Skip to main content
NCBI home page
The EMBO Journal logoLink to The EMBO Journal
. 2021 Apr 15;40(11):e106658. doi: 10.15252/embj.2020106658

Functional heterogeneity of cytotoxic T cells and tumor resistance to cytotoxic hits limit anti‐tumor activity in vivo

Roxana Khazen 1, Marine Cazaux 1,2, Fabrice Lemaître 1, Beatrice Corre 1, Zacarias Garcia 1, Philippe Bousso 1,✉
PMCID: PMC8167356  PMID: 33855732

Abstract

Cytotoxic T cells (CTLs) can eliminate tumor cells through the delivery of lethal hits, but the actual efficiency of this process in the tumor microenvironment is unclear. Here, we visualized the capacity of single CTLs to attack tumor cells in vitro and in vivo using genetically encoded reporters that monitor cell damage and apoptosis. Using two distinct malignant B‐cell lines, we found that the majority of cytotoxic hits delivered by CTLs in vitro were sublethal despite proper immunological synapse formation, and associated with reversible calcium elevation and membrane damage in the targets. Through intravital imaging in the bone marrow, we established that the majority of CTL interactions with lymphoma B cells were either unproductive or sublethal. Functional heterogeneity of CTLs contributed to diverse outcomes during CTL–tumor contacts in vivo. In the therapeutic settings of anti‐CD19 CAR T cells, the majority of CAR T cell–tumor interactions were also not associated with lethal hit delivery. Thus, differences in CTL lytic potential together with tumor cell resistance to cytotoxic hits represent two important bottlenecks for anti‐tumor responses in vivo.

Keywords: CAR T cells, CTL, intravital imaging, lethal hit, sublethal hit

Subject Categories: Cancer, Immunology

Intravital imaging of T cell‐tumor contacts reveals variability in the ability of cytotoxic T cells to kill tumor cells and that the majority of interactions are non‐lethal.

graphic file with name EMBJ-40-e106658-g006.jpg

Introduction

Cytotoxic T cells (CTLs) have the ability to kill infected and tumor cells. Upon CTL conjugation with cognate target cells, CTLs release perforin and granzymes to form pores in the plasma membrane and initiate cell death, respectively. The molecular and cellular dynamics of CTL‐mediated killing have been studied extensively in vitro (Dustin & Long, 2010; Dieckmann et al, 2016; Basu & Huse, 2017). These studies revealed in particular that CTL killing is a highly sensitive and rapid process (Wiedemann et al, 2006; Bertrand et al, 2013). Moreover, when CTL activity in culture was analyzed at the single cell level, extensive heterogeneity in CTL killing performance was recorded with a fraction of CTLs endowed with the ability to kill multiple targets (Varadarajan et al, 2011; Vasconcelos et al, 2015). These studies also highlighted that cytotoxic effectors such as NK cells or T cells can increase their killing abilities over time, exhibiting burst killing behaviors (Choi & Mitchison, 2013; Vasconcelos et al, 2015). In the context of cancer, another layer of complexity comes from the fact that tumor cells often exhibit resistance to cell death and reduced sensitivity to granzymes (Kashkar et al, 2006; Lickliter et al, 2007; Mohammad et al, 2015). Beyond the downregulation of antigen presentation, tumors may also exhibit specific mechanisms to resist CTL‐mediated cytotoxic hits (Medema et al, 2001; Khazen et al, 2016). Melanoma cells for instance can use lysosome secretion at the immunological synapse to counteract perforin activity (Khazen et al, 2016).

While these in vitro studies have been instrumental to delineate key features of CTL biology, it remains unclear how efficiently CTLs exert their cytolytic function in the microenvironment (TME) of a developing tumor. A few studies have relied on intravital imaging as a mean to monitor CTL killing activity at the single cell level in different contexts. In a model of solid tumor, the average killing frequency per CTL was one every 6 h (Breart et al, 2008). CTL delivery of lethal hits to B‐cell targets in lymph nodes was found to be largely suppressed by regulatory T cells (Mempel et al, 2006). In the context of CAR T cells, we recently reported extensive heterogeneity in CAR T‐cell signaling upon target encounter and subsequent cytotoxic activity (Cazaux et al, 2019). Finally, in a model of viral infection, CTL killing was only efficient when multiple T cells were collectively hitting the target cell (Halle et al, 2016). Together, these studies have indicated that CTL‐mediated killing may not be a very efficient process in vivo. Suboptimal CTL activity may arise from the inability to recognize tumor targets, failure to deliver cytotoxic hits, or lack of target cell death despite the delivery of cytotoxic cargo. Understanding the outcomes of individual CTL–tumor cell interactions in vivo is therefore essential to obtain a comprehensive picture of intratumoral CTL activity.

Here, we have used a combination of fluorescent reporters to track target cell damage and induction of apoptosis during CTL activity against a B‐cell lymphoma. We report that most CTL–tumor cell contacts are either unproductive or induce sublethal damages and that only a minority of interactions results in tumor cell killing. Furthermore, we provide evidence that individual CTLs exhibit heterogeneous behavior and lytic activities in situ. Our results suggest that suboptimal T‐cell cytotoxic activity and tumor resistance to lethal hits represent important bottlenecks for tumor elimination.

