cells captured by aptamer- or antibody-immobilized beads and removed from the column by plunger flush), and, in the case of aptamer-based selection, reversal agent elution (RAE, cells captured by aptamer-immobilized beads and eluted from the column by reversal agent treatment). The collected cells were stained for CD3, CD8, CD62L, CCR7, CD45RA and CD45RO and analyzed by flow cytometry.

Figure 3. Traceless selection of CD62L⁺CD8⁺ T cells from PBMCs using an aptamer-based, serial selection strategy.

A. Schematic representation of aptamer-based, traceless CD8⁺ cell isolation followed by aptamer-based, traceless CD62L⁺ cell isolation. B. Representative flow cytometry analysis of CD8 expression and CD62L expression in the starting PBMCs and in the flow through (FT) and reversal agent elution (RAE) fractions from the aptamer-based, serial positive selection strategy. APC: allophycocyanin; PE: phycoerythrin. C. Flow cytometry analysis of CD62L⁺CD8⁺CD3⁺ cell purity in different fractions of antibody- and aptamer-based isolations. Individual data points represent different PBMC donors from separate isolation experiments. Horizontal lines and error bars represent the mean ± s.d.; n = 3 independent experiments from separate PBMC donors; ns > 0.05, *P ≤ 0.05 (paired one-way ANOVA with Tukey’s test). D. Flow cytometry analysis of CD62L⁺CD8⁺CD3⁺ cell yield from PBMCs for antibody- and aptamer-based isolations. Individual data points represent different PBMC donors from separate isolation experiments. Horizontal lines and error bars represent the mean ± s.d.; n = 3 independent experiments from separate PBMC donors; *P ≤ 0.05 (paired one-way ANOVA). E. Flow cytometry analysis of the CD8⁺ T cell phenotypes in the starting PBMCs and the reversal agent elution (RAE) fractions after CD8 aptamer selection and CD62L aptamer selection. Numbers represent the mean values from n = 3 independent experiments from separate PBMC donors. Individual donor values and data for antibody-based isolations can be found in Fig. S16.

Using the aptamer-based strategy, we show near-complete selection of CD8⁺ cells in the first step, followed by near-complete selection of CD62L⁺ cells in the second step (Fig. 3B), on average enriching CD62L⁺CD8⁺ cells from a starting population of 5.70% in PBMCs to 72.97% or 75.03% by RAE or flush collection, respectively (Fig. S14A). These purity levels are similar to those obtained by the traditionally used antibody-based method. We analyzed the CD62L⁺CD8⁺ purity in both the RAE and flush fractions since both final elution strategies are viable approaches in eventual manufacturing processes. We also analyzed the CD62L⁺CD8⁺CD3⁺ purity to assess the specific enrichment of CD62L⁺CD8⁺ T cells. The average CD62L⁺CD8⁺CD3⁺ purity was 69.7% in the CD62L RAE fraction and 73.7% in the CD62L flush fraction, again commensurate to the purity of the cells selected using the commercial antibody-based method that had a large panel of antibodies to deplete non-CD8 T cells (Fig. 3C). Our results demonstrate that the aptamer-based serial positive selection method enables the isolation of CD62L⁺CD8⁺ T cells with high purity that is comparable to the conventional antibody-based method, while requiring only two synthetic selection agents (and their corresponding reversal agents) compared to more than ten selection antibodies needed in the conventional approach, thereby substantially reducing manufacturing costs, complexity, and variability.

We next determined the cell yields (percentage of target cells captured from the starting population) obtained by the two selection approaches. The aptamer-based method delivers an overall bulk CD62L⁺CD8⁺ yield (percentage of CD62L⁺CD8⁺ cells captured from the starting PBMCs) of 59.14% in the combined RAE and flush fractions, compared to 29.89% using the antibody-based method (Fig. S14B). Specifically, we compared the CD62L⁺CD8⁺ T cell yield (percentage of CD62L⁺CD8⁺CD3⁺ cells captured from the starting PBMCs) obtained by the two selection approaches. The aptamer-based method provides a significantly improved CD62L⁺CD8⁺ T cell yield over the antibody-based method, delivering on average an overall yield of 66.04% in the combined RAE and flush fractions, compared to 36.56% using the antibody-based method (Fig. 3D). Further analysis in each selection fraction revealed that the aptamer-based method affords slightly higher CD8⁺CD3⁺ cell yield in the CD8 selection step (67.43% for aptamer-based positive selection vs 62.42% for antibody-based negative selection; Fig. S14C) and significantly superior CD62L⁺ cell yield in the CD62L selection step (97.88% for aptamer-based positive selection vs 58.09% for antibody-based positive selection; Fig. S14D). Even if we consider only the RAE fraction, which contains label-free cells that are available for re-labelling and subsequent selections, the CD62L⁺ cell yield remains significantly higher for the aptamer-based method than the antibody-based method at 80.96% (Fig. S14D).

In our study, we observed the highest yield improvement from the CD62L selection step. We speculate that more efficient selection of CD62L⁺ cells from the CD62L selection step on the column results from the high affinity and specificity of CD62LApt.52, allowing the aptamer to tightly bind small, well-defined epitopes on CD62L. The aptamer’s small size relative to antibodies may additionally allow for a higher density of aptamers to be immobilized on microbeads, increasing the likelihood of binding to CD62L. The smaller size of aptamers also minimizes potential steric hindrance when targeting CD62L, which forms clusters on cell surfaces,^[49] thereby enabling more efficient magnetic labeling and isolation. While we did not observe a significant improvement in yield from the CD8 isolation step, positive CD8 isolation with reversible aptamer labeling enabled consistently high yield while requiring only one aptamer and its corresponding reversible agent, simplifying the selection workflow and obviating the need for large antibody panels for selection.

