B6.SJL-Ptprc^a (CD45.1; MGI 2164701) (11), C57BL/6-Tg(TcraTcrb)1100Mjb/J (OT-I; MGI 3054907) (12), Tg(UBC-GFP)30Scha (Ubc-GFP; MGI 305178) (13), Tg(Zp3-cre)82Knw (Zp3-Cre; MGI 4888024) (14), Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze} (Ai14; MGI 3809524) (15) and Dock2^tm1Ysfk (DOCK2-GFP; MGI 3772945) (16) mice were bred under specific pathogen-free conditions at the University of Fribourg. To obtain mice expressing tandem-dimer Tomato (tdT) driven by CAG promoter (tdT mice), Ai14 mice were crossed with Zp3-Cre mice to excise the floxed stop codon in the Ai14 alleles. Resultant tdT-expressing mice were backcrossed with wild-type C57BL/6J (B6J) mice to remove Zp3-Cre transgene. GFP-, tdT- or CD45.1-expressing OT-I mice were derived by crossing OT-I mice with Ubc-GFP, tdT or CD45.1 mice, respectively. In some experiments, GFP OT-I mice crossed with CD45.1 OT-I mice were used as GFP CD45.1/1 or GFP CD45.1/2 OT-I mice. Female and male B6J mice at the age of 4–6 weeks old were purchased from Janvier (Le Genest-Saint-Isle, France) and used for screening and as recipient of sex-matched OT-I cells. All animal experimentations have been approved by the Cantonal Committees for Animal Experimentation and performed in accordance with the federal guidelines.

Single cell suspension of spleen and peripheral lymph nodes (LNs) was prepared by mincing them using a 70-µm cell strainer. Untouched CD8⁺ T cells were magnetically isolated using EasySep™ Mouse CD8⁺ T Cell Isolation Kit (STEMCELL Technologies, Grenoble, France) or MojoSort™ Mouse CD8 T Cell Isolation Kit (BioLegend, San Diego, CA). Purity of the isolate CD8⁺ cells constantly exceeded 95%.

For activated CD8⁺ T cells, T-25 or T-75 flasks were coated with 5 µg/mL anti-CD3ϵ antibodies (clone 145-2C11; 100340, BioLegend) in phosphate-buffered saline (PBS) for 16–24 hr at 4°C before seeding CD8⁺ T cells. Immediately after magnetic isolation, CD8⁺ T cells were resuspended in complete medium (RPMI1640 medium supplemented with 10% fetal calf serum (FCS), 2 mM L-glutamine, 0.1 mM non-essential amino acids, 1 mM sodium pyruvate, 100 U/mL penicillin, 0.1 mg/mL streptomycin, 50 µM 2-mercaptoethanol and 10 mM HEPES) and seeded in the antibody-coated flask at 3–5×10⁵ cells/cm² together with soluble 1 µg/mL anti-CD28 antibodies (clone 37.51; 102116, BioLegend). Cells were cultured for 48 hr at 37°C in a humidified 5% CO₂ atmosphere. For naïve CD8⁺ T cells, magnetically isolated T cells were resuspended in the complete medium and were seeded in a 10-cm petri dish or T-75 flask at 1–2×10⁶ cells/cm² and cultured in the presence of 20 ng/mL recombinant mouse interleukin-7 (rmIL-7; 402-ML-020/CF, R&D Systems, Minneapolis, MN) for 24 hr at 37°C in a humidified 5% CO₂ atmosphere. In some experiments, up to 1×10⁷ naïve CD8⁺ T cells were labelled with 5 µM CellTrace™ Violet (C34571, Thermo Fisher) for 20 min at 37°C in 1 mL complete medium before culturing them with rmIL-7 prior to nucleofection.