Results

Diverse outcomes for tumor cells in response to CTL attack

To quantify the efficiency of CTL‐mediated killing, we relied on OT‐I T cells and a Myc‐driven B‐cell lymphoma tumor cell line expressing the ovalbumin (OVA) model antigen. We examined CTL activity in a classic in vitro cytotoxic assay performed in a U‐bottom plate. Tumor killing was very efficient in these settings even at low E:T ratio (2:1) with 65% of targets eliminated within 4 h (Fig 1A). This result confirmed that OVA‐expressing B lymphoma cells are sensitive to CTL killing. Of note, tumor cell elimination by CTLs significantly dropped when the cytotoxic assay was performed in a flat‐bottom plate at the same E:T ratio (Fig 1A). The perforin/granzyme pathway was likely dominant at least in the time frame of the assay as killing was largely abolished by calcium chelation (that blocks granule exocytosis) (Appendix Fig S1A). Since we cannot formally exclude that calcium chelation affects other pathways, we also analyzed granzyme B activity within targets cells loaded with a specific substrate. As shown in Appendix Fig S1B and C, we rapidly detected granzyme B activity within tumors cells cocultured with CTLs. The lower efficiency of CTL killing in flat‐bottom plates may reflect the fact that not all target cells were in contact with a CTL during the assay. It is also possible that most CTLs were efficiently paired to tumor cells but that a minority of contacts triggered cell death. To test the latter possibility, we imaged CTL–tumor cell conjugates with the aim to visualize both target cell damage and induction of apoptosis. Pore formation during delivery of cytotoxic hit has been associated with entry of extracellular calcium (and therefore an increase in intracellular calcium concentration) and of other small molecules present in the extracellular milieu (including propidium iodide (PI)) (Fig 1B; Poenie et al, 1987; Keefe et al, 2005; Lopez et al, 2013). We first labeled CTLs and tumor cells with the calcium sensitive dye Indo‐1 and added PI in the culture. As expected, CTLs exhibited a calcium signal upon conjugation with the target cells, indicative of effective antigenic recognition (Fig 1C and D, Movie EV1). In 54% (87/161) of the conjugates, this was followed by an elevation of intracellular calcium concentration in the target cell (Fig 1C and D, Movie EV1). In most of these conjugates, calcium elevation was concomitant with a progressive entry of PI in the target cell, suggesting that membrane damage (induced by perforin‐pore formation) was responsible for calcium entry (Fig 1C and D, Movie EV1). Two outcomes were then recorded. Target cell death (corresponding to lethal hits) as reflected by bright PI staining and persistent calcium elevation was detected in less than 20% of conjugates with membrane damage (Fig 1C and D, Movie EV1). The remaining target cells appeared to survive the attack (sublethal hit) with intracellular calcium returning to baseline levels and no further PI uptake being detected (Fig 1C and D, Movie EV1). This was most likely due to the existence of repair mechanisms that help restore membrane integrity after perforin pore formation (Keefe et al, 2005; Lopez et al, 2013; Khazen et al, 2016). Finally, in approximately half of the conjugates (46%), we did not detect evidence for membrane damage (no hits: no calcium elevation nor PI uptake in the target), despite the fact that all CTLs exhibited potent calcium elevation upon target cell conjugation (Fig 1C and D, Movie EV1). Overall, as summarized in Fig 1E, diverse tumor cell fates were observed following conjugate with a CTL. Moreover, in a different B‐cell lymphoma model (pro‐B cells driven by v‐Abl expression (Lenden Hasse et al, 2017), we again observed that the majority of CTLs contacts did not lead to lethal hit delivery (Fig 1E) indicating that sublethal hits are not restricted to Myc‐driven B‐cell lymphomas.

Figure 1. Visualizing the various tumor cell fates in response to CTL attack in vitro .

Figure 1

Open in a new tab
  • A
    OVA‐expressing lymphoma B cells are killed by OT‐I CTL in a bulk cytotoxic assay. OT‐I CTLs were cultured in the presence of OVA+ Eµ‐myc B cells in either U‐bottom (white) or flat‐bottom (gray) plates. Tumor cell killing was evaluated by flow cytometry at the indicated time points. Data were pooled from three independent experiments. (***P < 0.001, unpaired t‐test).
  • B
    Scheme illustrating that perforin‐mediated pore formation promotes the entry of calcium and propidium iodide (PI) in the cytosol of target cells.
  • C, D
    Diverse outcome for tumor cells following interactions with CTLs. OT‐I CTLs and OVA+ Eµ‐myc B cells were labeled with the calcium‐sensitive dye Indo‐1, cultured together in the presence of propidium iodide (PI) and subjected to live imaging. Representative time‐lapse images (C) and corresponding fluorescence quantification (D) showing the three distinct outcomes for CTL–tumor cell contacts (left, calcium response in T cells, and right, calcium and PI signal in targets): no detectable damage (no calcium signal in the target cells), reversible damage (transient calcium elevation and low uptake of PI), and irreversible damage (prolonged calcium elevation and bright PI staining). CTLs are outlined by white dashed lines. White arrowheads indicate tumor cells. Red arrowheads indicate tumor cells with internalized PI. Scale bar, 5 µm.
  • E
    Pie chart representing the various fates of tumor cells following conjugation with a CTL. Data were pooled from three independent experiments with 3 distinct image acquisitions in each experiment.

A minority of CTL–tumor cell interactions results in target cell apoptosis in vitro

To directly monitor the induction of apoptosis in target cells in conjunction with calcium signals, we used B‐cell tumors expressing a genetically encoded caspase 3 reporter (Cazaux et al, 2019) and labeled these cells with Indo‐1. We found that a majority of tumor cells paired with a CTL exhibited one or several calcium fluxes, while calcium fluxes were much less frequent and shorter in duration in isolated tumors (Fig 2A–D).

Figure 2. A minority of CTL–tumor cell interactions results in target cell apoptosis in vitro .

Figure 2

Open in a new tab
OVA+Eµ‐myc B cells expressing a FRET‐based reporter for caspase 3 activity were labeled with the calcium‐sensitive dye Indo‐1, incubated with OT‐I CTLs and subjected to live imaging.
  • A, B
    Calcium responses were recorded over time in (A) unconjugated tumor cells or in (B) tumor cells forming a conjugate with a CTL (during the duration of the contacts). Each line corresponds to an individual tumor cell and each square to a time point. Squares are colored in magenta upon calcium elevation or in blue otherwise. Apoptotic events (detected by caspase 3 activity) are indicated by an overlaid green dot. Data are representative of three independent experiments.
  • C
    The number of calcium fluxes in individual tumor cells is plotted for unconjugated tumors or conjugated tumors. Values were normalized to 100 min. Each dot represents one tumor cell. Bars represent mean values.
  • D
    The cumulated period of calcium signals was recorded for individual unconjugated or conjugated target cells and expressed as percentage of total time. Each dot represents one tumor cell. Bars represent mean values.
  • E, F
    Distinct tumor cell outcomes in response to CTL attack. Three categories were defined: no hit (no calcium response in the target cell), sublethal hit (transient calcium response and no caspase 3 activity), lethal hit (prolonged calcium response and induction of caspase 3 activity). (E) Pie chart representing the percentage of each category. (F) Time‐lapse images (left) of a representative CTL–tumor cell conjugate in each of the three categories are shown together with the corresponding quantification (right) of calcium response and caspase 3 activity over time. T cells are delineated by white dashed lines. Scale bar, 10 µm. Data were pooled from three independent experiments.

