Treatment Approaches to Chronic Lymphocytic Leukemia With High-Risk Molecular Features

The clinical course of chronic lymphocytic leukemia (CLL) is highly variable. Over the past decades, several cytogenetic, immunogenetic and molecular features have emerged that identify patients suffering from CLL with high-risk molecular features. These biomarkers can clearly aid prognostication, but may also be capable of predicting the efficacy of various treatment strategies in subgroups of patients. In this narrative review, we discuss treatment approaches to CLL with high-risk molecular features. Specifically, we review and provide a comprehensive overview of clinical trials evaluating the efficacy of chemotherapy, chemoimmunotherapy and novel agent-based treatments in CLL patients with TP53 aberrations, deletion of the long arm of chromosome 11, complex karyotype, unmutated IGHV, B cell receptor stereotypy, and mutations in NOTCH1 or BIRC3. Furthermore, we discuss future pharmaceutical and immunotherapeutic perspectives for CLL with high-risk molecular features, focusing on agents currently under investigation in clinical trials.


INTRODUCTION
Chronic lymphocytic leukemia (CLL) is the most prevalent form of leukemia in the Western world with an age-standardized incidence rate of 4-5 cases per 100,000 persons per year (1)(2)(3)(4)(5). The clinical course of CLL is characterized by marked heterogeneity, with some patients surviving for more than 10 years without treatment, whereas others suffer rapid disease progression and poor outcome, in spite of the availability of effective treatment regimens. Historically, prognostication in CLL has relied on clinical staging: the Rai and Binet staging systems, developed approximately 40 years ago, are still frequently used in clinical practice (6,7). However, as advanced molecular techniques such as fluorescence in situ hybridization (FISH), next-generation sequencing (NGS) and microarraybased genomic profiling have provided greater insight into the biology of CLL cells, the prognostication paradigm has shifted towards a perspective that not only relies on clinical features, but also incorporates high-risk genetic and molecular biomarkers. Indeed, the CLL international prognostic index (CLL-IPI) incorporates cytogenetic, immunogenetic and molecular features to predict the survival of CLL patients (8). In addition to prognostication, the presence of certain cytogenetic, immunogenetic and molecular features can predict differential responses to treatment. Several of the more newly discovered biomarkers are not routinely measured in clinical research and patient care, despite evidence of their predictive impact. Conversely, some routinely measured biomarkers that were historically associated with poor prognosis in the chemo (immuno)therapy era may have lost their predictive value in the context of novel agent-based therapies. Presently, the international workshop on CLL (iwCLL) guidelines recommend assessment of immunoglobulin heavy chain gene (IGHV) mutational status and TP53 aberrations in every CLL patient before the initiation of treatment (9). TP53 aberrations should be assessed by both targeted FISH and either Sanger sequencing or NGS. Whereas the IGHV mutational status is stable over time, additional (cyto) genetic abnormalities with therapeutic implications may be acquired over the course of the disease, necessitating reassessment at every subsequent line of therapy (9).
This narrative review discusses such therapeutic implications of CLL with high-risk molecular features. Specifically, we review the predictive impact, if any, of TP53 aberrations, deletion of the long arm of chromosome 11, complex karyotype or genomic complexity, unmutated IGHV, B cell receptor stereotypy, mutated NOTCH1 and mutated BIRC3. Furthermore, we discuss future perspectives for CLL with high-risk molecular features, focusing on upcoming agents in the therapeutic armamentarium of CLL.

