Targeting CD22 for B-cell hematologic malignancies

Chimeric antigen receptor (CAR)-T cell therapy has attracted much attention as a cellular immunotherapy. Although CD19 CAR-T cell therapy has achieved promising efficacy in clinical settings [1‐5], 30%–60% of patients eventually relapsed with a poor prognosis [6‐14]. One mechanism of relapse is the downregulation or loss of CD19 on the tumor cell surface [15‐17]. In B-cell acute lymphoblastic leukemia (B-ALL), CD19-negative relapse accounts for up to 83% of relapse cases [6‐11, 18]. In B-cell non-Hodgkin lymphoma (NHL), antigen loss also occurs in one–third of patients experiencing treatment failure after CD19 CAR-T cell therapy [15, 19].

CD22 is restrictively expressed in both normal and malignant B cells, which makes it a potential alternative target to CD19 [20, 21]. CD22 has been identified in the blasts of > 90% of B-ALL cases [22, 23]. Studies on the applicability of targeting CD22 have mainly focused on antibody–drug conjugate (ADC) and CAR-T cell therapy. Two CD22 ADCs, inotuzumab ozogamicin and moxetumomab pasudotox-tdfk, have been approved by the Food and Drug Administration for the treatment of relapsed or refractory (R/R) B-ALL and hairy-cell leukemia, respectively. While inotuzumab ozogamicin is increasingly used in therapeutic settings, moxetumomab pasudotox-tdfk was withdrawn from the market due to the low clinical uptake and complexity of the drug use. CD22 CAR-T cell therapies prove to be effective in treating R/R B-ALL [20, 21, 24‐31]. However, treatment failure is inevitable in CD22-targeted immunotherapies, further narrowing down the avaliable treatment options. Comprehensive knowledge of CD22 molecule will help us to understand the mechanisms of resistance and propose corresponding strategies. This review focuses on the development, the available clinical data and the strategies to improve the efficacy of CD22-targeted therapies, with the aim of providing a perspective on future directions in CD22-targeted immunotherapies.

CD22, also known as sialic acid-binding Ig-like lectin 2, belongs to the siglec family and immunoglobulin superfamily. It consists of seven extracellular IgG-like domains and a 141-amino acid-long cytoplasmic tail [32, 33] (Fig. 1A). CD22 can bind to α 2,6-linked sialic acid residues of surface molecules (such as CD22 itself, CD45 and IgM) on B cells in a “cis” configuration. It can also bind to ligands on other cells as an adhesion molecule in a “trans” configuration [34, 35]. Cis-ligation negatively tunes B-cell receptor signaling mainly in a SH2 domain-containing tyrosine phosphatases 1-dependent manner. Trans-ligation modulates the migration and B-cell receptor signaling threshold of B cells [33, 34]. CD22 is restrictively expressed in the B-cell lineage, particularly in malignant B cells [23, 34, 36‐40]. As an endocytic receptor, ligation of CD22 with its ligands triggers rapid internalization, which enables the application of CD22 ADCs [41].

ADC is composed of an antibody, a chemical linker, and a covalently attached cytotoxic agent (i.e., the payload). The ADC undergoes endocytosis after the antibody binds to the specific antigen on the tumor cell surface, and then releases the payload from the lysosomes. Choosing an optimal antibody is the first step in developing CD22 ADCs. The payload also plays a pivotal role in triggering the crosslinking or breakage of DNA, or inhibiting the tubulin activity, thus leading to cell cycle arrest and tumor cell apoptosis. The linker determines the stability of ADCs and also controls the release of payloads; it is designed to prevent unnecessary clustering of ADCs, which impairs the anti-tumor activity [42, 43]. The structures of CD22 ADCs/recombinant immunotoxins are summarized in Table 1.

