Introduction

Chimeric antigen receptor T-cell (CAR-T) immunotherapy is a form of adoptive cell therapy (ACT) consisting of genetically engineered T-lymphocytes recognizing specific surface targets of tumor cells. This target is usually CD19 for the treatment of B-lineage non-Hodgkin lymphomas (NHL). Peripheral T-lymphocytes are collected from the patient, transfected with a CAR construction, expanded ex-vivo, conditioned and reinfused.

Autologous CAR-T therapy, used in Europe since June 2018, is standard-of-care treatment for refractory and/or relapsed (R/R) B-cell NHL (B-NHL). Up to the summer of 2024, four FDA-approved CAR-T (tisagenlecleucel, [tisa-cel] Kymriah®, axicabtagene ciloleucel, [axi-cel] Yescarta®, brexucabtagene autoleucel, [brexu-cel] Tecartus® and lisocabtagene maraleucel, [liso-cel] Breyanzi®) have been developed in this context. They are reinfused in clinical practice after lymphodepletion (LD) with fludarabine and cyclophosphamide (FluCy) as recommended [1,2,3,4,5,6].

Although CAR-T represent a major advance in R/R B-NHL [1,2,3, 5, 7], less than half the patients have durable responses [2,3,4,5, 8,9,10,11,12,13]. CAR-T resistance is influenced by the construction and doses of CAR-T [3, 8, 14], high tumor volume [4, 11, 15] and tumoral intrinsic factors such as antigen loss, immune dysregulation and T-cell exhaustion [16,17,18], as well as by LD chemotherapy [19, 20]. The latter facilitates CAR-T engraftment, expansion, and persistence [21]. The response and durability of CAR-T therapy in B-NHL CAR-T is conditioned by CAR-T expansion kinetics (peak and first month area under the curve [AUC]) [2, 3, 22]. Many LD regimens have been tested, but FluCy is the most used, owing to its demonstrated superiority [3, 23].

The aim of this review is to present available clinical data on LD, its impact on ATC therapies, and outcomes in hematological malignancies, with a focus on NHL.

ACT historical background

ACT use cells of the immune system to target and eliminate tumor cells. This began with hematopoietic stem cell (HSC) transplantation (HSCT), followed by the reinfusion of tumor-infiltrating lymphocytes (TIL) and, later, CAR-T. The first transfer of HSC from a donor to a patient occurred in 1950, and the first allogenic HSCT (allo-HSCT) was performed in 1957 [24]. TIL were developed in the 1980’s and CAR-T two decades later.

The term “conditioning” for LD designates in HSCT the preparative regimen administered before HSC infusion. It aims at providing immunosuppression, favoring engraftment and reducing the risk of graft rejection by the recipient immune system. Initially, conditioning regimens were myeloablative, involving total body irradiation and Cy. This provided a potent anti-cancer effect and optimal engraftment, yet was associated to significant toxicity (83% mortality by day 100) [24, 25].

In 1983 [26] the introduction of busulfan and Cy allowed to alleviate toxicity, yet induced prolonged cytopenias, restricting HSCT to younger/fit patients. A shift occurred towards non-myeloablative or reduced-intensity regimens [27], using Flu and lower doses of alkylating agents [28]. Moreover, an optimal Flu exposure (AUC 20 mg*h/L), is a strong predictor of both non-relapse mortality and survival after allo-HSCT for hematological malignancies [29].

ACT using T-lymphocytes was found effective in murine models of nude rats [30], lacking a thymus and hence T-cells. Moreover, it was shown in murine models of cancers that LD (irradiation) prior to T-cell infusion favored their expansion and anti-tumor efficacy [31]. In the absence of LD, infused T-cells declined rapidly and failed to produce a response [31]. Moreover, these models showed that more effector T-cells were needed in the absence of LD, confirming the role of LD in curative T-cell expansion [31].

Initial studies using LD in TIL therapy were conducted in metastatic melanoma [32] where, without LD, responses were of short duration [33]. Yet Cy alone was not potent enough [33], while FluCy provided better and longer responses [34], after transient LD.

Although the doses and duration of FluCy varied between trials, it became clear that profound immune depletion was necessary to enhance the efficacy of TIL in solid cancer [33].