Despite using vastly distinct antibody sets, the yield we observed from using the antibody-based, negative-then-positive selection method corroborates the <30% cell recovery efficiencies previously reported by Terakura et al.,^[14–15] potentially reflecting intrinsic limitations associated with a negative-then-positive selection approach. In a clinical study that evaluated the potency of CAR T cells manufactured from defined CD4 and CD8 subsets, only 16 of 30 patients had sufficient CD8⁺ T_CM counts for selection using the protocols reported by Terakura et al.^[3] By offering a significant improvement in yield while ensuring a label-free product, our aptamer-based, serial selection method addresses the limitations of traditional antibody-based selection strategies and facilitates the manufacturing of defined CAR T cells products for clinical evaluation.

We also characterized the specific subsets of T cells that are enriched from aptamer- and antibody-based CD8 and CD62L selections. Overall, the isolated cells from aptamer-based and antibody-based selections display similar phenotypes, with CD45RA⁺CD45RO⁻ naïve CD8⁺ T cells and CD45RA⁺CD45RO⁺ transitional CD8⁺ T cells accounting for over 60% of the total enriched cells (Fig S16). Specifically, CD45RA⁺CD45RO⁻ naïve CD8⁺ T cells and CD45RA⁻CD45RO⁺/CCR7⁺ or CD62L⁺ central memory CD8⁺ T cells are significantly enriched from PBMCs following aptamer-based serial CD8 and CD62L selections by 13.56-fold and 13.44-fold, respectively (Fig 3E). Additionally, CD62L selection significantly reduced the percentage of CD45RA⁺CD45RO⁻CCR7⁻CD62L⁻ terminal effector CD8⁺ T cells from 20.81% to 7.86% and contaminating non-CD8⁺ T cells from 18.47% to 10.29% (Fig S16). These results further demonstrate that aptamer-based serial selection enables the selection of CD62L⁺CD8⁺ T cells that are phenotypically similar to antibody-selected cells while using only minimal synthetic reagents, and that aptamer-based CD62L selection following CD8 selection efficiently enriches for less differentiated CD8⁺ T cell subsets and depletes more differentiated CD8⁺ T cell subsets, to yield a CD8⁺ T cell population with defined and favorable phenotypes.

Interestingly, the average CD62L expression level is lower in aptamer-selected cells (CD62L Aptamer RAE and FL) than in antibody-selected cells (CD62L Antibody FL) (Fig. S14E). Similarly, the average CD62L expression level is lower in the flow through fraction (containing unselected cells) for aptamer selection (CD62L Aptamer FT) than that for antibody selection (CD62L Antibody FT). This indicates that the aptamer-based selection more effectively captures CD62L⁺ cells that have lower CD62L expression, leading to a higher overall CD62L cell yield. Given that CD62L expression varies among individuals, aptamer-based selection ensures consistently high yield despite donor-to-donor variability in CD62L expression patterns.^[19]

Serial positive selection afforded by aptamer-based selection also offers greater flexibility and modularity compared to the traditional negative-then-positive selection strategy. This flexibility is exemplified by our ability to achieve similarly high levels of CD62L⁺CD8⁺ T cell purity (77.77%) and yield (56.64%) for isolating CD62L⁺CD8⁺ T cells when the order of isolation was reversed to CD62L isolation followed by CD8 isolation, without additional optimization (Fig. S15).

CAR T Cell Manufacturing from Aptamer-Selected CD62L⁺CD8⁺ T Cells

The pre-selection of less differentiated T cells, including naive cells (T_N), stem cell-like memory cells (T_SCM) and central memory cells (T_CM), has emerged as a promising strategy in manufacturing more effective adoptive T cell therapy because these cells show consistently superior in vivo expansion, persistence, and antitumor capabilities compared to more differentiated effector memory (T_EM) and terminal effector (T_EFF) cells.^{[1–2, 6, 9]} Notably, the aptamer-based selection approach described here enriches for all less differentiated T_N, T_SCM, and T_CM subsets from CD8⁺ cells through CD62L selection alone. By contrast, conventional approaches for isolating separate T_N, T_SCM, and T_CM subsets require CD45RO/CD45RA selection or depletion steps in addition to CD8 and CD62L selection, which increases manufacturing complexity and cost.^{[2, 5]}

To investigate whether CD62L enrichment alone is sufficient to enhance CD8⁺ CAR T cell functionality without additional CD45RO/CD45RA selection, we produced CD19-directed CAR T cells from aptamer-isolated CD62L⁺CD8⁺ T cells and aptamer-isolated CD8⁺ T cells from healthy donor PBMC populations (Fig. 4A). We transduced isolated cells with a third-generation lentivirus vector encoding a second-generation CD19scFv(FMC63)-41BB-CD3ζ CAR, a truncated epidermal growth factor receptor (EGFRt) as a surrogate transduction marker, and a double mutant human dihydrofolate reductase (DHFRdm), which confers methotrexate resistance to transduced cells (Fig. 4B).^{[20, 50–51]} We observed similar transduction efficiencies (>40%) in both populations, with >95% EGFRt expression at similar levels after methotrexate selection (Fig. S17). Despite initial differences in T cell phenotypes between the two groups after isolation (Fig. 3E and S16), at the end of production we observed minimal difference in T cell phenotypes as analyzed by surface expression of CD45RA, CD45ARO, CD62L and CCR7 (Fig 4C). The high levels of CD45RA⁺CD45RO⁺ cells in the mock cells and both CAR T cell groups may reflect activation-induced CD45RO expression on naïve T cell precursors in the presence of interleukin (IL)-7 and IL-5 during expansion.^{[11, 52]} At the end of production, CAR T cells derived from CD62L-enriched CD8⁺ T cells and bulk CD8⁺ T cells show similar levels of exhaustion (PD-1, TIM-3 and LAG-3) and activation (HLA-DR) marker expressions (Fig. S18).