Data information: **P < 0.01, ***P < 0.001, unpaired t‐test.

Distinct outcomes were observed for CTL–tumor cell conjugates (Fig 2E and F, Movie EV2). Calcium elevation in target cells was most often transient and not associated with apoptosis induction. These results were consistent with the delivery of sublethal hits inducing pore formation but rapid repair. As shown in Fig 2E and F and Movie EV2, only a minority of CTL‐target pairs (15%) displayed evidence of cell damage (calcium flux) followed by apoptosis induction (lethal hits).

Killing events were associated with longer calcium increases (>12 min) while calcium elevation rarely exceeds 5 min in sublethal responses suggesting that sustained calcium elevation could be used as a proxy for target killing (Figs 1D and 2B and F).

In sum, the study of stable CTL–tumor cell conjugates strongly suggests that despite effective target cell recognition, CTLs often fail to trigger a lethal response in tumor targets, inducing instead either no detectable effects or a sublethal response.

Stable immunological synapse formation between CTLs and tumor cells does not predict target cell fate

It was possible that the quality of the immunological synapse formed between the CTL and the target cell dictated the efficacy of the killing process. We therefore compared the characteristics of CTL–tumor conjugates for interactions associated with no, sublethal, or lethal hits. We used LifeAct‐GFP‐expressing CTLs (to monitor actin accumulation) and the genetically encoded Twitch‐2B calcium indicator (to assess target cell responses to CTL attacks). One hallmark of conjugate formation is the rounding of CTLs (Donnadieu et al, 1994; Negulescu et al, 1996). As shown in Fig 3A–D, CTL rounding was similar for T cells triggering no, sublethal, or lethal hits. We also quantified actin accumulation at the CTL–tumor cell interface. Robust actin accumulation was detected in all conjugates, irrespective of the final tumor cell fate (Fig 3C and E). These results suggest that the low rate of successful killing by CTLs when analyzed at the single cell level is not the consequence of defective synapse formation.

Figure 3. The formation of the immunological synapse does not predict target outcome during CTL–tumor cell contacts.

Figure 3

Open in a new tab
LifeAct‐GFP OT‐I CTLs were cultured with B lymphoma cells expressing the OVA model antigen and the genetically encoded calcium indicator Twich2B.
  1. Representative time‐lapse image illustrating CTL rounding upon target engagement. Scale bar, 5 µm.
  2. Quantification of CTL shape (roundness coefficient) in unconjugated and conjugated CTLs. A circular shape corresponds to an index of 1. Each dot represents one CTL. Bars represent mean values.
  3. Representative time‐lapse images showing examples of CTL‐tumor conjugates resulting in either no detectable response, transient calcium response or sustained calcium response in the target cell. Arrows show actin accumulation at the immunological synapse. Scale bar, 5 µm.
  4. Quantification of CTL roundness for contacts in the three indicated categories. Each dot represents one CTL. Red bars represent mean values.
  5. Quantification of actin accumulation at the immunological synapse for CTLs engaged in contacts with the indicated outcome. Each dot represents one CTL. Red bars represent mean values.

Data information: Data were pooled from three independent experiments. (***P < 0.001, ns, not significant, unpaired t‐test).

Intratumoral CTL activity is dominated by the delivery of sublethal hits

Having established the diversity of tumor cell fates following CTL conjugation in vitro, we sought to quantify the outcome of CTL attacks at the single cell level in vivo. For this, we used mice with established B‐cell lymphoma expressing the OVA antigen and the Twich2B genetically encoded calcium reporter. As reported previously, bone marrow was the primary site of tumor development (Cazaux et al, 2019). Activated OT‐I CD8+ T cells were transferred as a model of adoptive T‐cell therapy, and intravital imaging was performed 2 days later. We examined the response of individual tumor cells during their interactions with CTLs in comparison with tumors developing in the absence of T cells. Consistent with our in vitro data, calcium fluxes were largely increased (both in duration and numbers) in tumor cells contacted by CTLs, indicating that a substantial fraction of tumor cells are responding to CTL attacks (Fig 4A–E, Appendix Fig S2A, Movie EV3). However, most of these calcium rises were transient, suggestive of sublethal interactions (Fig 4B, Movie EV3). A minority of contact resulted in prolonged calcium elevation in tumor cells, an observation compatible with irreversible membrane damage (Fig 4B). Although killing could not be directly ascertained in these experiments, our measurements of prolonged calcium responses indicated that, at most, 6% of the CTL–tumor contacts were lethal (Fig 4B and F). In sum, our single‐cell imaging of CTL activity in vivo suggested that tumor cells most often survive individual CTL attacks, a feature that most likely limits the efficacy of anti‐tumor T‐cell responses.

Figure 4. Delivery of sublethal hits dominates CTL activity in vivo .