Lymphodepletion for autologous CAR-T in NHL

Clinical application of CAR-T was not fully developed until the late 2010’s with tisa-cel approved by 2017 for acute lymphoblastic leukemia (ALL) [35]. In the field of R/R diffuse large B-cell lymphoma (DLBCL), the phase II JULIET (Table 1) trial administered tisa-cel, after various LD regimens of FluCy or bendamustine, recommended in case of Cy-related grade 4 hemorrhagic cystitis or demonstrated previous resistance to Cy [36]. LD was not required in case of leukopenia (1 × 109/L) in the week prior to infusion. Response rates were similar with FluCy or bendamustine, but FluCy allowed for a higher overall response rate (ORR; 57,6% versus 40,9%), 1-year progression-free (PFS; 39,1% vs 21,2%.) and better overall (OS) survival. Only 2/8 patients who did not receive LD achieved a response [3, 36].

Table 1 Description of lymphodepletion doses and anti-CD19 CAR – T kinetics in pivotal clinical trials in lymphoma.
Full size table

As shown in Table 1, FluCy at slightly different dosages was used with axi-cel in ZUMA-1, with tisa-cell in JULIET and ELARA, with liso-cel in TRANSCEND and with brexu-cel in ZUMA-2 [2, 3, 6, 36,37,38,39], in various types of NHL.

Effects of lymphodepletion on host and tumor cells (Fig. 1)

Endogenous host cells can impact the functionality of ACT by competing for homeostatic and activating cytokines or exerting an immunosuppressive activity. Indeed, successful ACT relies on the differentiation of infused T-cells into functional, long-lived memory cells, facilitated by interleukin (IL)-7, IL-12 and IL-15 through the JAK-STAT pathway [40,41,42], the levels of which are increased by LD. The removal of regulatory T-cells (Tregs), that would tamper these activities is one of the key mechanisms of LD.

Fig. 1: Schematic representation of the effects of lymphodepletion (1) on tumor cells and their immune microenvironment before CAR T-cells infusion (2).
figure 1

Lymphodepletion (LD), primarily consisting of combined fludarabine and cyclophosphamide (FluCy), is administered to induce changes in host cells and the microenvironment. The key effects of LD are: a Elimination of endogenous immune cells (T regs and probably NK cells); enhancing the recognition of tumor antigens by mature dendritic cells (DC) and M1 macrophages. b Induction of immunogenic cell death in tumor cells, which includes the downregulation of 2,3-dioxygenase (IDO), an enzyme that can suppress immune responses. c Improved availability of homeostatic cytokines (IL-7, IL-15, IL-21), which are crucial for survival and proliferation of transferred donor T cells. The final impact will induce the expansion of CAR T-cells (3).

Full size image

At the tumoral level, LD downregulates the expression of indoleamine 2,3-dioxygenase (IDO) [43], an intracellular enzyme metabolizing tryptophan in derivates inhibiting T-cell activity and cytokine production. CAR-T therapy indeed showed no efficacy in a xenograft murine model of IDO-positive NHL cells [44, 45].

LD, particularly with low Cy doses, induces the conversion of suppressive tumor-promoting M2 macrophages into M1 pro-inflammatory and antigen-presenting cells (APC). LD positively influences the production of oxygen radicals through innate immunity [46].

LD activates dendritic cells (DCs), the most potent professional APCs [40, 47], notably liver and spleen DCs during the early phase of lymphopenia. DCs maturation is crucial for them to participate to anti-tumoral immunity. It is triggered, via Toll-like or other receptors by the uric acid issued tumor cell apoptosis [40] and by translocation of the microbiota upon LD-induced damages to mucosal barriers.

A less positive effect of LD, although controversial, is that they could increase the levels of myeloid-derived suppressor cells (MDSCs), liable to impair CAR-T efficacy, through mobilization of hematopoietic progenitor cells from the bone marrow [16, 20, 43, 48, 49].