BL22 contains a disulfide-stabilized anti-CD22 fragment variable derived from a murine RFB4 antibody and a 38 kDa truncated form of pseudomonas exotoxin A [44]. HA22 is a refined form of BL22 that shows a higher affinity to CD22. In HA22, SSY residues in the hot spot region of complementarity-determining region 3 of heavy chain of BL22 were mutated to THW residues [45, 46]. HA22 showed improved in vivo and in vitro anti-tumor activity compared to BL22 [45]. Unlike moxetumomab pasudotox deriving from HA22, inotuzumab ozogamicin comprises the humanized anti-CD22 monoclonal IgG4 antibody G544 and a chemically linked DNA-damaging payload Calich-DMH. It recognizes the IgG-like domain 1 of CD22 and exerts a potent cytotoxic effect on tumor cells, leading to an obvious tumor mass regression in two lymphoma-xenograft bearing models [47, 48]. Pinatuzumab vedotin incorporates the humanized anti-CD22 monoclonal IgG1 antibody Hu10F4 and a microtubule inhibitor monomethyl auristatin E. Pinatuzumab vedotin demonstrated better anti-tumor activity than the standard rituximab plus CHOP regimen in a xenograft model bearing Ramos cells and also showed anti-lymphoma effects in vitro [49]. CD22-NMS249, with the same antigen-binding region of pinatuzumab vedotin and a more potent anthracycline analogue PNU-159682, displayed better cytotoxicity than pinatuzumab vedotin [50]. Anti-CD22-(LC:K149C)-SN36248 uses the same antibody of pinatuzumab vedotin and two SN36248 molecules as payload. The K149C conjugation site in the light chain promotes the in vivo conjugation stability of the ADC. SN36248 is a non-cleavable linker drug compound with a seco-CBI homodimer tethered to a maleimide linker. Anti-CD22-(LC:K149C)-SN36248 yielded a longer response duration than pinatuzumab vedotin and showed anti-lymphoma effects in several mouse models, including two resistant to pinatuzumab vedotin (Raji and WUS-DLCL2) [51].

The available clinical results [52‐62] for CD22 ADCs in treating B-ALL and B-cell NHL are summarized in Table 2. Inotuzumab ozogamicin was evaluated in 90 patients with R/R B-ALL in a phase 2 study [54] and showed a complete response (CR) rate of 58% and a median duration of remission of 7 months. Seven patients (6.7%) experienced veno‐occlusive disease (VOD)/sinusoidal obstruction syndrome (SOS), among whom six developed VOD/SOS after transplantation. A two-arm, randomized phase 3 study [55, 56] compared the efficacy of inotuzumab ozogamicin (n = 164) and standard chemotherapy (n = 143) in adult patients with R/R B-ALL. The CR rate was also higher in the inotuzumab ozogamicin group (74% vs 31%). The median progression-free survival (PFS) and overall survival (OS) in the inotuzumab ozogamicin group were 5 and 7.7 months, respectively. Notably, the incidence of VOD/SOS was higher in the inotuzumab ozogamicin group (14% vs 2%). A phase 2 study assessed inotuzumab ozogamicin in 48 pediatric and adolescent patients with R/R B-ALL, among whom 10 (20.8%) underwent prior CD19 CAR-T cell therapy, 1 (2.1%) received CD22 CAR-T cell infusion, and 14 (29.2%) were treated with blinatumomab (a CD3/CD19 bispecific T cell engager). This study reported a CR rate of 58.3%, and minimal residual disease-negativity (< 0.01%) was achieved in 66.7% of patients. Twenty-one patients then proceeded to hematopoietic stem cell transplantation (HSCT) after ADC treatment, and six patients (28.6%) developed SOS after HSCT.

Pinatuzumab vedotin, a novel anti-CD22 ADC, has rarely been used as a single agent in clinical settings. The efficacy of pinatuzumab vedotin with or without rituximab was tested in a phase 1 study of adult patients with diffuse large B-cell lymphoma (DLBCL), indolent B-cell NHL, and chronic lymphoblastic leukemia (CLL) [62]. At its recommended phase II dose, the overall response rate (ORR) was 36% in DLBCL and 50% in indolent B-cell NHL. Notably, no therapeutic effect was observed in CLL. The most common toxicity of pinatuzumab vedotin was peripheral neuropathy, especially peripheral sensory neuropathy.