CD90 crisprRNAs (crRNA) were designed in the previous study (4). DOCK2 crRNAs were designed using DESKGEN online tool (www.deskgen.com; discontinued). Alt-R^® CRISPR-Cas9 crRNA (custom design) and trans-activator RNA (tracrRNA) (1072534) were purchased from Integrated DNA Technologies (Coralville, IA) and reconstituted at 100 µM with nuclease free duplex buffer (Integrated DNA Technologies). Nucleofection was performed as described previously (4) with some modifications. In brief, one microliter each of crRNA and tracrRNA were annealed to form gRNA at 95°C for 5 min using a thermal cycler and cooled to room temperature. Annealed gRNA was mixed with TrueCut Cas9 v2 (A36499, Thermo Fisher Scientific, Basel, Switzerland) at a ratio of gRNA: Cas9 = 1.8 µL: 1.2 µL (equivalent to 90 pmol: 36 pmol) and left at room temperature for > 10 min to generate RNP complex. Pre-stimulated T cells were harvested, counted, and spun down at 80×g for 7 min. Pellets were resuspended in Primary Cell 4D-Nucleofector™ X Kit S (Lonza) buffer solution at a cell concentration of 1–4×10⁶ cells in 20 µL. The entire cell suspension was mixed with 3 µL per complex RNP solution and added to Nucleocuvette™ strip well. Cells were then nucleofected using a 4D-Nucleofector™ with X-Unit (V4XP-4032 and V4XP-9096, Lonza, Basel, Switzerland). Up to three RNP complexes in 9 µL were used per reaction. Immediately after nucleofection, one hundred microliter of pre-warmed complete medium containing 10 ng/mL recombinant mouse IL-2 (rmIL-2; 402-ML-020/CF, R&D Systems) or 20 ng/mL rmIL-7 was added to each Nucleocuvette™ strip well for activated or naïve CD8⁺ T cells, respectively. Then, cells were gently mixed by pipetting and aliquoted into a flat-bottom 96-well plate. Cells were cultured in a total volume of 250 µL complete medium containing rmIL-2 or rmIL-7 at 1×10⁵ and 2×10⁶ cells per well for activated and naïve CD8⁺ T cells, respectively, for 2–10 days at 37°C in a humidified 5% CO₂ atmosphere. crRNAs used in this study are listed in Table 1. Buffer and pulse code are indicated in the corresponding figure legends.,

For activated CD8⁺ T cells, mice subcutaneously infected with 5×10⁴ plaque-forming unit (pfu) herpes simplex virus-1 expressing tdT and ovalbumin (HSV-OVA) (17) via hock (18) received a mix of 1.5×10⁵ each of nucleofected or control in vitro-activated OT-I cells on day 4 post-infection. For naïve CD8⁺ T cells, mice were intravenously injected with a total of 1.5×10⁴ nucleofected OT-I cells comprising 5×10³ each of nucleofected and control OT-I cell populations. One day later, mice were subcutaneously infected with 5×10⁴ pfu HSV-OVA in 20 µL PBS via hock. For secondary infection, mice were kept for 31 days after adoptive transfer of in vitro-activated CD8⁺ T cells and infected intraperitoneally with 10⁵ pfu recombinant lymphocytic choriomeningitis virus expressing ovalbumin (LCMV-OVA) (19) in 100 µL. Each mouse received three OT-I cell populations that were differently nucleofected (or left non-nucleofected) and marked by the expression of GFP, tdT or CD45.1.

Following antibodies were purchased from BioLegend and used for staining: Brilliant Violet 421™ (BV421)-conjugated anti-CD45.1 (clone A20; 110732), CD62L (clone MEL-14; 104436) and anti-TNF-α (clone MP6-XT22; 506328), Brilliant Violet 605™ (BV605)-conjugated anti-CD44 (clone IM7; 103047) and anti-CXCR3 (clone CXCR3-173; 126523), Brilliant Violet 711™ (BV711)-conjugated anti-CD45.1 (clone A20; 110739), CD62L (clone MEL-14; 104445) and anti-IFN-γ (clone XMG1.2; 505836), PE-conjugated streptavidin (405203), PerCP/Cy5.5-conjugated anti-CD43 activation associated glycoform (clone 1B11; 121223) and anti-CD90.2 (clone 53-2.1; 105337), PE/Cy7-conjugated anti-CD45.1 (clone A20; 110729), anti-CD90.2 (clone 53-2.1; 140309) and anti-KLRG1 (clone 2F1/KLRG1; 138415) and anti-CD127 (clone A7R34; 135012), Alexa Fluor™ 700 (AF700)-conjugated anti-CD90.2 (clone 53-2.1; 140323), APC/Fire™ 750-conjugated anti-CD8α (clone 53-6.7; 100766), and Ultra-LEAF™ purified anti-CD16/32 (clone 93; 101330). APC-conjugated anti-IL-2 (clone JES6-5H4; 554429) was purchased from BD Biosciences. PE-conjugated anti-CD49d (clone PS/2; 1520-09) was purchased from SouthernBiotech. Biotinylated anti-CCR7 (clone 4B12; 13-1971-85) was purchased from eBioscience/Thermo Fisher Scientific. Unlabeled anti-CD16/32 (clone 2.4G2; 60161) was purchased from STEMCELL Technologies. Data were acquired using an LSRFortessa (BD Biosciences) and analyzed using FlowJo™ software (FlowJo, Ashland, OR).