Effect of lymphodepletion on CAR-T kinetics

LD significantly impacts the expansion of CAR-T (i.e. peak level and persistence), higher peaks/AUC being associated with better PFS and OS [50, 51], notably through higher levels of LD-induced IL-15 [49]. In the same line, high baseline cytokine levels correlate with those after LD and CAR-T peak [16]. Moreover, CAR-T express high levels of receptors for key homeostatic cytokines, providing them with an advantage over other cell types [41]. Compared to no LD, Flu-based LD was shown to positively impact the kinetics of tisa-cel in DLBCL [19]. In a retrospective study of axi-cel [52], strong expanders had more objective ORR at day30 (91% vs. 40%) and better PFS. Similarly, after liso-cel, higher CAR-T expansion resulted in better overall response and CR rate [37, 53].

The next chapter examines the impact of drug type and dosages used for LD.

Lymphodepletion regimens and CD19-AUTOLOGOUS CAR-T in hematological malignancies

High dose cyclophosphamide

Several studies, primarily in ALL, have demonstrated the advantage of using high doses of Cy [54, 55]. Comparing CAR-T therapy in patients with R/R chronic lymphocytic leukemia (CLL) or ALL [54], split in two cohorts respectively without LD or with Cy alone, it was shown that peripheral CAR-T were respectively undetectable at 1 month vs. still detected at 5 weeks. Cy moreover provided a longer B-cell aplasia (BCA), a good sign of CAR-T efficacy. Another trial in R/R ALL using Cy at two doses (1.5 or 3.0 g/m2) [55] found a better lymphodepletion after 3 g/m² Cy, followed by a higher peak of CAR-T in responders (13 days vs 1). This highlights the importance of achieving a profound lymphodepletion and high post-infusion CAR-T peak, in patients with minimal pre-treatment disease burden, in order to achieve clinical response and longer OS [55, 56].

Addition of fludarabine improves CAR-T KINETICS and response. (Table 1)

Although the first CAR-T LD with FluCy in NHL occurred in 2010 [57], protocols only began to incorporate Flu in conditioning for hematological malignancies from 2016 on, owing to the report of improved CAR-T expansion after Flu-based LD [57, 58].

In a phase I clinical trial in ALL [14], FluCy, Cy alone, or etoposide-based LD were used, showing 100% engraftment after FluCy, i.e. detectable peripheral CAR-T, development of BCA and MRD-negative remission, with a median time to peak of 10 days. The peak and AUC were significantly higher after FluCy compared to Cy alone or etoposide, and BCA lasted longer (6,4 months vs 2.1).

In a CLL trial [59], patients receiving Flu-based LD achieved the lowest lymphocyte nadir and greatest peak expansion of CAR-T.

In a trial of liso-cel for B-NHL, FluCy conditioning was also associated to greater CAR-T expansion, higher response rates (50% CR, 72% ORR) and better PFS compared to Cy alone (8% CR, 50% ORR) [23]. Another trial with higher FluCy LD intensity in NHL comparing two dosages of Cy [60] found that Cy 60 mg/kg/d was associated with a favorable cytokine profile correlating with better PFS than Cy 30 mg/kg/d. However, not all patients with high intensity LD achieved such a favorable cytokine profile, indicating that biological individual factors may be also determinant [60]. Using Flu 30 mg/m2 and 3 daily doses of Cy 300 mg/m2 in a phase I first-in-human trial in NHL, yielded an ORR of 70% [61]. Similar CR rates were observed in ALL patients treated with Cy 3 g/m2 alone as LD [62].

Flu is currently dosed based on body surface area and administered IV as a monophosphate prodrug (F-ara-AMP), converted to the circulating metabolite F-ara-A, mainly cleared by the kidney. In ALL, fludarabine exposure is measured by an F-ara-A assay [29], and correlates with leukemia-free survival, CAR-T expansion and better duration of BCA [63]. These results were confirmed in a retrospective real-world analysis of 152 patients with ALL treated by tisa-cel [56], as well as in another with axi-cel in NHL, where optimal Flu exposure correlated with improved PFS [64].

Other agents than flucy

Shortage of Flu in the USA [65] has led to search alternative regimens, bendamustine being a viable option [66, 67]. This alkylating agent offers anti-lymphoma activity, potent lymphodepleting effects, and good tolerability [68]. Moreover, it is metabolized in the liver and not excreted in the kidney, making it appropriate for patients with impaired renal function [66].