Figure 4. Manufacturing and in vitro characterization of CAR T cells from CD8 T cells and CD62L-enriched CD8 T cells by aptamer-based isolation.

A. Schematic summary of CAR T cell manufacturing with aptamer-based cell isolations. B. The second-generation anti-CD19 CAR construct used for producing CD19-directed CAR T cells. The polycistronic construct is driven by the EF1α promoter and encodes a CD19-directed CAR, a methotrexate selection marker (DFHRdm), and a EGFRt reporter, separated by 2A self-cleaving sequences. C. Flow cytometry analysis of the cell phenotypes at the end of CAR T cell production. Naïve CD8⁺ T cells are defined as CD8⁺CD3⁺CD45RA⁺CD45RO⁻/CCR7⁺ or CD62L⁺, transition cells are defined as CD8⁺CD3⁺CD45RA⁺CD45RO⁺ and central memory CD8⁺ T cells are defined as CD8⁺CD3⁺CD45RA⁻CD45RO⁺/CCR7⁺ or CD62L⁺. Data are mean ± s.d., n = 6 independent experiments from separate PBMC donors, ns > 0.05 (paired two-way ANOVA with Tukey’s test). D. In vitro antitumor cytotoxicity against CD19⁺ Raji cells and control CD19⁻ K562 cells at a 10:1 effector to target cell ratio. Data are mean ± s.d., n = 4 independent experiments from separate PBMC donors, ns > 0.05, **P ≤ 0.01, ***P ≤ 0.001 (paired two-way ANOVA with Tukey’s test). E. In vitro TNF-ɑ release from mock and CAR T cells against Raji or K562 cells when incubated at a 2:1 effector to target cell ratio. Data are mean ± s.d., n = 4 independent experiments from separate PBMC donors, *P ≤ 0.05 (paired two-away ANOVA with Tukey’s test). F. Flow cytometry analysis of TIM-3 expression on the CAR T cells 72 hours after co-culture with Raji cells at a 2:1 effector to target cell ratio. Data are mean ± s.d., n = 4 independent experiments from separate PBMC donors, **P ≤ 0.01 (paired two-tailed t-test). G. Flow cytometry analysis of combined relative expression PD-1, TIM-3 and LAG-3 on the CAR T cells 72 hours after co-culture with Raji cells at a 2:1 effector to target cell ratio. The combined relative expression was calculated as the mean of the relative expression of each exhaustion marker on the CD62L-enriched CD8⁺ CAR T cells normalized against the expression on donor-matched bulk CD8⁺ CAR T cells. Data are mean ± s.d., n = 4 independent experiments from separate PBMC donors, **P ≤ 0.01 (paired one-away ANOVA with Tukey’s test).

Although we did not detect an obvious difference in T cell subset distribution after expansion of the two CAR T cell populations, we found distinctions in effector functions of the CAR T cells when challenged against CD19⁺ B lymphoma Raji cells. Direct tumor cell killing efficiencies of CD19-directed CAR T cells derived from CD62L-enriched CD8⁺ T cells and from bulk CD8⁺ T cells are similar (Fig. 4D). However, during co-culture experiments, CAR T cells derived from CD62L-enriched CD8⁺ T cells produced substantially higher levels of tumor necrosis factor (TNF)-α and slightly higher levels of IL-2 (Fig. 4E and Fig. S19A). Similar levels of interferon (IFN)-γ secretion were observed (Fig. S19B). Importantly, after a 3-day co-culture with Raji cells, CAR T cells derived from CD62L-enriched CD8⁺ T cells expressed lower levels of the exhaustion marker TIM-3, characterized by consistently lower TIM-3 MFI and a lower percentage of TIM-3⁺ CAR T cells, across all PBMC donor groups (Fig. 4F and Fig. S20). We observed no significant differences in individual PD-1 or LAG-3 expression between the CAR T cell populations (Fig. S20). However, CAR T cells derived from CD62L-enriched CD8⁺ T cells overall exhibited a reduced exhausted-like status, as displayed by lower cumulative expression levels of PD-1, LAG-3 and TIM-3 (Fig. 4G). These results demonstrate that aptamer-mediated enrichment of less differentiated CD8⁺ T cells facilitates the manufacturing of CAR T cells with similar tumor-killing efficiency but distinct functional advantages compared to CAR T cells manufactured from bulk CD8⁺ T cells. CAR T cells derived from CD62L-enriched CD8⁺ T cells exhibit increased pro-inflammatory cytokine secretion and reduced exhaustion marker expression, potentially leading to a more potent and sustained therapeutic effect.