For samples from virus-infected mice, popliteal LNs (popLN) and spleen were harvested and minced using a 70-µm cell strainer. Single cell suspension of peripheral blood leukocytes (PBL) was prepared from whole blood by lysing red blood cells. Cells were counted and plated at ≤ 3×10⁶ cells per well in a round-bottom 96 well plate for antibody staining. First, cells were resuspended in 15 µL of 2 µg/mL anti-CD16/32 in staining buffer (PBS containing 2% FCS, 2 mM EDTA and 0.05% (w/v) NaN₃) and incubated for 5 min on ice to block Fc receptors. Fifteen microliters of antibody mix prepared at twice higher concentration than the working concentration was added to cells without washing away anti-CD16/32 and incubated for 30 min on ice (except for CCR7, which was stained at 37°C). After staining, cells were washed twice with staining buffer and resuspended in the staining buffer containing 0.5 µg/mL propidium iodide for dead cell exclusion. For cells derived from virus-infected mice, Zombie Red™ Fixable Viability Dye (423110, BioLegend) was added to anti-CD16/32 solution at a 1: 100 dilution for dead cell exclusion and fixed with 4% paraformaldehyde (15710, Electron Microscopy Sciences, Lucerne, Switzerland) in PBS for 20 min on ice after staining.

For intracellular cytokine staining, cells were restimulated with 1 µM SIINFEKL (BAP-201, EMC Microcollections, Tübingen, Germany) and 5 µg/mL brefeldin A (B6542, Sigma-Aldrich, St. Louis, MO) in 250 µL complete medium for 5 hr at 37°C in a humidified 5% CO₂ atmosphere before staining. After surface staining, dead cell labeling with Zombie Red™ dye and fixation, cells were permeabilized by resuspending in 200 µL PermWash buffer (554723, BD Biosciences, San Jose, CA) and labeled with antibodies against cytokines diluted in PermWash buffer.

DOCK2-GFP CD8⁺ T cells were nucleofected with three CD90 crRNAs with or without two DOCK2 crRNAs (Table 1) on day 0. On day 7 or 10 after nucleofection, cells were harvested and stained with APC/Fire™ 750-conjugated anti-CD8α and PE/Cy7-conjugated anti-CD90.2. After two rounds of wash, cells were resuspended in staining buffer containing 0.5 µg/mL propidium iodide and sorted into CD90^lo and CD90^hi fractions using a FACSAria™ Fusion cell sorter (BD Biosciences). Sorted cells were collected in the complete medium and processed for total RNA isolation immediately after sorting.

Total RNA isolation and cDNA preparation from sorted cells were performed using TRI Reagent™ (12044977, Sigma-Aldrich) and PrimeScript™ RT Reagent Kit (RR037B, Takara Bio, Saint-Germaine-en-Laye, France), respectively, according to the manufacturer’s instruction. cDNA was prepared using 200 ng total RNA and random primers. Quantitative PCR was performed in duplicates with 20 ng template cDNA per reaction using KAPA SYBR^® FAST (KK4603, Roche, Basel, Switzerland) and a CFX96™ Real Time System C1000 Touch™ Thermal Cycler (BioRad, Hercules, CA). Cycling condition was 95°C for 5 min, followed by 40 cycles of 95°C for 15 sec and 60°C for 60 sec. After the completion of PCR reaction, amplification specificity was confirmed by generating dissociation curves. Relative expression values were calculated by ΔΔCT method. First, relative expression levels of Dock2 were calculated as 2^–ΔCt, where ΔCt value represents the difference in Ct values between Dock2 and Gapdh (Ct(Dock2) – Ct(Gapdh)). Then, Dock2 expression levels were normalized to the level in control, DOCK2 non-targeted samples. Following primers were purchased from Microsynth (St. Gallen, Switzerland): Dock2-Fwd, 5’-GGCTCATAGGATTCTCCATCCG-3’; Dock2-Rev, 5’-GATTGGGATGGTGGCTTTCCTG-3’; Gapdh-Fwd, 5’- AGAACATCATCCCTGCATCC-3’; Gapdh-Rev, 5’- TCATCATACTTGGCAGGTTTCTC-3’.