A small study in CLL used pentostatin combined with Cy in five patients, bendamustine alone in six, and FluCy in three, without difference in response rates [69].

For tisa-cel, a retrospective multicenter comparison between bendamustine and FluCy resulted in similar efficacy, with lower rates of CRS, neurotoxicity, and hematological toxicity in the bendamustine arm [67]. ORR and PFS were similar with 50% ORR after bendamustine and 42.9% after FluCy. Median PFS were 3.26 and 3.06 months respectively. CRS of any grade were higher in the FluCy group (66.7% versus 40%), without differences in severe CRS. ICANS of any grade were present in 7.8% of patients with bendamustine versus 21.4% for FluCy, severe in 1.1% vs. 9.5%. There were respectively 15.6% and 50% of infections, possibly related to grade 3 neutropenia (28.9% and 90.5%).

For axi-cel, with bendamustine vs FluCy, no differences were seen in PFS and OS, yet there was a lower incidence of ICANS and severe neutropenia with bendamustine [68, 70]. Lymphocyte counts decreased of similar rates after LD. The best ORR/CR were 77.8%/48.1% with bendamustine and 81.0%/50.0% with FluCy. Six-month PFS were 43.8% and 55.6% and 6-month OS 81.5% and 90.4%. Grade ≥3 CRS were observed in 3.7% vs. 4.8% of the patients, grade ≥3 ICANS in 19% vs. 31% and grade ≥3 neutropenia in 68% vs. 100%, while grade 3 infections were similar at 24% vs. 19% respectively [68]. This differs from data obtained with axi-cel in another bendamustine vs. FluCy trial, that reported more febrile neutropenia (13.6% vs. 78.4%) and infections (27.3% vs. 78.4%) [70] with FluCy.

The role of the hypomethylating agent decitabine is currently investigated in CAR-T therapy, with regard to the abnormal hypermethylation in lymphoma. Under decitabine, there is an increased expression of CD19, less T-cell exhaustion, more T-cell activation and modification the tumor microenvironment [71]. Lymphoma cell-lines exposed to decitabine also increase CD19 expression and show no impairment of CAR-T efficacy [71]. Two patients with NHL, conditioned with decitabine and FluCy, achieved optimal responses [71].

Clofarabine has also been used with Cy or bendamustine before tisa-cel or ARI-0001 in pediatric ALL [72,73,74].

Toxicity related to lymphodepleting regimens

Cytopenias and subsequent infections are the most common adverse effects of LD, causing an initial drop in leukocyte counts within the first two weeks [74]. Hematology toxicity is an early and long-lasting complication with a biphasic pattern [75] which may require blood or platelet transfusions or the use of granulocyte-colony stimulating factor (G-CSF) [50, 76,77,78]. Neutropenia is the most common cytopenia (72%) after CAR-T, followed by thrombopenia [79]. Early neutropenia is mostly related to LD while persisting neutropenia beyond day 28 post infusion can be related to the CAR-T construct, CRS or ICANS [79]. EBMT (European group for bone marrow transplantation) recommendations consider using prophylactic G-CSF in neutropenic patients from day +5, although earlier use has no effect on immunotoxicity, CAR-T expansion nor prognosis, yet reduces febrile neutropenia [77, 78].

High dose LD (Flu 125 mg/m2, Cy 60 to 120 mg/kg) has been reported to result in 67% of patients needing platelet transfusion versus 1% after low-dose LD (Flu 90 mg/m2, Cy 300 mg/m2) [68]. As reported above, bendamustine seems less toxic than FluCy [67, 68, 70].

LD, while not being the main factor, might also influence CRS and ICANS. FluCy conditioning has been correlated with the severity of CRS [80]. For axi-cel in DLBCL and follicular lymphoma, any-grade CRS occurred in 91.9% of the patients after FluCy vs. 72.7% after bendamustine, and any-grade neurotoxicity in 45.9% vs18.2% [70].

Flu was initially thought to be related to the neurotoxicity of CAR-T, as it may lead to reversible somnolence and peripheral neuropathy at the time of infusion. However, the evidence to date does not support a direct role for Flu in ICANS, although it could be a contributing factor in case of impaired renal function [81].