An in vitro study [63] indicated that internalization ability influenced the cytotoxicity of inotuzumab ozogamicin. Clinical data showed that CD22 density on the tumor cell surface correlated with clinical outcomes [55, 61, 64]. KMT2A translocations/rearrangement is a high-risk cytogenetic factor closely associated with low CD22 expression, and positive minimal residual disease status after treatment with inotuzumab ozogamicin [23, 65]. The patients with higher baseline CD22 expression and normal cytogenetics benefited most from inotuzumab ozogamicin. Patients treated with inotuzumab ozogamicin also showed decreased CD22 expression in blasts at relapse [61, 64]. The mechanisms are unknown. Bryostatin-1 is a natural substance that can specifically elevate CD22 surface distribution in a dose- and time-dependent manner [66]. Bryostatin-1 upregulates CD22 expression on CLL cells by activating protein kinase C [66] and on B-ALL cells through potential membrane trafficking [67]. Consequently, bryostatin-1 can be used for pretreatment or in combination with CD22 ADCs [66], though clinical effects need exploration.

Unlike CD22 CAR-T cell therapy, most patients receiving CD22 ADCs are naive to CD19 CAR-T cell therapy. Inotuzumab ozogamicin can effectively reduce the tumor burden and usually serves as a bridging therapy to HSCT or CAR-T cell therapy. Many clinical trials are also exploring the combination of CD22 ADCs with rituximab or other chemotherapies to increase the response depth. Inotuzumab ozogamicin plus mini-hyper-CVD chemotherapy, with or without blinatumomab, represents a feasible therapeutic regimen for elderly patients with newly diagnosed Ph- B-ALL [68‐70]. Inotuzumab ozogamicin plus rituximab, with or without other chemotherapy agents, has also elicited encouraging clinical results in treating R/R B-cell NHL [71‐73].

Unlike ADC, CAR consists of an extracellular antigen-recognizing single-chain variable fragment (scFv), a hinge and transmembrane domain, and an intracellular signaling domain. It can mimic the T-cells’ intrinsic activation mode and initiates their killing to tumor cells without the restriction of the major histocompatibility complex [74]. The investigation on CD22 CAR-T cell therapy is started with the scFvs from two recombinant immunotoxins, HA22 and BL22.

CAR-T cells using HA22-derived scFv did not exert enhanced cytolytic effects than that using BL22-derived scFv for the possible reason that HA22-derived scFv cannot produce a strong activation signal under sufficient antigen stimulation [75]. The novel fully human anti-CD22 antibody m971 has gradually become an appealing alternative to HA22. The epitope recognized by m971 is distinct from that by HA22 and BL22. HA22 and BL22 recognize IgG-like domain 3 of CD22, while m971 targets the most proximal three extracellular domains 5–7 with a relatively low avidity (Fig. 1A) [76]. Moreover, increasing the affinity of m971 scFv did not improve in vitro and in vivo CAR-T cell activity against CD22-low leukemia cells [67]. Second-generation CAR-T cells incorporating m971 scFv displayed a better anti-tumor activity than those with BL22 or HA22-derived scFv [37, 77]. This may be due to the decreased density of CD22 on the tumor cell surface caused by the HA22-mediated internalization, which was not observed with m971 antibody [76, 78]. The novel hCD22.7 scFv was shown to bind to the distal Ig-like domain 1 with high affinity without causing evident CAR-T cell-mediated antigen loss in a mouse model bearing the primary leukemia cells [79]. Several other CD22-targeted scFv structures have shown remarkable preclinical anti-tumor activity and have been adopted in clinical practice [21, 31]. Linkers between heavy and light chains [80] also affect the targeting capacity of CD22 CAR-T cells. Short linker facilitates the formation of immune synapse and spontaneous clustering of CARs without antigen stimulation, thus inducing a tonic activation and improved CAR-T cell function [81] (Fig. 1B).

The available clinical results on CD22 CAR-T cell therapy [20, 21, 24‐31] are summarized in Table 3. The first in-human phase 1 trial of CD22 CAR-T cells, conducted at the National Cancer Institute [20, 24], reported an ORR of 72% (n = 57) and CR rate of 70%. At a median follow-up of 2 years, median OS and relapse-free survival in CR patients were 13.4 months and 6.0 months, respectively; 35% of patients relapsed, and the majority developed CD22-dim or negative disease. Cytokine release syndrome (CRS) and neurotoxicity occurred in 72% and 86% of patients, respectively. Two pilot studies recruited three adult and five pediatric patients with B-ALL [81]. The ORR was 50% and all of the responders achieved CR, with the longest CR being 7 months. CRS occurred in 75% of patients, and one developed grade 3 CRS.