TP53 ABERRATIONS
Among the numerous prognostic and predictive biomarkers that have been identified over the previous decades, TP53 aberrations indisputably remain the single most impactful genetic lesion in CLL. The TP53 locus encodes the tumor-suppressor protein p53, which plays a key role in cell division, apoptosis and genomic stability. TP53 signaling can be impaired through deletion of the TP53 gene in the chromosomal locus 17p13.1, i.e. del(17p), or through genetic lesions, including missense and nonsense mutations, deletions, insertions or splice-site mutations. Cumulatively, these aberrations are present in 4-8% of all CLL patients at diagnosis, 10% at start of first-line therapy and 30-40% in relapsed or refractory (R/R) CLL patients who were previously treated with chemoimmunotherapy (10). In CLL patients, impaired TP53 signaling is associated with a poor response to chemotherapy and chemoimmunotherapy. Chemotherapeutic agents exert their cytotoxic effects by causing DNA damage. In TP53 wildtype cells, such irreparable damage results in apoptotic cell death. However, if TP53 is not expressed or not functional, chemotherapy-induced DNA damage does not lead to apoptosis and as consequence, administration of these drugs leads to accumulation of detrimental genomic mutations while failing to induce cell death, possibly worsening disease prognosis.
Based on the evidence from the trials discussed above, chemoimmunotherapy is no longer considered an acceptable first-line treatment for patients with TP53 aberrations. Initially, the anti-CD52 monoclonal antibody (mAb) alemtuzumab, with or without methylprednisolone, was considered the only effective pharmaceutical treatment in patients with TP53 aberrations, despite limited efficacy (overall response rate [ORR], 82%, median PFS 11.8 months, median OS 23 months) and an unfavorable toxicity profile (18). In addition, an allogeneic hematopoietic stem cell transplantation (alloHSCT) was considered as a first-line consolidation treatment for patients with TP53 aberrations (19)(20)(21). The GCLLSG CLL3X trial demonstrated that the 6-year OS rate after alloHSCT was approximately 60%, irrespective of TP53 aberrations (19). Although potentially curative, alloHSCT is only available for a highly selected young and fit CLL patient population, given the high risk of transplant-related morbidity and mortality. Thus, historically, treatment of patient with TP53 aberrations has been challenging. Fortunately, alemtuzumab and alloHSCT have been replaced as treatment of choice for these patients by newer agents with greatly improved efficacy.
Advanced insight in the pathophysiology of CLL have led to the development of small molecule inhibitors that have revolutionized the treatment of patients with TP53 aberrations. Ibrutinib is an irreversible inhibitor of Bruton's tyrosine kinase (Btk), a signaling molecule downstream of the B cell receptor (BCR) (22)(23)(24). In the phase III RESONATE-2 trial, the efficacy of ibrutinib was compared with chlorambucil monotherapy  (32). This observation has not yet been validated in larger cohorts, or in different treatment settings. Importantly, ibrutinib monotherapy needs continuation until progression and is associated with adverse events such as fatigue, diarrhea, nausea, bleeding complications, cardiac arrhythmias and, rarely, sudden death (in 1% of treated patients) (33). In addition, in some population-based studies, adverse events were more severe resulting in a higher discontinuation rate, compared with those in clinical trials (34)(35)(36). The second-generation Btkinhibitor acalabrutinib has a more selective binding profile and could therefore potentially overcome ibrutinib-associated toxicities. In the phase III ELEVATE-TN study, acalabrutinib and acalabrutinib plus obinutuzumab (O-acalabrutinib) were compared with O-Clb in unfit, treatment-naïve CLL patients (37). The estimated median 24-month PFS rate in patients with del(17p) was longer following treatment with O-acalabrutinib, compared with O-Clb (88%; 95%CI, 61-97% versus 22%; 95%CI, 5-45%). Similar results were obtained in patients with mutated TP53 (24-month PFS rate 95%; 95%CI 70-99% versus 19%; 95% CI 5-41%).