LD in other contexts: ANTI-CD30, dual and allogenic CAR-T. (Table 2)

For the treatment of R/R Hodgkin lymphoma (HL) with anti-CD30 CAR-T, different regimens have been evaluated, mostly bendamustine alone, bendamustine and fludarabine or FluCy [82]. Although bendamustine is a potential therapy in R/R HL, it is unlikely to enhance post-CAR-T responses. The combination of Flu with bendamustine promoted a favorable homeostasis of IL-7 and IL-15 compared with FluCy or bendamustine alone, leading to higher antitumor activity and longer CAR-T persistence. However, the 94% 1-year OS did not differ between these regimens [82]. In a phase I trial for CD30-expressing NHL, FluCy allowed for a CAR-T peak in some patients, but all progressed within 6 months, which precluded further development [83].

Table 2 Description of the lymphodepletion used in CAR T-cells under evaluation in clinical trials in lymphoma, other than anti-CD19 CAR T-cells.
Full size table

Dual CAR-T directed to both CD19 and CD22 are under investigation to overcome antigen escape. LD consists of decitabine and FluCy and, so far, some transient grade 3/4 neutropenias have been observed [84].

Allogenic “off-the-shelf” CAR-T, currently used in clinical trials, are generated from healthy donor T-cells. TRAC and CD52 genes are inactivated using TALEN® gene editing to minimize the risk of graft-versus-host disease and improve the compatibility and persistence of CAR-T. For conditioning, alemtuzumab (anti-CD52) may be added in some circumstances. In R/R ALL, UCART19 is under phase I evaluation with a FluCy conditioning with or without alemtuzumab [85]. In R/R aggressive DLBCL, a 3-day LD with FluCy and alemtuzumab has been used [86]. Alemtuzumab seems to be necessary in combination with FluCy before UCART19 [87] for positive IL-7 exposure and UCART19 kinetics, but higher toxicities are expected [74]. In ALL, peripheral UCART19 was detectable from D7, with peak expansion in 72% of the patients between D10 and D17, and a median persistence duration of 28 days. No expansion was observed in 5/18 patients, 3 of whom did not receive alemtuzumab. However, the definitive LD regimen and its impact on UCART19 expansion are currently investigated [85]. Dual allogenic CAR-T against CD19 and CD22 (UCART20x22) are also in phase I/IIa study in R/R NHL with FluCy and alemtuzumab [88]. Preliminary results showed expansion in 3/3 patients with initial detection on day 7 followed by peaks between days 9-14, predominantly CD8+ [88].

Several studies stress the importance of administering LD before each CAR-T infusion, if the procedure needs to be repeated [89, 90]. Better responses have been observed in B-cell malignancies if FluCy was administered before both CART1 and CART2, although an increased dose of CART2 was also needed [88]. In ALL, pre-CAR-T2 intensified LD yielded 71% responders with higher CART2 expansion (yet lower than with CAR-T1) than after standard LD [90].

Another approach is to use NKT-cells which are also potent cytotoxic cells of the innate immune system [91]. In this context, FluCy is being used. Preliminary results indicate responses at day 30 with high peaks [92].

Fourth generation CAR-T or TRUCKs :“T-cells redirected for antigen-unrestricted cytokine-initiated killing” could avoid LD. Indeed, tumor targeting and CAR-T activation and cytotoxicity is completed by the transfection of cytokine genes yielding endogenous production that self-promotes their survival, proliferation and activation [93].

Perspectives

Many questions regarding LD and its future remain unanswered and clinical randomized trials are needed. LD is nonetheless an important factor that significantly affects CAR-T kinetics and response with the goal of achieving the lowest lymphocyte nadir and greatest peak expansion. Flu exposure seems to be a major part of LD, easily modifiable. Moreover, the importance of a personalized approach using PK-directed dosing, based on weight, renal function and drug monitoring, has been stressed, to achieve better outcomes with minimal toxicities [75, 82]. In terms of toxicity, bendamustine appears interesting as LD with a favorable profile, reducing treatment costs and hospitalization duration.