Graph preparation, statistical tests and calculation of Pearson correlation coefficient were performed using GraphPad Prism software (GraphPad Software; San Diego, CA). Thick lines in scatter plots and error bars represent mean and standard deviation, respectively. Differences among two or more groups were considered statistically significant when p < 0.05. Statistical tests used for comparison are indicated in the corresponding figure legends.

First, we examined the optimal nucleofection conditions for in vitro-activated mouse CD8⁺ T cells. Here, we chose CD90 as the model target because of its uniform expression on the surface of CD8⁺ T cells, which allows for its easy detection by flow cytometry. For nucleofection, we selected seven electric pulse codes based on a previous report by Seki and Rutz (4) and recommendation from the manufacturer, Lonza: CA137, CM137, DN100, DN107, DO100, DS138 and DV100. Seki and Rutz used pulse codes indicated in the manufacturer’s instruction for the optimization kit for primary cells. Because this recommendation is not tailored for mouse T cells, we surmised that the previous screening panel might have missed certain pulse codes suitable for mouse CD8⁺ T cells. Therefore, we partly revised the screening panel to include DN107, DO100 and DV100, while retaining only those that performed well in the previous study as other four candidates. We applied no electric pulse as control for the last condition to build our screening panel consisting of 40 conditions with five buffers P1–P5 and eight electric pulse conditions.

We activated polyclonal CD8⁺ T cells isolated from B6J mice in vitro using antibodies against CD3ϵ and CD28 for 48 hr before nucleofection. We then nucleofected activated CD8⁺ T cells with RNP complex containing three different CD90-targeting crRNAs (Table 1) and kept them in culture for additional two days in the presence of rmIL-2 but without anti-CD3ϵ/CD28 (Figure 1A). To determine the optimal condition, we evaluated the frequency of CD90^lo population (knockout efficacy) and cellular yield (calculated by dividing the recovered cell number by the number of cells seeded after nucleofection). Consistent with the notion that activated T cells are amenable to transfection and resultant CRISPR/Cas9-mediated genome editing (4), more than 90% of the harvested CD8⁺ T cells decreased the expression of CD90 protein under many of the nucleofection conditions (Figures 1B, C). In contrast, there was a sizable difference in the cellular yield among the tested nucleofection conditions, ranging from less than 10% to 150% or higher (Figure 1D).

To make an objective decision based on these two readouts, we derived a unified parameter, nucleofection score, from knockout efficacy and cellular yield, defined as ([%CD90^lo] × [%Yield])/100. By definition, a condition that exhibits 100% knockout efficacy and 100% cellular yield gives a nucleofection score of 100, while the one with 0% knockout efficacy gives 0 irrespective of cellular yield. We found that two conditions, P3 buffer with the pulse code CA137 (P3/CA137) and P4/CA137, marked clearly higher mean score than others (Figure 1E). P3/CM137, P3/DN100, P4/CM137 and P5/CA137 yielded comparable nucleofection scores, which were slightly lower than P3/CA137 and P4/CA137. Differences in nucleofection score among these conditions likely stem from cellular yield which is conceivably influenced also by cell death early after nucleofection (Figure 1F). Based on these results, we decided to test the in vivo immune responsiveness of in vitro-activated CD8⁺ T cells nucleofected using P3/CA137, P3/CM137 and P4/CM137. We selected P4/CM137 from the above listed four conditions because this was what Seki and Rutz identified as the optimal condition (4).