We conducted a subgroup analysis based on the available clinical results of CD22 CAR-T cell therapy (Table 4). While the pooled ORR did not differ with age, the CR rate was higher in children than that in adults (74% vs 57%, P = 0.05). Neurotoxicity tended to occur more frequently in children than adults (28% vs 11%, P = 0.04). In addition, CRS had a predilection for patients with B-ALL instead of those with B-cell lymphoma (87% vs 74%,P < 0.01). The relapse rate was higher in patients with B-ALL than that in patients with B-cell lymphoma (31% vs 8%, P = 0.04). Young patients also had a higher risk of relapse than adult patients (36% vs 6%, P = 0.01). Moreover, previous treatment with CD19 CAR-T cells did not influence the efficacy or safety of CD22 CAR-T cell therapy (CR 73% vs 79%, P = 0.63; CRS 76% vs 83%, P = 0.59; Neurotoxicity 13% vs 9%, P = 0.72).

Lower CD22 density in blasts and a higher tumor burden are associated with poor outcomes. Zheng et al. [82] reported two CD22 splicing isoforms—a Δex5-6-splicing and Δex2-skipping isoform—from the RNA sequencing databases of newly diagnosed pediatric B-ALL patients. The former isoform causes epitope recognition failure in the IgG-like domain 3, while the latter is a CD22 Δex2 variant (Δex2 encodes mRNA containing the initiation codon AUG) that cannot be translated into any identifiable protein. The CD22 variants may partially explain the initial low CD22 density on tumor cell surface. In addition, prior CD22-targeted immunotherapy is a potential predictor of poor efficacy. Tumors can escape the killing of CAR-T cells by decreasing the antigen density on the surface, which has been observed in both CD19 and CD22 CAR-T cell therapy. Notably, Ramakrishna et al. [67] found that the genetic sequence and mRNA levels of CD22 remained unchanged in the tumor cells of relapsed patients, but the surface distribution of CD22 decreased. Moreover, CD22 CAR-T cells did not induce obvious trogocytosis of CD22 on tumor cells [83]. Membrane trafficking resulted in the internalization of CD22 molecules after antigen–antibody interactions [78], which might explain the downregulation of CD22 after CAR-T cell therapy. Notably, CD22 splicing isoforms were not detected after treatment failure with CD22 CAR-T cells.

CD22 downregulation is an important mechanism of treatment failure after CD22 CAR-T cell therapy [20, 21, 29‐31, 81]. Bryostatin-1 can sensitize tumor cells to CD22 CAR-T cell therapy [67]. However, bryostatin-1 cannot affect CD22 expression on tumor cell lines with low primary CD22 expression. Direct exposure to bryostatin-1 can dampen interferon-γ production while enhancing granzyme B secretion in CD22 CAR-T cells. In a mouse model, bryostatin-1 effectively prolonged in vivo persistence and promoted the memory phenotype of CD22 CAR-T cells [67]. Epigenetic modifiers, such as 5-azacytidine and all-trans retinoic acid can also modulate CD22 expression in different cell lines [84]. Although these studies provide proof for combination therapies with CD22 CAR-T cells, further research is needed to determine the optimal timing and duration.