Conclusion

In this work, we report the discovery and extensive characterization of a novel, high-affinity CD62L/L-selectin aptamer and demonstrate the application of this aptamer and our previously discovered CD8 aptamer for the traceless and serial positive selection of CD62L⁺CD8⁺ T cells with high purity and yield. We show that CD62L⁺CD8⁺ T cells, comprising of naïve, stem-cell like memory, and central memory CD8⁺ T cells, can be isolated and enriched to ~70% purity, comparable to antibody-based isolations, with >65% yield, a near two-fold improvement over the antibody-based, negative-then-positive selection approach. Our studies demonstrate that the aptamer-based serial positive selection approach is a feasible method for the selection and enrichment of specific T cell subsets for manufacturing CAR T cells with distinct functionalities. With the future discovery of aptamers against CD4, CD45RA and CD45RO, among other targets, aptamers can be used to realize clinical scale, cost-effective multi-parameter cell selection in a single synthetic system to facilitate clinical evaluation of CAR T cells derived from complex T cell subsets in large cohorts.^[53] Ultimately, aptamer-based cell isolation strategies may be inexpensively integrated into existing adoptive T cell manufacturing processes to improve the accessibility and clinical performance of adoptive T cell therapies. Importantly, this aptamer-based selection approach can be broadly applied to the manufacturing of other cell therapies where well-defined cell subsets are desired.^[54–55]

Materials and Methods

Cell culture and PBMC isolation:

The Jurkat (TIB-152), J.RT3-T3.5 (TIB-153), Raji (CCL-86) and K562 (CCL-243) cell lines were purchased from ATCC. Positively selected CD4⁺ T cells, T cells, and T cell-depleted apheresis were provided by Juno Therapeutics. Peripheral blood mononuclear cells (PBMCs) were isolated from Leukocyte Reduction System (LRS) chambers (Bloodworks Northwest) by density gradient centrifugation over Ficoll-Paque (GE). The above cells were cultured in RPMI 1640 medium (Gibco and Corning) supplemented with 10% heat-inactivated FBS (Life Tech and VWR). The CD4⁺ T cells used in SELEX were rested in RPMI 10% FBS supplemented with 20 ng/mL recombinant human IL-2 (Miltenyi Biotec) overnight prior to selection to remove the cell-freezing reagent, DMSO. For cell selection experiments, PBMCs were used either immediately after isolation from LRS chambers or after incubation in RPMI 10% FBS for 15 hours post-thaw to restore CD62L expression. HEK293T cells were also purchased from ATCC and were cultured in DMEM (Gibco) supplemented with 10% heat-inactivated FBS (Life Tech and VWR).

Oligonucleotides and buffers:

Aptamer sequences used in the study are listed in Table S4. All oligonucleotides were synthesized by Integrated DNA Technologies (IDT). The starting library and individual aptamers were purified by high performance liquid chromatography (HPLC) and the primers were purified by a standard desalting procedure. Buffer compositions were adapted from previously published protocols.^[1] The base wash buffer (WB) was prepared by supplementing Dulbecco’s phosphate-buffered saline (DPBS) with calcium and magnesium (Corning) with 5 mM MgCl₂ and 25 mM D-(+)-Glucose (Sigma-Aldrich). To anneal individual aptamers, aptamers were diluted to 1 μM in WB and then heated at 95°C for 5 min, followed by snap cooling on ice for at least 15 min. For SELEX, the libraries were annealed at the indicated concentrations.

Competitive cell-SELEX:

Cell-SELEX protocols were adapted from previously published protocols.^[1–2] In round 1, 40 nmol of the initial N52 library (theoretically ~10¹⁶ unique sequences) was annealed and incubated with 40 × 10⁶ mixed T cells in 2800 μL WB supplemented with 0.1% BSA (Miltenyi Biotec) for 1 hour at 4°C to avoid aptamer internalization. After incubation, the cells were washed with WB to remove unbound sequences. From rounds 2–7, positive selection was performed in 400 μL total volume with increasing stringency by decreasing the library concentration, decreasing the number of target cells, reducing the incubation time, increasing the number of washes, and/or increasing the concentration of protein blocker. In round 8, PBMCs were used for competitive positive selection, where the sequences are selected against the target (CD4⁺ T cells) in the presence of competitors (non-CD4⁺ T cells in PBMCs). Untouched T cells were then isolated by magnetic activated cell sorting (MACS) using a Pan T Cell Isolation Kit (Miltenyi Biotec). After positive selection, bound sequences were eluted by heating the cells at 95°C for 10 min in 400 μL molecular biology grade water. Cell debris was pelleted by centrifugation at 13,100g for 5 min at 4°C, while the eluted sequences were simultaneously re-folded. The supernatant containing eluted sequences was used for PCR amplification (rounds 1 and 8) or negative selection (rounds 2–7). For negative selection, the eluted sequences were incubated with 10 × 10⁶ J.RT3-T3.5 cells (rounds 2–3) or T cell-depleted PBMCs (rounds 4–7) with the same protein blocker and for the same time that were used in positive selection. The sequences bound to the undesired cells were removed by centrifugation at 13,100g for 5 min at 4°C, and the supernatant was collected. After each round, the PCR cycles were optimized to amplify the libraries from the supernatant with Phusion High-Fidelity DNA polymerase (NEB), dNTP (QIAGEN), FAM-labeled forward primer (IDT), and biotinylated reverse primer (IDT). To prepare single stranded DNA (ssDNA) from the PCR product, the amplicons were first captured on High Capacity Neutravidin Agarose Resin (Thermo Scientific) via the biotinylated reverse strand, and the forward strand was eluted from the resin by denaturation with 500 μL of 200 mM NaOH (Sigma-Aldrich). The resulting ssDNA was desalted into 1 mL molecular biology grade water (Corning) using a NAP-5 desalting column (Cytiva), dried on a Savant ISS110 SpeedVac concentrator (Thermo Scientific), and annealed in WB at 1 μM for selection in the next round. Detailed experimental conditions in each round are summarized in Table S1.