To examine in vivo antigen-specific responses of nucleofected in vitro-activated CD8⁺ T cells, we nucleofected OT-I cells as we did for the screening using three CD90 crRNAs (Table 1). After letting the cells rest for 2 days in the presence of rmIL-2 (i.e. 4 days after activation), 150,000 each of the nucleofected cells were adoptively transferred into HSV-OVA-infected hosts (Figure 2A; Supplementary Table 1). Recipients were split into two groups, and each group received two out of four populations of nucleofected OT-I cells (three with CD90 crRNAs and one no RNP control nucleofected using P4/CM137). To gauge the influence of nucleofection itself, all recipients received a third group of 150,000 OT-I cells that were cultured in the same manner as CD90-targeted cells but without nucleofection (denoted as NTC). Thus, each recipient received three populations of OT-I cells expressing GFP, tdT or CD45.1. Recipients were infected on the day when OT-I cells were put in culture for activation (i.e. 4 days before the adoptive transfer).

On days 7 and >40 post-infection, we determined the number of OT-I cells in antigen-draining popLN and spleen. Although there was a weak trend on day 7 that OT-I cells nucleofected using P3 buffer outperformed other groups (Figure 2B), none of the nucleofected OT-I cell populations displayed a statistically significant difference from NTC OT-I cells, except that no RNP control in the spleen was slightly less abundant. After >40 days, the number of nucleofected OT-I cells were less abundant for all groups in both popLN and spleen as compared to NTC OT-I cells, resulting in a subtly more pronounced contraction of nucleofected OT-I cells (Figures 2C, D). Longitudinal analysis of OT-I cells among PBL revealed that a more pronounced contraction of nucleofected cells manifested after day 21 post-infection (Supplementary Figure 1A). Frequency of CD90^lo cells among OT-I cells remained high in all three test groups, while no RNP control OT-I cells retained CD90 expression at the same frequency as NTC OT-I cells (Supplementary Figures 1B–E). Thus, the loss of CD90 did not account for the greater degree of contraction of nucleofected OT-I cells.

To further characterize in vitro-activated, nucleofected OT-I cells, we analyzed their phenotype after adoptive transfer into HSV-OVA-infected host. On day 7 post-infection, almost all OT-I cells were CD44^hi and displayed variable frequencies of CD62L^hi among different anatomical compartments and different conditions (Supplementary Figure 2). Whereas the majority of OT-I cells expressed high levels of CD62L in popLN, only 30–60% were CD62L^hi in the spleen and blood (Supplementary Figures 2B, C). Although there were statistically significant differences among groups, the trend did not correspond to the one that we observed for OT-I cell number. In contrast, most OT-I cells exhibited CD127⁺ KLRG1^– memory precursor effector cell phenotype (Figure 3A; Supplementary Figures 3A, 4A), with CD127^– KLRG1^– early effector cells being the second most abundant population. Such composition of effector OT-I cells in popLN and spleen was common to all five conditions (and hence irrespective of the loss of CD90). Yet, nucleofected OT-I cells contained higher frequency of memory precursor effector cells than NTC OT-I cells did.

CD127⁺ KLRG1^– cells remained the dominant population until > 40 days post-infection, with small fractions of cells being CD127^– KLRG1^– CD127⁺ KLRG1⁺ or CD127^– KLRG1⁺ (Figure 3B; Supplementary Figures 3B, 4B). As a result, all five groups generated memory cells with a largely comparable composition as assessed by the expression of CD127 and KLRG1. In the memory phase, frequencies of CD44^hi and CD62L^hi cells were nearly identical among the groups (Supplementary Figure 2); the minute differences in the frequency of CD44^hi cells are unlikely to be biologically significant. To confirm that memory OT-I cell heterogeneity is indeed comparable, we also evaluated the expression of activation-associated glycoform of CD43 (recognized by the antibody clone 1B11) and CXCR3. The majority of OT-I cells in all five groups comprised CXCR3⁺ CD43^lo cells (Figure 3C; Supplementary Figures 3C, 4C), which had been reported to self-renew at an intermediate rate but exhibit greater clonal expansion upon recall responses (20). There was a weak trend that NTC OT-I cells contained slightly higher frequency of CXCR3⁺ CD43^hi cells, but the difference did not reach a statistically significant level. Even though there were small differences in the phenotype between nucleofected and NTC OT-I cells, none of them explains the lower number of memory cells generated by nucleofected OT-I cells (Figures 2C, D). Furthermore, nucleofected OT-I cells displayed subtle differences in cytokine producing ability on day 7 as compared to NTC OT-I cells (Supplementary Figure 5), whereas there was a slight decrease in the frequency of IFN-γ⁺ cells (but not IFN-γ⁺ TNF-α⁺ cells) >40 days after infection (Supplementary Figure 6).