Figure 4

Open in a new tab
Mice with established B‐cell lymphoma expressing the Twitch‐2B calcium indicator were adoptively transferred with GFP‐expressing OT‐I CTLs. Intravital imaging of the bone marrow was performed on day 2 post‐T‐cell transfer.
  1. Representative time‐lapse images showing tumors without CTLs (upper panels) or with OT‐I CTLs (lower panels). Tumor cells with low and high intracellular calcium appear in blue and red, respectively. CTLs appear in white. Red arrows indicate tumor cells with elevated calcium. Scale bar, 5 µm.
  2. Calcium responses in individual tumors cells were recorded in (upper table) control tumors (developing without T cells) or (lower table) after T‐cell transfer in tumor cells contacted by at least one CTL during the imaging period. Data are representative of three independent experiment.
  3. Calcium responses of a representative tumor cell in control tumors (blue) or of a tumor cell contacted by a CTL in adoptively transferred mice (red).
  4. Quantification of the number of distinct calcium fluxes normalized to 100 min in individual tumor cells. Each dot represents one tumor cell. Red bars represent mean values.
  5. Quantification of the cumulated period of calcium signals in individual tumor cells expressed as the percentage of total time. Each dot represents one tumor cell. Bars represent mean values.
  6. Pie chart showing the outcomes of CTL‐tumor contacts. Contacts were classified into three categories: no calcium elevation in the target (no hit), transient calcium elevation (sublethal hit), prolonged calcium response in the target (putative lethal hit).

Data information: Data were pooled of four independent experiments. (*P < 0.05, ***P < 0.001, unpaired t‐test).

CTL functional heterogeneity during interactions with tumor cells in vivo

In addition to potential heterogeneity within tumor cells, it is possible that not all CTLs exhibited the same capacity of hit delivery at the time of imaging. To test this possibility, we tracked individual CTLs for 2–4 h and analyzed the outcome of all their interactions with tumor cells. First, we examined calcium responses elicited in all target cells contacted by individual CTLs. As illustrated in Fig 5A and B, extensive differences were observed in the calcium responses elicited by distinct CTLs. Some CTLs induced no or little calcium elevation in the contacted targets, while other efficiently triggered a calcium response in the majority of contacted targets. Second, we imaged apoptosis induction in tumor cells contacted by individual CTLs in vivo, using B lymphoma cells expressing the FRET‐based reporter for caspase 3 activity. Extensive heterogeneity in killing efficiency was apparent in these experiments, with some T cells killing rapidly and successfully multiple targets, while others failed to kill any tumor cells over multiple hours (Fig 5C–E and Movie EV4 and EV5). These observations likely represent in vivo evidence for the previous in vitro description of super‐killer cells within the CTL pool (Vasconcelos et al, 2015). Only 20% of CTLs were capable of inducing tumor cell death during the imaging period, and killing events represented 5% of all total contacts established (Fig 5C–E). Taken together, our results suggest that functional heterogeneity in CTLs contributes to the diverse outcomes during CTL–tumor contacts in vivo.

Figure 5. Extensive functional heterogeneity within tumor‐infiltrating CTLs in vivo .

Figure 5

Open in a new tab
  • A, B
    CTLs exhibit distinct capacities to trigger calcium responses in contacted tumor cells. Mice with established B‐cell lymphoma expressing OVA and the Twitch‐2B calcium indicator were adoptively transferred with GFP‐expressing OT‐I CTLs. Intravital imaging of the bone marrow was performed two days after T‐cell transfer. (A) Left. Two‐photon time‐lapse image and corresponding scheme showing representative CTLs triggering either limited (upper images) or numerous (lower images) calcium signals in 5 contacted targets. Red arrows mark tumor cells with elevated calcium. Scale bar, 5 µm. Right. The periods of contact for the indicated targets are shown as white boxes, and periods of calcium elevation are shown in magenta. (B) Total number of distinct calcium fluxes triggered by individual CTLs in all contacted tumor cells during an imaging period of one hour. For a rigorous comparison, only CTLs which activity could be monitored for at least 1 h were included in the analysis. Data are representative of 4 independent experiments.
  • C–E
    CTLs exhibit extensive differences in lytic activity in vivo. Mice with established B‐cell lymphoma expressing OVA and the caspase 3 reporter were adoptively transferred with LifeAct‐GFP‐expressing OT‐I CTLs. Intravital imaging of the bone marrow was performed two days after T‐cell transfer. The frequency of CTLs engaged in lytic activity (induction of apoptosis in at least one contacted tumor cell) is shown (C) together with the frequency of lytic contacts within all CTL–tumor cell contacts recorded (D). (E) Representative two‐photon time‐lapse images showing a representative CTL with high lytic capacity (killing 4 targets) and a CTL with no lytic activity for up to 4 h. Blue arrows mark apoptotic tumor cells. Live and apoptotic tumors appear in gray and blue, respectively. CTLs appear in green. Scale bar, 10 µm. Representative of 3 independent experiments.

A minority of interactions established by CAR T cells lead to lethal hit delivery

To extend our findings to a therapeutic setting, we analyzed the tumor cell fate upon contact with anti‐CD19 CAR T cells. As observed with CTLs, CAR T cells triggered calcium elevation in target cells (Fig 6A–D, Movie EV6). Although tumor killing by CAR T cells was more efficient than by conventional CTLs (compared Figs 2 and 6), still in most cases, calcium signals in tumors were reversible, suggesting frequent delivery of sublethal hit by CAR T cells (Fig 6E, Movie EV6). Next, we examined CAR T‐cell interaction with B‐cell tumors in vivo by intravital imaging. CAR T cells increased the frequency of calcium responses in contacted targets as compared to non‐contacted targets in the same region or tumors without CAR T‐cell transfer (Fig 7A–D, Appendix Fig S2B). However, only a small percentage of contacted targets (10%) exhibited a prolonged calcium response compatible with a lethal hit delivery (Fig 7E, Movie EV7). Functional heterogeneity in CAR T cells contributed to the diversity of contact outcome as some CAR T cells induced calcium responses in multiple contacted targets while others were seen interacting extensively with tumors eliciting little or no detectable response (Fig 7F–G). Thus, the outcome of CAR T‐cell interactions with tumors appears highly diverse with only a fraction of contacts leading to lethal hit delivery.

Figure 6. A minority of CAR T cell–tumor interactions results in target cell apoptosis in vitro .