NGS and Sequence Analysis:

The aptamer pools from each round of SELEX were PCR amplified with barcoded primers as listed in Table S2, purified by gel extraction, and sequenced using the MiSeq Reagent Kit v2 (300 cycles) on a MiSeq System (Illumina) with custom sequencing primers. FASTAptamer v1.0.3 toolkit was used to analyze the FASTA files.^[3] FASTAptamer-count was used to identify top aptamers with the highest frequency (high count) and FASTAptamer-enrich was used to analyze the fold-enrichment of each unique sequence in adjacent rounds. MEME suite v5.2.0 Motif Discovery tool was used to predict binding motifs of top aptamers in the final round based on their sequence similarities.^[4] The top 50 aptamer sequences in the final round were used for phylogenic tree generation with Simple Phylogeny^[5] and FigTree toolkit v1.4.4 (tree.bio.ed.ac.uk/software/figtree/) to assess the similarities among sequences. The NUPACK web application was used to simulate aptamers folding into their thermally stable secondary structures.^[6]

Antibodies and flow cytometry:

The following antibodies were used for flow cytometry: FITC anti-human CD3 (1:100, BioLegend, UCHT1), PE anti-human CD3 (1:200, BioLegend, HIT3a), APC anti-human CD8a (1:100, BioLegend, RPA-T8), PE anti-human CD62L (1:100, Miltenyi Biotec, 145/15), Alexa Fluor 488 anti-human CD62L (1:25, BioLegend, DREG-56), PE/Cyanine7 anti-human CD45RA (1:200, BioLegend, HI100), Brilliant Violet 605 anti-human CD45RO (1:100, BioLegend, UCHL1), PerCP/Cyanine5.5 anti-human CD197 (1:50, BioLegend, G043H7), FITC anti-human CD14 (1:50, BioLegend, M5E2), APC-Cy7-antihuman CD14 (1:200, Molecular probes, 6ID3), Super Bright 702-antihuman CD19 (1:100, eBioscience, SJ25C1), Super Bright 600-antihuman CD56 (1:100, eBioscience, TULY56), FITC-antihuman CD8a (1:100, BioLegend, RPA-T8), Erbitux–biotin (1:500, Jensen Lab), APC anti-human EGFR (1:100, BioLegend, AY13), PerCP/Cyanine5.5 anti-human CD366 (1:100, BioLegend, F38–2E2), APC/Fire^™ 750 anti-human CD223 (1:100, BioLegend, 11C3C65), FITC anti-human CD279 (1:20, BioLegend, EH12.2H7), APC/Fire^™ 750 anti-human HLA-DR (1:100, BioLegend, L243). Flow cytometry was performed on an Attune NxT (Invitrogen) and flow cytometry data were analyzed with FlowJo V10 (FlowJo LLC).

Binding assays:

Cells were washed with DPBS without calcium and magnesium (Gibco) and incubated with Zombie Violet (1:500, 10⁷ cells/mL, BioLegend) in DPBS without calcium and magnesium for live/dead staining for 15 min at room temperature. After live/dead staining, cells were washed with WB supplemented with 1% BSA to quench the remaining dye and aliquoted in a 96-well plate at 1–2 × 10⁵ cells/well for immortalized cells and sorted T cells and 5–10 × 10⁵ cells/well for PBMCs. Prior to staining with antibodies, cells were first blocked with 10 μL/well FcR Blocking Reagent (Miltenyi Biotec) for 10 min at 4°C. For antibody staining without aptamers, cells were stained with antibodies diluted in 100 μL WB supplemented with 1% BSA for 20–30 min at 4°C. For aptamer staining without antibodies, cells were stained with annealed aptamer pools or individual aptamers at the indicated concentrations in 100 μL binding buffer (WB supplemented with 1–2% BSA and 0.1 mg/mL salmon sperm DNA (Invitrogen) and/or 0.1 mg/mL yeast tRNA (Invitrogen)) for 20–30 min at 4°C. For antibody co-staining and competition experiments, the antibody was added together with fluorescently labeled aptamer at the indicated concentrations. For staining cells with biotinylated aptamers, secondary staining with streptavidin-AF647 (1:500, BioLegend) was performed after the washing steps in 100 μL binding buffer for 15 min at 4°C. For reversal agent testing, cells were first stained with 10 nM biotinylated aptamers, washed twice with 200 μL WB supplemented with 1% BSA, and then incubated with the reversal agent (pre-warmed at 37°C) in 100 μL WB supplemented with 0.5% BSA at the indicated concentrations for 10 min at room temperature. The cells were washed twice with 200 μL WB supplemented with 1% BSA to remove released aptamers and excess reversal agent. Finally, the cells were stained with streptavidin-AF647 in 100 μL binding buffer for 15 min at 4°C. After staining, cells were washed twice with 200 μL WB supplemented with 1% BSA, fixed in WB supplemented with 1% BSA and 0.2% PFA, and analyzed on an Attune NxT Cytometer (Invitrogen).

Membrane protein extraction:

Protocols for extracting membrane proteins and performing aptamer-based pulldowns were adapted from previously published protocols.^[7–9] 240 × 10⁶ J.RT3-T3.5 cells were washed 3 times with DPBS and split into two groups, each containing 120 × 10⁶ cells. The cells were lysed with 5 mL hypotonic buffer comprised of 10 mM Tris-HCl pH7.5 supplemented with EDTA-free cOmplete Protease Inhibitor Cocktail (Roche) and 1 mM PMSF (Thermo Scientific) for 30 min at 4°C with end-over-end mixing. Cell membrane debris was pelleted by centrifuging the cell suspension at 16,000g for 15 min at 4°C, after which the pellet was washed 3 times with 5 mL hypotonic buffer to remove intracellular proteins. To extract and solubilize membrane proteins, the membrane pellet was resuspended in 1 mL WB supplemented with 1% Triton X-100, protease inhibitor cocktail, and 1 mM PMSF and incubated for 30 min at 4°C with end-over-end mixing. The mixture was sonicated for 5 min in an ice water bath and pelleted at 16,000g for 15 min at 4°C. The supernatant containing solubilized membrane proteins was collected, snap frozen on dry ice for 10 min, and stored at −80°C before use.