Dual-targeting effectively mitigates the antigen loss associated with CAR-T cell therapy and improves treatment outcomes. Many dual-targeting strategies have been reported, including sequential infusion [85‐88] and co-administration [89] of two CAR-T cell products, co-transduction [90, 91] or sequential transduction [92] of T cells with two vectors conveying different CARs, and transduction of T cells with one vector encoding two separate CARs (bicistronic structure) [93, 94], using tandem [95‐101] or loop [102, 103] scFv structures. Preclinical studies demonstrated that tandem CD19/CD22 CAR-T cells eradicated CD19 + tumor cells effectively but had a limited cytotoxicity on CD22 + tumor cellss compared to single-target CAR-T cells [104]. However, tandem CD19/CD22 CAR-T cell therapy resulted in a higher CR rate than single-target or sequential infusion of CD19 and CD22 CAR-T cells [105, 106]. Recently, Kokalaki et al. [107] screened out 9A8 as a better scFv in CAR design based on its sensitivity to CD22-low tumors after sequential antigen stimulation. They constructed two CAR vectors incorporating the FMC63 or 9AB scFv and developed a dual-target CAR-T cell product using a co-transduction method. CD19/CD22 dual-target CAR-T cell therapy showed relative better efficacy than single-target CAR-T cell therapy, with a CR rate ranging from 83%–100% in B-ALL and 50%–62.5% in B-cell lymphoma [85, 86, 90‐101] (Table 5). Trispecific CAR-T cells targeting CD19, CD20 and CD22 also showed a promising ability for eliminating antigen-heterogeneous tumor cells [108].

CD22 is a potential target especially for patients who have experienced treatment failure with CD19-targeted immunotherapies. CD22-targeted immunotherapies have shown promising efficacy and acceptable toxicities in hematologic malignancies. Despite the difference in patient characteristics, CD22 CAR-T cells outperformed CD22 ADCs in terms of clinical efficacy, reflecting the superiority of T-cell-mediated immunity over the cytotoxic agents. We also found that prior exposure to CD19 CAR-T cells did not profoundly affect the efficacy of CD22 CAR-T cells. Therefore, CD22 CAR-T cell therapy could either be an upfront choice if CD22 density is far higher than CD19 on the tumor cell surface or be a candidate for CD19 CAR-T cell therapy. CRS and neurotoxicity are typical toxicities caused by CAR-T cells. CRS could sometimes be intense and leads to multi-organ damage. But CD22 ADC provides a safer option for those at high risk to develop CAR-T-associated toxicities and the elders, at least as a bridging therapy. CD22 ADCs are not suitable for patients who relapsed after CD22 CAR-T treatment due to the associated antigen loss. Instead, CD22 ADCs can prepare the patients for subsequent CD19 CAR-T cell therapy, as the counts of CD3 + T cells are maintained after the treatment [61]. However, the prolonged B-cell aplasia caused by CD22 ADCs may be an unfavorable factor for subsequent CAR-T cell therapy [109], given that a lower percentage of CD19 + B cells (< 15%) in the bone marrow at infusion correlated with a shorter persistence of CD19 CAR-T cells [11]. One pediatric patient with B-ALL experienced a myeloid lineage switch after the infusion of CD19 CAR-T cells following the treatment with inotuzumab ozogamicin. Notably, KMT2A rearrangement was previously detected in the case, emphasizing the importance of detecting genetic abnormalities at enrollment [61].

Both CD22 ADCs and CAR-T cells were more effective in treating B-ALL than B-cell NHL. This was especially evident with CD22 ADCs, demonstrating a more suitable therapeutic profile against B-ALL. Therefore, both of CD22 ADC and CAR-T cell therapy are feasible for patients with B-ALL. But CD22 ADCs are not recommended to patients with DLBCL or large B-cell lymphoma. As for indolent lymphoma, limited data is accessible on CD22 CAR-T cell therapy, while CD22 ADCs  demonstrate modest efficacy compared to other targeted therapies [110, 111].

Nevertheless, relapse after CD22-targeted immunotherapies remains an unsolved problem. Particularly, the antigen downregulation is apparent in CD22-targeted therapies, and the mechanism is unclear. The preclinical results [66, 67, 83, 84] of bryostatin-1 plus CD22-targeted therapies uphold the combination therapies, but how to schedule and administrate the use of bryoststin-1 still needs more clinical exploration. CD19/CD22 dual-targeting CAR-T cells demonstrated excellent efficacy and can avoid antigen escape to a great extent. Based on the different mechanisms of immunotherapies, CD22 ADC plus CD22 CAR-T, CD19 CAR-T, or CD19 bispecific T cell engager provide new directions for designing effective therapeutic strategies. However, whether this would produce a result greater than one plus one remains unexplored.

Collectively, optimization of CD22 CAR-T cells and ADCs, combination strategies and dual-targeting designs are future perspectives, which shed light on the better clinical application of CD22-targeted therapies.