Another small molecule inhibitor is the Bcl2-inhibitor venetoclax. Although this compound mainly acts through induction of apoptosis, it does so in a p53-independent manner (38) Venetoclax has demonstrated comparable efficacy in R/R CLL with TP53 aberrations. In the phase II Pivotal trial, venetoclax monotherapy was evaluated in R/R CLL patients with del(17p). The ORR was 77% for the overall cohort. The estimated 24months PFS and OS were 66% (95%CI, 55%-74%) and 73% (95% CI, 65%-79%). In the phase III MURANO trial, venetoclax plus rituximab (Ven-R) was compared with BR in physically fit, R/R CLL patients (52)(53)(54). Outcome with Ven-R was superior in terms of median PFS as compared with BR in patients with del (17p) (NR after 48 months versus 15.4 months, respectively). However, in a pooled analysis of four early-stage trials, patients with either del(17p) or TP53 mutations remained at higher hazard of relapse, following venetoclax-based treatment (HR 1.7; 95%CI, 1.2-2.4) (55).
Taken together, given the availability of more efficacious drugs, chemoimmunotherapy is no longer a suitable treatment option for CLL patients with TP53 aberrations, both in first-line and in R/R settings. As a consequence, ideally every CLL patient with an indication for treatment should undergo cytogenetic and molecular testing for TP53 disruption. At present, patients with TP53 aberrations qualify for treatment with novel agents such as ibrutinib, acalabrutinib or venetoclax. Still, even after novel agent-based treatment, patients with TP53 disruption seem to have inferior outcome, compared with patients with intact TP53, although these differences are not statistically significant in every trial. At present, there is no preference for either ibrutinib, acalabrutinib or venetoclax if considering treatment for patients with TP53 aberrations. As such, the choice should be determined by the physician and the patient jointly, taking into account treatment duration, side effects, comorbidities and previous lines of therapy. A complete overview of clinical trials comparing treatment regimens in previously untreated or R/R CLL patients with TP53 aberrations is given in Table 1 and Table 2, respectively.

DELETION OF THE LONG ARM OF CHROMOSOME 11
Deletion of the long arm of chromosome 11, i.e. del(11q), is one of the most common structural chromosomal aberrations in CLL. At diagnosis, del(11q) is present in 10% of patients with early-stage and 25% of patients with advanced stage, treatmentnaïve CLL (82,83). Patients carrying a del(11q) characteristically have bulky disease, rapid progression and a shorter OS (82). The minimally deleted region in del(11q) encompasses several tumor suppressor genes such as ATM, FDX, MLL and RDX. The tumor suppressor gene ATM encodes the serine-threonine kinase ATM, which is important in the repair of double-strand DNA breaks. Deleterious ATM mutations in the residual allele can be found in 36% of the CLL patients with a del(11q) and are associated with a poorer prognosis, compared with patients with del(11q) alone (84)(85)(86). Since ATM functions as a positive upstream regulator of p53, loss of ATM could interfere with chemotherapyinduced apoptosis.

First-Line Setting
Concordantly, in the chemotherapy era, prognosis for CLL patients with del(11q) was poor (83). In the US E2997 trial, first-line treatment with FC did not improve PFS in patients with del(11q), compared with F alone (median PFS 25.2 and 16 months, respectively). In the FC arm, PFS was 25.2 months and NR after a follow-up of 60 months for patients with del(11q) and those without, respectively. Moreover, in a multivariable analysis, del(11q) was strongly associated with reduced PFS (HR 1.904; P=0.006) (13,87). Similarly, in the CLL8 trial, first-line treatment with FC resulted in a complete response (CR) rate of only 19% in del(11q) patients (15,16). In contrast, in this trial, patients with del(11q) responded very well to treatment with FCR (CR rate 71%). In addition, the 5-year PFS rate for patients with del(11q) was 11.4% after FC, compared with 31.4% after FCR (HR 0.47; 95%CI, 0.32-0.68). Similarly, the 5-year OS rate of patients with del(11q) was superior after treatment with FCR, compared with FC (HR 0.35; 95%CI, 0.20-0.61). Indeed, subgroup analysis revealed that patients with del(11q) and mutated IGHV genes responded very well to FCR, with outcomes similar to patients without del(17p) and del(11q) (88). In a retrospective observational cohort study by Rossi et al., physically fit, treatment-naïve CLL patients were treated with FCR (17). After a median follow-up of 70 months, median PFS was 43.5 months (95%CI, 32.2-54.7) for patients with del(11q) and 56.9 months (95%CI, 47.1-66.6) for patients without del (11q) (P=0.01). In a multivariable analysis, del(11q) was identified as an independent predictor of PFS (HR 1.67; 95% CI, 1.13-2.46). Furthermore, in the GCLLSG CLL10 trial, a phase III non-inferiority trial comparing FCR to BR in physically fit, treatment-naïve CLL, patients with del(11q) had shorter PFS in the BR arm (HR 2.33; 95%CI, 1.47-3.67) (89). In summary, although the introduction of chemoimmunotherapy, especially FCR, has significantly improved the outcome of patients with del (11q), their prognosis after chemoimmunotherapy is still inferior, compared with patients without del(11q).