Aptamer-based pulldown:

1 mL thawed protein extracts were spiked with 100 nM biotinylated non-specific (NS) aptamer and supplemented with 0.1 mg/mL yeast tRNA and incubated at 4°C for 30 min with end-over-end mixing to pre-clear the extract. 2 mg washed Dynabeads^™ MyOne^™ Streptavidin C1 (Invitrogen) were added to the mixture and incubated for an additional 15 min to magnetically remove proteins bound to the biotinylated NS aptamer. For the control group, 1.5 mg washed Dynabeads^™ MyOne^™ Streptavidin C1 were first saturated with 50 nmol biotin for 15 min at 4°C, then added to the pre-cleared extract and incubated for 30 min at 4°C with end-over-end mixing. For the aptamer group, the pre-cleared extract was spiked with 100 nM biotinylated CD62LApt.88 and supplemented with 0.1 mg/mL salmon sperm DNA (Invitrogen) and incubated at 4°C for 30 min with end-over-end mixing. 1.5 mg washed Dynabeads^™ MyOne^™ Streptavidin C1 was added to the mixture and incubated for an additional 15 min at 4°C to capture CD62LApt.88-bound proteins. After incubation, beads in both groups were washed 5 times in 1 mL WB with 0.01% Triton X for 5 min at 4 °C with end-over-end mixing. To mildly elute the proteins, the beads were resuspended in 50 μL proteolytic digestion-compatible buffer composed of 5 M urea, 20 mM Tris pH 7.5, 10 mM EDTA and 100 mM NaCl and heated at 37°C for 10 min followed by an additional 15 min at 47°C. The eluted proteins were stored at −80°C before SDS-PAGE and mass spectrometry analyses.

SDS-PAGE:

10 μL of the elution was combined with 10 μL 2x Laemmli sample buffer (Bio-Rad) and heated at 47°C for 15 min to denature the proteins. The proteins were separated by SDS-PAGE on a Novex WedgeWell 4–20% Tris-Glycine gel (Invitrogen) in an XCell SureLock Mini-Cell Electrophoresis System (Invitrogen) according to the manufacturer’s instructions. The gel was stained with a Colloidal Blue Staining Kit (Invitrogen) and imaged on a Gel Doc EZ system (Bio-Rad).

Receptor Identification:

40 μL of the elution from both the control group and the aptamer group were submitted to the Proteomics & Metabolomics Shared Resources at Fred Hutchinson Cancer Center for analysis. The samples were processed and analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) on an Orbitrap Fusion mass spectrometer (Thermo Scientific). Proteins were identified by searching the data against a Uniprot Human database along with common contaminants. The search results were filtered to only include identifications from peptides with a false discovery rate of 1% or less.

Biolayer Interferometry (BLI):

BLI was performed in an Octet RED96e system (ForteBio). Prior to running, Octet SA biosensors (Sartorius) were pre-soaked in 200 μL diluent (WB supplemented with 1% BSA, 0.1 mg/mL yeast tRNA, 0.1 mg/mL salmon sperm DNA and 0.01% Tween-20). For BLI, the SA biosensors were first rinsed in 200 μL diluent for 100 seconds to establish the first baseline. The rinsed sensors were transferred to wells containing 50 nM biotinylated aptamers in 200 μL diluent for aptamer immobilization onto the SA biosensors until a threshold of 0.5 nm is reached. Next, aptamer-immobilized SA biosensors were rinsed in 200 μL diluent for 100 seconds to remove excess aptamers and for an additional 100 seconds in 200 μL diluent to establish the second baseline. The sensors were then transferred to wells containing different concentrations of CD62L (R&D Systems), CD62P (ACROBiosystems), or CD62E (ACROBiosystems) in 200 μL diluent to allow aptamer association with the protein. Finally, the sensors were transferred to wells containing 200 μL diluent to allow dissociation. The association and dissociation durations are indicated on the BLI curves.

siRNA knockdown:

5 × 10⁶ J.RT3-T3.5 cells in log phase growth were washed with EP buffer (MaxCyte) and mixed with 5 pmol each of CD62L siRNA 1 and 2 (10 pmol total) or 250 pmol control siRNA in 50 μL EP buffer. The sample was then loaded in an R-50×3 processing assembly (MaxCyte) and electroporated using the Jurkat protocol on the ExPERT GTx system (MaxCyte). Electroporated cells were stained with 50 nM CD62L antibody (clone DREG-56) or CD62LApt.88 aptamer in 100μL binding buffer at 24 hours and 48 hours post-electroporation, washed, fixed, and analyzed by flow cytometry.