Figure 6

Open in a new tab
Eµ‐myc B cells expressing a FRET‐based reporter for caspase 3 activity were labeled with the calcium‐sensitive dye Indo‐1, incubated with GFP‐expressing anti‐CD19 CAR T cells and subjected to live imaging.
  • A, B
    Calcium responses were recorded over time in (A) unconjugated tumor cells or in (B) tumor cells forming a conjugate with a CAR T cell. Each line corresponds to an individual tumor cell and each square to a time point. Squares are colored in magenta upon calcium elevation or in blue otherwise. Apoptotic events (detected by caspase 3 activity) are indicated by an overlaid green dot. Data are representative of three independent experiment.
  • C
    The number of calcium fluxes in individual tumor cells is plotted for unconjugated tumors or conjugated tumors. Values were normalized to 100 min. Each dot represents one tumor cell. Bars represent mean values. **P < 0.01, unpaired t‐test.
  • D
    The cumulated period of calcium signals was recorded for individual unconjugated or conjugated target cells and expressed as percentage of total time. Each dot represents one tumor cell. Bars represent mean values. **P < 0.01, unpaired t‐test.
  • E
    Distinct tumor cell outcomes in response to CAR T‐cell attack. Three categories were defined as previously described: no hit, sublethal hit, lethal hit. Pie chart representing the percentage of each category. Data were pooled from three independent experiments with 3 distinct image acquisitions in each experiment.

Figure 7. Heterogeneity in CAR T cell–tumor cells outcome in vivo .

Figure 7

Open in a new tab

Mice with established B‐cell lymphoma expressing the Twitch‐2B calcium indicator were adoptively transferred with GFP‐expressing anti‐CD19 CAR T cells. Intravital imaging of the bone marrow was performed on day 2 post CAR T‐cell transfer.

  • A
    Representative time‐lapse images showing tumors without CAR T cells (upper panels) or with CAR T Cells (lower panels). Tumor cells with low and high intracellular calcium appear in blue and red, respectively. CAR T cells appear in light green. Red arrows indicate tumor cells with elevated calcium. Scale bar, 5 µm.
  • B
    Calcium responses in individual tumors cells were recorded in (upper table) control tumors (developing without CAR T cells) or (lower table) after CAR T‐cell transfer in tumor cells contacted by at least one CAR T cell during the imaging period. Representative of 3 independent experiments.
  • C
    Quantification of the number of distinct calcium fluxes normalized to 100 min in individual tumor cells. Each dot represents one tumor cell. Red bars represent mean values. *P < 0.05, unpaired t‐test.
  • D
    Quantification of the cumulated period of calcium signals in individual tumor cells expressed as the percentage of total time. Each dot represents one tumor cell. Bars represent mean values. ***P < 0.001, unpaired t‐test.
  • E
    Pie chart showing the outcomes of CAR T cell‐tumor contacts. Contacts were classified into three categories: no calcium elevation in the target (no hit), transient calcium elevation (sublethal hit), prolonged calcium response in the target (putative lethal hit). Data were pooled from three independent experiments.
  • F,G
    CAR T cells exhibit distinct capacities to trigger calcium responses in contacted tumor cells. (F) Left. Two‐photon time‐lapse image and corresponding scheme showing representative CAR T‐cell triggering either limited (upper images) or numerous (lower images) calcium signals in 4 contacted targets. Red arrows mark tumor cells with elevated calcium. Scale bar, 10 µm. Right. The periods of contact for the indicated targets are shown as white boxes and periods of calcium elevation are shown in magenta. (G) Total number of distinct calcium fluxes triggered by individual CAR T cells in all contacted tumor cells during an imaging period of one hour is plotted. For a rigorous comparison, only CAR T cells which activity could be monitored for at least 1 h were included in the analysis. Data are representative of three independent experiments.

Discussion

By means of intravital imaging and functional reporters in tumor cells, we have established here that despite effective antigen recognition, intratumoral CTLs and CAR T cells often fail to deliver lethal hits. Overall, less than 10% of CTL–tumor cell contacts resulted in tumor cell killing. The large majority of CTL–tumor cell contacts appeared to induce reversible damage or no detectable damage at all.

Our results extend findings by Halle et al that have observed low killing rate and sublethal interactions in the context of virally infected cells in lymph node (Halle et al, 2016). One notable difference with the viral infection model in which killing occurred during swarming of multiple T cells around the target is that the CTLs density within the TME was found to be lower, limiting CTL cooperativity. Here and in previous tumor imaging studies (Breart et al, 2008; Cazaux et al, 2019), tumor killing (although unfrequent) was most often the result of a monogamous interaction. Virus and tumors may also have both shared and unique mechanisms for suppressing CTL activity. In particular, several mechanisms could concur to the surprisingly low rate of intratumoral CTL‐ or CAR T‐cell‐mediated killing when this phenomenon was analyzed at the single cell level. Dampening of T‐cell responsiveness in the TME, the presence of regulatory T cells and functional exhaustion may decrease the ability of CTL to arrest on their target and degranulate (Mempel et al, 2006; Boldajipour et al, 2016; Michonneau et al, 2016). Additionally, tumor cells can develop mechanisms to resist cytotoxic hits and rapidly repair membrane damages (Medema et al, 2001; Lopez et al, 2013; Khazen et al, 2016). Consistently, we provide evidence for frequent delivery of sublethal hits in the TME. Our result implies that a high density of CTLs is most likely required for tumor regression, in order to increase the probability of killing (Deguine & Bousso, 2013; Halle et al, 2016; Hamieh et al, 2019; Beck et al, 2020). This has been indeed predicted by modeling approach (Budhu et al, 2010), and consistently, we had previously observed in two distinct models that tumor regression was associated with the progressive accumulation of cytotoxic effectors (Breart et al, 2008; Cazaux et al, 2019). In addition, our present finding that only a fraction of CTLs exhibit strong cytotoxic potential at a given time point also suggests that functional heterogeneity in CTLs represents an important bottleneck in tumor eradication. Previous in vitro studies have indicated that not all CTLs are equally efficient at killing cognate targets (Snyder et al, 2003; Vasconcelos et al, 2015) and it is possible that this heterogeneity is further increased by stochastic events in vivo such as CTL history of previous cellular encounters. It is important to note that in addition to CTLs, tumor cells present in a given individual can also exhibit extensive heterogeneity (Marusyk et al, 2012; Milo et al, 2018), further contributing to the diversity of outcomes during CTL–tumor contacts.