Treatment with ibrutinib has proven very effective in patients with del(11q), yielding outcomes similar to patients without del (11q) (25-27, 30, 90). For example, after first-line treatment with ibrutinib alone in the RESONATE-2 trial, the 60-month PFS rate was 79% in patients with del(11q), compared with 67% for patients without del(11q) (25)(26)(27). Surprisingly, pooled data from three phase III ibrutinib trials (RESONATE, RESONATE-2 and HELIOS) demonstrated that ibrutinibtreated patients with del(11q) had slightly longer PFS, compared with ibrutinib-treated patients without del(11q) ( Table 3.       Table 4. In summary, in the historical context of chemotherapy, del (11q) was considered a marker of poor prognosis. Treatment with chemoimmunotherapy, specifically FCR, largely mitigates the deleterious outcome associated with del(11q), although survival of patients without del(11q) after FCR remains somewhat superior. Treatment with ibrutinib, acalabrutinib or venetoclax yields equal outcomes in patients with and without del(11q).

COMPLEX KARYOTYPE/GENOMIC COMPLEXITY
The prognostic and predictive value of complex karyotype (CK) or genomic complexity (GC) in CLL has been reported in an increasing number of studies over the past decade. CK is traditionally defined as CLL with 3 or more cytogenetic aberrations, as measured by chromosomal banding analysis (CBA) or interphase FISH. When chromosomal aberrations are detected using chromosomal microarray analysis (CMA), the presence of 3 or more aberrations it is generally referred to as genomic complexity (GC). CBA and interphase FISH are complementary techniques (94). More specifically, CBA can detect abnormalities that cannot be found by a limited panel of FISH probes and FISH can detect abnormalities that cannot be found by CBA due to an inherently lower resolution (95)(96)(97)(98). As such, performing CBA in combination with interphase FISH results in a more sensitive measurement of CK. Alternatively, CMA, often referred to as digital karyotyping, allows for screening the entire genome, and therefore has a higher detection rate of cytogenetic abnormalities, but may fail to detect cytogenetic aberrations without copy number aberrations, such as balanced translocations. Another advantage of CMA over CBA is that it is a DNA-based analysis and therefore culturing of B-cells in combination with mitogen stimulation is not needed to generate chromosome metaphases (99)(100)(101).