To test whether the comparable composition and functionality of memory OT-I cells (Figure 3; Supplementary Figures 2-6) leads to a similar recall response capability, we intraperitoneally infected the recipient with recombinant lymphocytic choriomeningitis virus expressing OVA (LCMV-OVA) 35 days after HSV-OVA infection (Figure 4A). We found that, despite slight variability in OT-I cell number in the blood (Figure 4B), all five groups of OT-I cells generated a comparable number of OT-I cells five days after LCMV-OVA infection (Figures 4B, C). The expanded secondary effector cells displayed similar expression pattern of CD127 and KLRG1 (Figure 4D; Supplementary Figure 7), suggesting that nucleofected, in vitro-activated OT-I cells that remained as memory cells are free from the impact of nucleofection observed at earlier time point (Figures 2C, D) and possess recall capabilities comparable to non-nucleofected memory cells.

In sum, nucleofection appeared to impair the fitness of in vitro-activated CD8⁺ T cells for a certain period of time but without reshaping the differentiation spectrum of effector and memory CD8⁺ T cells. Based on these results, we concluded P3/CA137 as the best condition for CRISPR/Cas9-mediated gene inactivation of in vitro-activated primary mouse CD8⁺ T cells because it marked the highest nucleofection score among the tested conditions (Figure 1E). In addition, the observation that inactivation of CD90 led to minimal to no functional alteration as compared to no RNP control confirmed its suitability for use in our screening and functional validation.

Next, we tested the optimal nucleofection conditions for naïve primary mouse CD8⁺ T cells. Similar to the screening for in vitro-activated CD8⁺ T cells, we selected seven electric pulse codes based on the previous publication (4) and inputs from the manufacturer: CM137, DN100, DP100, DS137, DS150, DZ100 and EA100. Of these, DP100, DZ100 and EA100 were not included in the previous study (4). With the no pulse control for each buffer, we evaluated again 40 conditions.

CD8⁺ T cells were nucleofected with three CD90 crRNAs (Table 1) after incubating with rmIL-7 for 24 hr and analyzed for the loss of CD90 protein and cellular yield on day 5 post-nucleofection (Figure 5A). Unlike in vitro-activated CD8⁺ T cells, naïve CD8⁺ T cells lost the expression of CD90 protein to a variable degree, ranging from less than 60% to approximately 90% on average (Figures 5B, C). On top of that, cellular yield was much lower than that of activated CD8⁺ T cells (Figure 5D). To objectively rank all the tested conditions, we used the nucleofection scores again and found that P4/CM137 and P5/CM137 exhibited slightly higher score than other conditions (Figure 5E), owing largely to higher cellular yield. Slightly greater degree of proliferation in P4/CM137 condition than other nucleofected conditions is consistent with the cellular yield (Figures 5F, G). The previously reported optimal condition, P4/DS137, was one of the conditions with the highest knockout efficacy but ranked only fifth in our screening because of the lower cellular yield as compared to P4/CM137 and P5/CM137. P1/CM137 and P2/CM137, the third and fourth position in our result, yielded marginally higher score than P4/DS137 because of higher cellular yield. Yet these two conditions achieved less than 80% knockout efficacy, which is lower than that of P4/DS137 at 88.9%. Of note, the fact that the extent of cell division (Figures 5F, G) does not correlate with knockout efficacy (Figure 5C) indicates that assessment of nucleofection conditions based on knockout efficacy and cellular yield was unlikely to be influenced by facilitated loss of CD90 at a protein level by greater proliferation in certain conditions. Considering that higher knockout efficacy is more beneficial than slightly higher cellular yield for most experiments, we chose P4/CM137, P4/DS137 and P5/CM137 to evaluate the immune responsiveness of nucleofected naïve CD8⁺ T cells.