One limitation of the study is that assessment of target killing and tumor damage was performed during and immediately after CTLs contacts. This is due in part to the technical difficulty to image individual tumors for extended periods in vivo. While calcium signals in target cells have been largely attributed to the perforin/granzyme pathway (Halle et al, 2016), it remains theoretically possible that distinct types of cell death are induced at later time points. The generation of genetically encoded reporter for other cell death pathways may help clarify the relative contribution of caspase 3‐independent cell death pathways in tumor regression.

As illustrated here, the ability to measure the consequence of single CTL–tumor cell interactions should provide a framework to evaluate strategies aimed at boosting lethal hit delivery by a maximal number of tumor‐infiltrating T cells.

Materials and Methods

Mice and cell lines

Six‐ to eigth‐week‐old C57BL/6J mice were purchased from Charles River and ENVIGO. Rag1 −/− LifeAct‐GFP OT‐I TCR or Ubi‐GFP Rag1 −/− OT‐I TCR, Eμ‐myc, and Rag2 −/− mice were bred in our animal facility under specific pathogen‐free conditions. All animal studies were approved by the Institut Pasteur Safety Committee in accordance with French and European guidelines (CETEA 2017‐0038).

A lymphoma B‐cell line was isolated from a male Eμ‐myc transgenic mouse (Harris et al, 1988), which develops spontaneous Burkitt‐like B‐cell lymphomas. Immortalized pro‐B cells were generated by infecting bone marrow cells with retro virus encoding for viral‐Abelson kinase (v‐abl) (Lenden Hasse et al, 2017). Both cell lines were retrovirally transduced to express OVA antigen, and Eµ‐myc cells were additionally transduced to express either the FRET‐based Twitch‐2B calcium indicator (Eµ‐myc‐Twitch‐2B) (Thestrup et al, 2014) or a FRET‐based reporter for caspase‐3 activity (Eµ‐myc‐DEVD) (Cazaux et al, 2019). Cells were cultured in complete RPMI medium and were routinely tested for the absence of Mycoplasma contamination (Venor‐GeM Advance mycoplasma detection kit, Minerva Biolabs).

In vitro cytotoxicity assay

Target cells were washed and subsequently transferred to a 96‐well U‐bottom or flat bottom plate at 105 in 100 μl of complete RPMI. Activated OTI‐GFP T cells were added to the target cells at an effector:target ratio of 2:1. Cells were pelleted for 1 min and incubated at 37°C for the indicated periods. The number of live cells was counted on a Cytoflex LX (Beckman Coulter).

In vitro imaging of CTL–tumor cell interactions

Plastic dishes were coated with Poly‐d‐Lysine (Sigma, 0.01% dilution in H2O) for 10 min at 37°C and then with 5 μg/ml recombinant mouse ICAM‐1 (R&D systems) for 1 h at 37°C. Targets and/or T cells were stained with Indo‐1/AM (2.5 μM, Molecular Probes) for 40 min at 37°C. Cells were then washed and transferred in plastic dishes in complete RPMI without phenol red, and when indicated 100 µM propidium iodide (Sigma) was added at the beginning of image acquisition. In vitro imaging was performed using a two‐photon microscope. Excitation was provided by an Insight DS + Dual laser (Spectra‐Physics) tuned at either 720, 880, or 1040 nm. The following filter sets were used for imaging Indo‐1 and propidium iodide (Indo‐1: 483/32 and 390/40, and PI: 624/40), Twitch‐2B, and DEVD (CFP: 483/32 and YFP: 542/27). Ca2+ signals were quantified using the ratio of calcium bound to calcium‐free fluorescence (Indo‐1, 720 nm excitation) or using the ratio of Venus to Cerulean fluorescence (Twitch‐2B, 880 nm excitation) as described (Bohineust et al, 2020). PI uptake was measured by evaluating PI fluorescence in target cells over time. Images were analyzed using Fiji software.

Generation of CTLs and CAR T cells

Splenocytes were isolated from Rag1 −/− LifeAct‐GFP OT‐I TCR or Ubi‐GFP OT‐I TCR transgenic mice, and red blood cells were removed by ammonium–chloride–potassium lysis. One‐third of the cells was then pulsed with 50 µM of OVA257–264 peptide (SIINFEKL) for 2 h at 37°C. The rest of the cells was incubated at 37°C in complete medium. The two populations were mixed and cultured for 3 days. Cells were then subjected to Ficoll gradient centrifugation to remove dead cells and cultured in complete medium containing human IL‐2 (210 IU/ml; R&D) for 4 additional days. Anti‐CD19 CAR T cells were generated as described previously (Cazaux et al, 2019) using lymph nodes from Rag1 −/− Ubi‐GFP OT‐I TCR transgenic mice. T cells were activated in plates pre‐loaded with 2.5 µg anti‐CD3 mAb (clone 17.A2; BioLegend) and of 2.5 µg/ml soluble anti‐CD28 mAb (clone 37.51; BioLegend) with 10 ng/ml murine IL‐12 (SRP3204; Sigma‐Aldrich). T cells were transduced using the tCD34.2A.amCD19.CD28IEVζ retroviral vector (Cazaux et al, 2019). CAR T cells were cultured for 4 additional days in the presence of hIL‐2.

Tumor establishment and adoptive T‐cell transfer

B‐cell lymphomas (Eµ‐myc‐Twitch‐2B or Eµ‐myc‐DEVD lines) were established by i.v injection of 0.5 × 106 cells in Rag2−/− mice or in C57BL/6 mice conditioned by a sublethal irradiation (4 Gy). T cells or CAR T cells (10–20 × 106) were injected i.v. on the indicated day.

Intravital imaging

Bone marrow intravital imaging was performed 40 h after T‐cell transfer as previously described (Cazaux et al, 2019). Two‐photon imaging was performed with an upright microscope FVMPE‐RS (OLYMPUS) and a 25×/1.05 NA water‐dipping objective (OLYMPUS). Excitation was provided by an Insight deep see dual laser (Spectra Physics) tuned at 880 nm. A 25‐µm‐thick volume of tissue was scanned at 5‐µm Z‐steps and 30–40 s intervals. Images were processed and analyzed using Fiji software. Videos and figures based on two‐photon microscopy are shown as two‐dimensional maximum intensity projections of three‐dimensional data.