Optimization of labeling and releasing conditions for selection of CD62L⁺ cells:

CD62L⁺ J.RT3-T3.5 cells were stained with 1 μM CellTrace^™ CFSE dye (Invitrogen) at 1×10⁶ cells/mL in DPBS with calcium and magnesium for 20 min at room temperature. The remaining dye was inactivated by diluting the cells 1:3 with RPMI 10% FBS. Stained cells were resuspended in RPMI 10% FBS at 1×10⁶ cells/mL and incubated for an additional 30 min at 37°C to allow excess dye to leak from the cells. Stained J.RT3-T3.5 cells were then combined with washed CD62L⁻ K562 cells at a 40:60 ratio to model a PBMC mixture. For isolation with separate aptamer and microbeads labeling, 20×10⁶ cells were first stained with 1 to 60 nM biotinylated aptamer in 200 μL binding buffer composed of WB supplemented with 1% BSA and 0.1 mg/mL yeast tRNA for 20 min at 4°C with end-over-end rotation, after which 20 μL Anti-Biotin Microbeads were added and allowed to incubate for an additional 15 min at 4°C with end-over-end rotation. For isolation with aptamer immobilized on microbeads, biotinylated aptamer at the indicated concentration was first incubated with the Anti-Biotin Microbeads for 15 min at 4°C with end-over-end rotation, after which the cells were resuspended in the aptamer-microbead solution and allowed to incubate for 20 min at 4°C with end-over-end rotation. The cells were washed with 1 mL WB supplemented with 0.5% BSA and resuspended in 1000 μL WB supplemented with 0.5% BSA. The suspension was applied to a washed LS column mounted on a QuadroMACS separator (Miltenyi Biotec) according to the manufacturer’s instructions. The flow through containing unlabeled cells was collected, and the column was washed 2 times with 3 mL DPBS supplemented with 0.5% BSA and 1 time with 1 mL DPBS supplemented with 0.5% BSA and 2 mM EDTA to collect all unlabeled cells. For one-step elution, 1 mL of reversal agent solution (10 nM to 1 μM reversal agent in 1 mL DPBS supplemented with 0.5% BSA, 2 mM EDTA and 5 mM MgCl₂, pre-warmed at 37°C) was applied onto the column. 600 μL was allowed to drain, and the column was plugged with an M/F Luer Lock Plug and let incubated for 10 min at room temperature. The remaining volume was drained after incubation, and the column was washed 3 times with 3 mL DPBS supplemented with 0.5% BSA and 5 mM EDTA to collect eluted cells. For two-step elution, 2 mL of pre-warmed reversal agent solution (1 μM reversal agent in 2 mL DPBS supplemented with 0.5% BSA, 2 mM EDTA and 5 mM MgCl₂, pre-warmed at 37°C) was applied onto the column. 600 μL was allowed to drain, and the column was plugged with an M/F Luer Lock Plug and let incubated for 10 min at room temperature. After incubation, 800 μL was allowed to drain and the column was plugged again for a second 10-min incubation with the remaining reversal agent solution. The remaining volume was drained after incubation, and the column was washed 3 times with 3 mL DPBS supplemented with 0.5% BSA and 5 mM EDTA to collect eluted cells. Finally, the washed column was removed from the separator and the column-bound cells were collected by flushing with 5 mL DPBS supplemented with 0.5% BSA and 2 mM EDTA using the provided plunger.

Selection of CD62L⁺CD8⁺ T cells:

Frozen PBMCs were thawed and rested in RPMI 10% FBS at 5 × 10⁶ cells/mL for 15 hours to recover CD62L expression. CD8⁺ T cells were tracelessly selected from 100 × 10⁶ PBMCs as described previously.^[2] Tracelessly selected cells were immediately centrifuged at 500g for 3 min at 4°C to remove the supernatant containing the reversal agent and EDTA. In parallel, CD8⁺ T cells were selected from 100 × 10⁶ PBMCs of the same donor using a CD8⁺ T Cell Isolation Kit (Miltenyi Biotec) according to the manufacturer’s instructions. Aptamer-immobilized MicroBeads were prepared by incubating 10 nM biotinylated CD62LApt.52.s20 aptamers with 10 μL Anti-Biotin Microbeads in 100 μL binding buffer composed of WB supplemented with 1% BSA and 0.1 mg/mL yeast tRNA for 15 min at 4°C with end-over-end rotation. To isolate CD62L⁺CD8⁺ cells, 10 × 10⁶ tracelessly isolated CD8⁺ cells were incubated with 100μL of the aptamer-bead solution for 20 min at 4°C with end-over-end rotation. The suspension was centrifuged at 500g for 3 min at 4°C and the supernatant was removed. The cells were washed with 1 mL WB supplemented with 0.5% BSA, spun down, and resuspended in 500 μL WB supplemented with 0.5% BSA. The suspension was applied to a washed MS column mounted on a MiniMACS separator (Miltenyi Biotec) according to the manufacturer’s instructions. The flow through containing unlabeled cells was collected, and the column was washed 2 times with 500 μL DPBS supplemented with 0.5% BSA and 1 time with 500 μL DPBS supplemented with 2 mM EDTA and 0.5% BSA to collect all unlabeled cells. To elute the CD62L⁺ CD8⁺ cells from the column, 1 mL of reversal agent solution (1 μM reversal agent in 1 mL DPBS supplemented with 2 mM EDTA, 0.5% BSA and 5 mM MgCl₂, pre-warmed at 37°C) was applied onto the column. The column volume (250 μL) was allowed to drain, and the column was plugged with an M/F Luer Lock Plug and let incubated for 10 min at room temperature. After incubation, 400 μL was allowed to drain and the column was plugged again for a second 10-min incubation with the remaining reversal agent solution. The remaining volume (350 μL) was drained after incubation, and the column was washed 3 times with 500 μL DPBS supplemented with 5 mM EDTA and 0.5% BSA to collect eluted cells. The washed column was removed from the separator and the column-bound cells were collected by flushing with 1 mL DPBS supplemented with 2 mM EDTA and 0.5% BSA using the provided plunger. In parallel, 10 × 10⁶ negatively selected CD8⁺ cells were subjected to CD62L selection with CD62L MicroBeads (Miltenyi Biotec) according to the manufacturer’s instructions. The yield, purity and phenotype of the isolated cells were evaluated by flow cytometry after staining for CD62L, CD8, CD3, CD45RA, CD45RO and CCR7 expression. CD62L isolation with > 10e6 cells was performed on an LS column as described in the previous section.