To test in vivo immune responses of nucleofected naïve CD8⁺ T cells, we nucleofected OT-I cells with three CD90-targeting crRNAs using the above selected three conditions and cultured for 5 days in the presence of rmIL-7 (Figure 6A; Supplementary Table 1). One day before subcutaneous infection with HSV-OVA, we injected 5,000 each of nucleofected OT-I cells into recipient mice. Similar to the assay for in vitro-activated CD8⁺ T cells, we split recipients into two groups and injected two groups of nucleofected OT-I cells (two of P4/CM137, P4/DS137, P5/CM137 or no RNP control nucleofected using P5/CM137) per recipient, together with freshly-isolated OT-I cells as a third group as benchmark. Thus, each recipient received a total of 15,000 OT-I cells, which were identifiable by the expression of GFP, tdT or CD45.1.

The number of nucleofected OT-I cells day 7 post-infection was indistinguishable from that of freshly-isolated OT-I cells in both popLN and spleen (Figure 6B). In all five groups, OT-I cells in popLNs exhibited a diverse effector cell differentiation profile identified by the expression of CD127 and KLRG1 (Figure 6C; Supplementary Figure 8). In contrast, the majority of OT-I cells in the spleen displayed CD127^– KLRG1⁺ short-lived effector cell phenotype, irrespective of the condition (Figure 6D; Supplementary Figure 8). After undergoing massive clonal expansion, OT-I cells nucleofected with CD90-targeting RNPs maintained the frequency of CD90^lo cells similar to the level found in the screening, whereas mock-nucleofected and freshly-isolated OT-I cells retained the high expression of CD90 (Figures 6E–H, 5C). These results indicate that loss of CD90 before priming affected neither clonal expansion nor heterogeneity of effector OT-I cells. Taken together, once CD8⁺ T cells survive nucleofection, they are as competent as freshly-isolated CD8⁺ T cells to elicit effector responses. Although P4/CM137 and P5/CM137 marked slightly higher nucleofection scores (Figure 5E), we concluded P4/DS137 as the best condition for naïve primary mouse CD8⁺ T cells based on its higher knockout efficacy (Figure 5C).

Our optimal nucleofection conditions constantly yielded > 75% knockout efficacy even for naïve CD8⁺ T cells (Figures 5B, C, 6E–H) that were more resistant to gene inactivation than in vitro-activated cells. However, use of such “partial knockout” cells conceivably imposes a hurdle for interpretation of experimental results when the experiment aims to identify functions of the gene of interest. For example, residual wild-type cells can outnumber knockout cells or affect their behavior in vivo. Thus, it is highly useful to devise a tool to enrich successfully targeted cells to a purity of 90% or higher even when the knockout efficacy does not reach such a high level after nucleofection. For targets that are expressed uniformly on the surface of naïve CD8⁺ T cells (e.g. CD90), isolation of successfully targeted cells is readily attainable by simple antibody staining followed by cell sorting. Yet this is not the case for intracellular proteins (e.g. transcription factors) or surface proteins that are completely or partially absent on normal naïve CD8⁺ T cells (e.g. PD-1). Therefore, we tested whether a simultaneous targeting of a “reporter” gene in naïve CD8⁺ T cells enables to identify and enrich successfully targeted cells. To this end, we chose the intracellular protein DOCK2 as the model target because a reporter mouse line that expresses a GFP-fused form of DOCK2 (16) allows for single-cell analysis of its expression levels by flow cytometry. We used CD90 as the reporter because of its uniform expression on naïve CD8⁺ T cells and the lack of functional impact on them (Figure 6; Supplementary Figure 8). To assess the effectiveness of enrichment across a wide range of knockout efficacy, we deliberately varied knockout efficacy of DOCK2 by using different combinations of crRNAs. We kept the knockout efficacy for CD90 low by decreasing the amount of RNP complex to deliver into CD8⁺ T cells to 0.5 μL (instead of 9 μL used in other experiments), such that the frequency of CD90^lo cells always remains lower than that of DOCK2-GFP^lo cells.