Statistical analysis

All statistical tests were performed with Prism v.6.0b (GraphPad). Unpaired t‐tests were used as indicated.

Author contributions

Experiments: RK, MC, BC, ZG, and FL; Experiment design: RK, MC, and PB; Data analysis and manuscript writing: RK and PB.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Click here for additional data file. (4.3MB, docx)

Movie EV1

Click here for additional data file. (2.1MB, zip)

Movie EV2

Click here for additional data file. (1.3MB, zip)

Movie EV3

Click here for additional data file. (1.7MB, zip)

Movie EV4

Click here for additional data file. (1.3MB, zip)

Movie EV5

Click here for additional data file. (4.8MB, zip)

Movie EV6

Click here for additional data file. (1.4MB, zip)

Movie EV7

Click here for additional data file. (2.3MB, zip)

Review Process File

Click here for additional data file. (360.4KB, pdf)

Acknowledgements

We thank members of the Bousso laboratory for critical review of the manuscript. We thank the mouse facility and Technology Core of the Center for Translational Science (CRT) at Institut Pasteur for support in conducting the present study. The work was supported by Institut Pasteur, INSERM and an advanced grant (ENLIGHTEN) from the European Research Council. R.K. was supported by fellowships from the Canceropole and Fondation ARC.

The EMBO Journal (2021) 40: e2020106658.

Data availability

This study includes no data deposited in external repositories.