CD19 CAR lentivirus production:

Lentivirus was generated by transient transfection of HEK293T cells with the third-generation lentiviral packaging system using the BioT reagent (Bioland). One day before transfecting the HEK293T cells, the media was replaced with DMEM 10% FBS supplemented with 1 mM sodium pyruvate (Gibco) and 2 nM L-glutamine (Gibco), and again replaced with DMEM 10% FBS just prior to transfection. To prepare the transfection cocktail, 10 μg of total DNA (containing the transfer plasmid, the packaging plasmids and the envelop plasmid) in 10 μL was added to 50 μL DMEM and combined with 440 μL BioT cocktail containing 15 μL BioT and 425 μL DMEM. The transfection cocktail was incubated for 10 min at room temperature and added to HEK293T cells grown on 10-cm TC plates. 24 hours later, the supernatant was discarded, and the media was replaced with 10 mL DMEM 10% FBS. 72 hours after transfection, the media was collected and filtered through a 0.22-μm PES membrane (Corning). Virus media was concentrated by centrifugation at 18,500 rpm for 2 hours at 4°C. The virus pellet was resuspended in DMEM, vortexed for 2 hours at 4°C for homogenization, and stored at −80°C until use.

CAR T cell manufacturing:

CD8⁺ and CD62L⁺CD8⁺ T cells were isolated from fresh PBMCs with aptamers as described previously and the isolated cells were stimulated with Dynabeads^™ Human T-Activator CD3/CD28 (Gibco) at a 1:1 bead to cell ratio in RPMI 10% FBS supplemented with 50 IU/mL IL-2 (Miltenyi Biotec), 5 ng/mL IL-7 (Miltenyi Biotec) and 5 ng/mL IL-15 (Miltenyi Biotec). 2 days later, activated T cells were transduced with the CD19 CAR lentivirus at a multiplicity of infection (MOI) of 3 with 5 μg/mL polybrene (Sigma-Aldrich). Media changes were conducted every 2–3 days to replenish cytokines. Methotrexate was added 2 days post transduction to the transduced cells at 50–100 nM to enrich for CAR-expressing cells. The activator beads were removed 9 days post stimulation.

Antitumor cytotoxicity assay:

CD19⁺ Raji and CD19⁻ K562 target cells were labeled with 0.5 μM CellTrace^™ Far Red dye (Invitrogen) at 1× 10⁶ cells/mL in DPBS with calcium and magnesium for 20 min at room temperature. The remaining dye was inactivated by diluting the cells 1:3 with RPMI 10% FBS. Labeled cells were resuspended in RPMI 10% FBS at 1 × 10⁶ cells/mL and incubated for an additional 30 min at 37°C to allow excess dye to leak from the cells. T cells and labeled tumor cells were combined in a 96-well TC plate at the indicated ratios in 200 μL RPMI 10% FBS. Media with 2% Triton-X 100 was added to the target cells as maximum lysis control. The cells were incubated for 18 hours at 37°C and analyzed by flow cytometry after live/dead staining.

Antitumor cytokine release assay:

5 × 10⁴ cells CD19⁺ Raji and CD19⁻ K562 target cells were co-incubated with 1 × 10⁵ CAR T cells in 200 μL RPMI 10% FBS in a 96-well U-bottom TC plate. After 24 hours, the cells were pelleted and 120 μL of the media was collected and frozen at −80°C until ready for analysis. The media was diluted 1:10 in serum-free RPMI and the cytokine concentrations were quantified using a LEGENDplex^™ kit (BioLegend) according to the manufacturer’s protocol. Briefly, diluted media were incubated with 1x IL-2, IFN-γ and TNF-α capture beads in a 96-well V-bottom plate (BioLegend). The beads were washed, stained with the Human Th Panel Detection Antibodies V02 cocktail (BioLegend) and additionally stained with LEGENDplex^™ SA-PE (BioLegend) for fluorescence detection. The plate was analyzed on a flow cytometer and FCS files were imported into the LEGENDplex^™ Data Analysis Software (BioLegend) to calculate cytokine concentrations. The media was replenished in the original culture plate, and the effector and target cells were allowed to co-incubate for an additional 48 hours. After co-incubation, the cells were stained for PD-1, TIM-3 and LAG-3 expression, and analyzed by flow cytometry.

Statistical analysis:

Data are expressed as mean ± standard deviation unless otherwise stated. The number of technical or biological replicates for each experiment is indicated in the figure captions. A two-tailed t-tests was used for hypothesis testing when comparing two groups. ANOVA was used for hypothesis testing when comparing more than two groups with each other. Paired hypothesis testing was performed when accounting for large donor-to-donor variability. Tukey’s test was used when every mean was compared to every other mean. Sidak correction was used when multiple comparisons were made following two-way ANOVA, assuming that each comparison is independent of the others. Unless otherwise stated, graphing and statistical tests were performed using GraphPad Prism version 10.2.1, GraphPad Software, Boston, Massachusetts USA, www.graphpad.com.