We followed changes in the frequency of DOCK2-GFP^lo cells among total, CD90^hi and CD90^lo CD8⁺ T cells up to 10 days post-nucleofection. In most cases, the frequency of DOCK2-GFP^lo and CD90^lo among total CD8⁺ T cells rapidly increased by day 5 and remained relatively stable thereafter. CD90^lo cells were indeed enriched for DOCK2-GFP^lo cells as compared to total CD8⁺ T cells at all timepoints in all experiments (Figures 7A–F). We confirmed lower expression levels of Dock2 mRNA among CD90^lo cells as compared to CD90^hi cells (Figure 7G). To evaluate the effectiveness of simultaneous CD90 targeting as the reporter, we calculated %enrichment as ([%DOCK2-GFP^lo in CD90^lo cells] – [%DOCK2-GFP^lo in total CD8⁺ T cells])/(100 – [%DOCK2-GFP^lo in total CD8⁺ T cells]). This value shows to what extent the use of reporter filled the gap for the observed %DOCK2-GFP^lo among total CD8⁺ T cells to reach 100% knockout efficacy, without being affected by deliberately varied %DOCK2-GFP^lo. The mean %enrichment values on days 5, 7 and 10 were almost identical at approximately 50 (Figure 7H), indicating that enrichment efficacy does not depend on the time after nucleofection. Furthermore, these data demonstrate that the enrichment using CD90 as reporter does not always yield > 90% purity of DOCK2^lo cells. Of note, within the tested ranges of knockout efficacy for DOCK2 and CD90, %enrichment positively correlated with %CD90^lo and %DOCK2-GFP^lo cells among total CD8⁺ T cells (Figures 7I, J), where %DOCK-GFP^lo showed better correlation than %CD90^lo did. Taken together, simultaneous targeting of a reporter gene can provide a means to enrich the cells that underwent successful editing of the primary target gene, albeit to a limited degree.

The relatively stable frequency of CD90^lo and DOCK2^lo cells (Figure 7E) prompted us to ask whether the loss of DOCK2 and CD90 protein after gene inactivation in naïve CD8⁺ T cells would also progress at a comparable rate. To answer this question, we traced changes in the mean fluorescence intensity (MFI) of DOCK2-GFP and CD90 among successfully targeted cells by flow cytometry. To take into account inter-experimental variabilities in the staining, we obtained normalizers for each experiment by dividing the measured MFI value of non-targeted control by the mean values of non-targeted control from six experiments. Then, all the measured MFI values were normalized using the normalizer of corresponding experiment. Interestingly, despite the deliberate variability in the frequency of knockout cells, normalized MFI values turned out to decay at a comparable rate in all the experiments, as shown by the extremely small variability (Figure 8A). We found that the MFI values of both DOCK2-GFP among DOCK2-GFP^lo cells and CD90 among CD90^lo cells kept decreasing until day 10 (Figure 8A). However, levels of two other proteins, CD49d (integrin α4) and CCR7, decayed faster among successfully-targeted cells, reaching the lowest MFI values by day 5 post-nucleofection (Figure 8A).

To estimate the half-life of these proteins after nucleofection, we calculated relative MFI by normalizing the measured values in each experiment to that of the corresponding non-targeted cells. By fitting the data to a one phase decay model, we obtained the half-life of 2.57, 1.53, 1.27 and 1.28 days for DOCK2, CD90, CD49d and CCR7, respectively (Figure 8B), showing that the decay rate of DOCK2 is slower than that of other three targets. In contrast, changes in %DOCK2-GFP^lo and %CD90 from day 5 onwards appeared comparable and smaller than those between day 0 and 5 (Figure 7E). Thus, we hypothesized that gene inactivation in successfully nucleofected cells was largely completed within 5 days after nucleofection for both targets. To test this, we assumed the frequency of knockout cells on day 10 as plateau of each experiment and normalized the frequency of knockout cells on days 5 and 7 to that on day 10 (Figure 8C). After curve fitting, we obtained the half-time to reach the putative plateau of 1.48, 1.80, 1.98 and 0.805 days for DOCK2, CD90, CD49d and CCR7, respectively. These results indicate that the loss of protein proceeds at a disparate rate for different proteins. Considering that there was no evident link between the decay of target protein and appearance of successfully targeted cells, turnover rate and abundance of target proteins are likely to be the determinant of their decay rate after gene inactivation. Moreover, extremely small variability in the normalized MFI values (Figure 8A) also implies that degradation of protein (and mRNA) encoded by the target gene occurs at a rate that is dependent on its biochemical properties of the target but independent of the knockout efficacy.