References

  1. Basu R, Huse M (2017) Mechanical communication at the immunological synapse. Trends Cell Biol 27: 241–254 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Beck RJ, Bijker DI, Beltman JB (2020) Heterogeneous, delayed‐onset killing by multiple‐hitting T cells: Stochastic simulations to assess methods for analysis of imaging data. PLoS Comput Biol 16: e1007972 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bertrand F, Muller S, Roh KH, Laurent C, Dupre L, Valitutti S (2013) An initial and rapid step of lytic granule secretion precedes microtubule organizing center polarization at the cytotoxic T lymphocyte/target cell synapse. Proc Natl Acad Sci USA 110: 6073–6078 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Bohineust A, Garcia Z, Corre B, Lemaitre F, Bousso P (2020) Optogenetic manipulation of calcium signals in single T cells in vivo. Nat Commun 11: 1143 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Boldajipour B, Nelson A, Krummel MF (2016) Tumor‐infiltrating lymphocytes are dynamically desensitized to antigen but are maintained by homeostatic cytokine. JCI Insight 1: e89289 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Breart B, Lemaitre F, Celli S, Bousso P (2008) Two‐photon imaging of intratumoral CD8 T cell cytotoxic activity during adoptive T cell therapy in mice. J Clin Invest 118: 1390–1397 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Budhu S, Loike JD, Pandolfi A, Han S, Catalano G, Constantinescu A, Clynes R, Silverstein SC (2010) CD8(+) T cell concentration determines their efficiency in killing cognate antigen expressing syngeneic mammalian cells in vitro and in mouse tissues. J Exp Med 207: 223–235 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Cazaux M, Grandjean CL, Lemaitre F, Garcia Z, Beck RJ, Milo I, Postat J, Beltman JB, Cheadle EJ, Bousso P (2019) Single‐cell imaging of CAR T cell activity in vivo reveals extensive functional and anatomical heterogeneity. J Exp Med 216: 1038–1049 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Choi PJ, Mitchison TJ (2013) Imaging burst kinetics and spatial coordination during serial killing by single natural killer cells. Proc Natl Acad Sci USA 110: 6488–6493 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Deguine J, Bousso P (2013) Dynamics of NK cell interactions in vivo . Immunol Rev 251: 154–159 [DOI] [PubMed] [Google Scholar]
  11. Dieckmann NM, Frazer GL, Asano Y, Stinchcombe JC, Griffiths GM (2016) The cytotoxic T lymphocyte immune synapse at a glance. J Cell Sci 129: 2881–2886 [DOI] [PubMed] [Google Scholar]
  12. Donnadieu E, Bismuth G, Trautmann A (1994) Antigen recognition by helper T cells elicits a sequence of distinct changes of their shape and intracellular calcium. Curr Biol 4: 584–595 [DOI] [PubMed] [Google Scholar]
  13. Dustin ML, Long EO (2010) Cytotoxic immunological synapses. Immunol Rev 235: 24–34 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Halle S, Keyser KA, Stahl FR, Busche A, Marquardt A, Zheng X, Galla M, Heissmeyer V, Heller K, Boelter J et al (2016) In vivo killing capacity of cytotoxic T cells is limited and involves dynamic interactions and T cell cooperativity. Immunity 44: 233–245 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hamieh M, Dobrin A, Cabriolu A, van der Stegen SJC, Giavridis T, Mansilla‐Soto J, Eyquem J, Zhao Z, Whitlock BM, Miele MM et al (2019) CAR T cell trogocytosis and cooperative killing regulate tumour antigen escape. Nature 568: 112–116 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Harris AW, Pinkert CA, Crawford M, Langdon WY, Brinster RL, Adams JM (1988) The E mu‐myc transgenic mouse. A model for high‐incidence spontaneous lymphoma and leukemia of early B cells. J Exp Med 167: 353–371 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Kashkar H, Seeger JM, Hombach A, Deggerich A, Yazdanpanah B, Utermohlen O, Heimlich G, Abken H, Kronke M (2006) XIAP targeting sensitizes Hodgkin lymphoma cells for cytolytic T‐cell attack. Blood 108: 3434–3440 [DOI] [PubMed] [Google Scholar]
  18. Keefe D, Shi L, Feske S, Massol R, Navarro F, Kirchhausen T, Lieberman J (2005) Perforin triggers a plasma membrane‐repair response that facilitates CTL induction of apoptosis. Immunity 23: 249–262 [DOI] [PubMed] [Google Scholar]
  19. Khazen R, Muller S, Gaudenzio N, Espinosa E, Puissegur MP, Valitutti S (2016) Melanoma cell lysosome secretory burst neutralizes the CTL‐mediated cytotoxicity at the lytic synapse. Nat Commun 7: 10823 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Lenden Hasse H, Lescale C, Bianchi JJ, Yu W, Bedora‐Faure M, Deriano L (2017) Generation and CRISPR/Cas9 editing of transformed progenitor B cells as a pseudo‐physiological system to study DNA repair gene function in V(D)J recombination. J Immunol Methods 451: 71–77 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Lickliter JD, Cox J, McCarron J, Martinez NR, Schmidt CW, Lin H, Nieda M, Nicol AJ (2007) Small‐molecule Bcl‐2 inhibitors sensitise tumour cells to immune‐mediated destruction. Br J Cancer 96: 600–608 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lopez JA, Susanto O, Jenkins MR, Lukoyanova N, Sutton VR, Law RH, Johnston A, Bird CH, Bird PI, Whisstock JC et al (2013) Perforin forms transient pores on the target cell plasma membrane to facilitate rapid access of granzymes during killer cell attack. Blood 121: 2659–2668 [DOI] [PubMed] [Google Scholar]
  23. Marusyk A, Almendro V, Polyak K (2012) Intra‐tumour heterogeneity: a looking glass for cancer? Nat Rev Cancer 12: 323–334 [DOI] [PubMed] [Google Scholar]
  24. Medema JP, de Jong J, Peltenburg LT, Verdegaal EM, Gorter A, Bres SA, Franken KL, Hahne M, Albar JP, Melief CJ et al (2001) Blockade of the granzyme B/perforin pathway through overexpression of the serine protease inhibitor PI‐9/SPI‐6 constitutes a mechanism for immune escape by tumors. Proc Natl Acad Sci USA 98: 11515–11520 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Mempel TR, Pittet MJ, Khazaie K, Weninger W, Weissleder R, von Boehmer H, von Andrian UH (2006) Regulatory T cells reversibly suppress cytotoxic T cell function independent of effector differentiation. Immunity 25: 129–141 [DOI] [PubMed] [Google Scholar]
  26. Michonneau D, Sagoo P, Breart B, Garcia Z, Celli S, Bousso P (2016) The PD‐1 axis enforces an anatomical segregation of CTL activity that creates tumor niches after allogeneic hematopoietic stem cell transplantation. Immunity 44: 143–154 [DOI] [PubMed] [Google Scholar]
  27. Milo I, Bedora‐Faure M, Garcia Z, Thibaut R, Perie L, Shakhar G, Deriano L, Bousso P (2018) The immune system profoundly restricts intratumor genetic heterogeneity. Sci Immunol 3: eaat1435 [DOI] [PubMed] [Google Scholar]
  28. Mohammad RM, Muqbil I, Lowe L, Yedjou C, Hsu HY, Lin LT, Siegelin MD, Fimognari C, Kumar NB, Dou QP et al (2015) Broad targeting of resistance to apoptosis in cancer. Semin Cancer Biol 35(Suppl): S78–S103 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Negulescu PA, Krasieva TB, Khan A, Kerschbaum HH, Cahalan MD (1996) Polarity of T cell shape, motility, and sensitivity to antigen. Immunity 4: 421–430 [DOI] [PubMed] [Google Scholar]
  30. Poenie M, Tsien RY, Schmitt‐Verhulst AM (1987) Sequential activation and lethal hit measured by [Ca2+]i in individual cytolytic T cells and targets. EMBO J 6: 2223–2232 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Snyder JE, Bowers WJ, Livingstone AM, Lee FEH, Federoff HJ, Mosmann TR (2003) Measuring the frequency of mouse and human cytotoxic T cells by the Lysispot assay: independent regulation of cytokine secretion and short‐term killing. Nat Med 9: 231–235 [DOI] [PubMed] [Google Scholar]
  32. Thestrup T, Litzlbauer J, Bartholomaus I, Mues M, Russo L, Dana H, Kovalchuk Y, Liang Y, Kalamakis G, Laukat Y et al (2014) Optimized ratiometric calcium sensors for functional in vivo imaging of neurons and T lymphocytes. Nat Methods 11: 175–182 [DOI] [PubMed] [Google Scholar]
  33. Varadarajan N, Julg B, Yamanaka YJ, Chen H, Ogunniyi AO, McAndrew E, Porter LC, Piechocka‐Trocha A, Hill BJ, Douek DC et al (2011) A high‐throughput single‐cell analysis of human CD8(+) T cell functions reveals discordance for cytokine secretion and cytolysis. J Clin Invest 121: 4322–4331 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Vasconcelos Z, Muller S, Guipouy D, Yu W, Christophe C, Gadat S, Valitutti S, Dupre L (2015) Individual human cytotoxic T lymphocytes exhibit intraclonal heterogeneity during sustained killing. Cell Rep 11: 1474–1485 [DOI] [PubMed] [Google Scholar]
  35. Wiedemann A, Depoil D, Faroudi M, Valitutti S (2006) Cytotoxic T lymphocytes kill multiple targets simultaneously via spatiotemporal uncoupling of lytic and stimulatory synapses. Proc Natl Acad Sci USA 103: 10985–10990 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Appendix

Click here for additional data file. (4.3MB, docx)

Movie EV1

Click here for additional data file. (2.1MB, zip)

Movie EV2

Click here for additional data file. (1.3MB, zip)

Movie EV3

Click here for additional data file. (1.7MB, zip)

Movie EV4

Click here for additional data file. (1.3MB, zip)

Movie EV5

Click here for additional data file. (4.8MB, zip)

Movie EV6

Click here for additional data file. (1.4MB, zip)

Movie EV7

Click here for additional data file. (2.3MB, zip)

Review Process File

Click here for additional data file. (360.4KB, pdf)

Data Availability Statement

This study includes no data deposited in external repositories.

Articles from The EMBO Journal are provided here courtesy of Nature Publishing Group

RESOURCES