CK is associated with poor OS in CLL patients (98,(102)(103)(104)(105)(106)(107)(108)(109)(110)(111)(112)(113). Indeed, in a large multicenter cohort study, Baliakas et al. confirmed that the presence of CK was associated with a shorter time-to-first-treatment (TTFT), compared with CLL patients with a normal karyotype (P=0.01) (114). Furthermore, TTFT was even shorter in patients harboring ≥5 chromosomal aberrations (P<0.001). The latter was confirmed in 2019, when the same research group proposed a hierarchical model Fischer et al.    (49). In contrast, in a pooled analysis of three phase III ibrutinib trials for both treatment-naïve and previously treated CLL patients (the RESONATE-2, RESONATE and HELIOS trials), no effect of the presence of CK on PFS could be demonstrated (91). Specifically, the 42-month PFS rate was 63% in patients with CK, compared with 69% in those without (HR 1.02; P=0.95). Of note, the RESONATE-2 and HELIOS trials excluded patients with del (17p) (25,92). Likewise, the ALLIANCE trial did not substantiate inferior survival in patients with CK after first line treatment with BR, ibrutinib and R-ibrutinib, compared with those without CK (HR 1.01; 95%CI, 0.68-1.51) (30). The impact of CK on treatment with acalabrutinib has only been evaluated in a single phase II trial in R/R CLL patients (81). PFS was shorter for patients with CK, compared with the overall cohort (33 months versus NR after 41 months).
The MURANO trial analyzed the impact of GC (≥3 cytogenetic aberrations, as measured by CMA) after treatment with Ven-R (53). In this trial, GC was stratified into low GC (3-4 aberrations) and high GC (≥5 aberrations). After treatment with Ven-R, patients without GC had superior PFS, compared with patients with low-and high GC (HR 2.9; 95%CI, 1.1-3.6). In addition, the co-occurrence of TP53 aberrations only negatively affected PFS in patients with high CK, but not low CK. In concordance, in a pooled analysis of three phase I trials, Anderson et al. showed that the presence of CK was associated with shorter PFS in R/R CLL patients treated with venetoclax monotherapy (HR 6.61; 95%CI, 1.47-29.75) (120).
Altogether, the predictive impact of CK is challenging to disentangle, mostly due to inconsistent reporting and cooccurrence of TP53 aberrations. While patients with CK seem to have a high chance of relapse, even in the absence of TP53 aberrations, after treatment with chemoimmunotherapy, evidence of the impact of CK on the efficacy of novel agents such as ibrutinib and venetoclax is contradictory. Consequently, further research into the impact of CK is warranted, including clear reporting on the definition of CK and multivariable analysis to correct for the co-occurrence of TP53 aberrations. For an overview of clinical trials comparing treatment regimens in CLL patients with and without CK, see Table 5.

UNMUTATED IGHV
The prognostic and predictive impact of the somatic hypermutation (SHM) imprint on the IGHV gene of the leukemia-specific BCR rearrangement has been recognized over the past two decades (121,122). Whereas CLL with abundant SHM (IGHV germline identity <98%, 'IGHV mutated', or M-CLL) generally has an indolent disease course, a paucity of SHM (IGHV germline identity ≥98%, 'IGHV unmutated', or U-CLL) is a biomarker of high-risk disease and is associated with a shorter TTFT and OS, compared to M-CLL patients (121)(122)(123). Consequently, in early stage CLL, the proportion of patients with unmutated IGHV is around 50%, whereas this prevalence enriches to approximately 60% at firstline therapy and up to 80% in R/R CLL.

First-Line Setting
After first-line treatment with FCR, a significant proportion of patients with M-CLL experience durable remission, whereas almost all patients with U-CLL eventually relapse (88,124 Controversy remains as to whether chemoimmunotherapy or novel agent-based regimens are more appropriate as first-line treatment for patients with U-CLL. Head-to-head comparisons in U-CLL patients have demonstrated that, compared with chemoimmunotherapy, treatment with ibrutinib results in longer PFS. In the ALLIANCE trial, while U-CLL patients in both ibrutinib-based arms had similar median PFS (NR after 38 months), median PFS in the BR arm was markedly shorter (39 months, 95%CI,32-NR) (30). Likewise, the ECOG-1912 phase III trial compared the efficacy of R-ibrutinib to FCR in previously untreated CLL (90). In this trial, the 3-year PFS rate after FCR was 62.5% in patients with U-CLL, compared with 90.7% after Ribrutinib (HR 0.26; 95%CI, 0.14-0.50). However, an OS benefit for patients with U-CLL after first-line treatment with ibrutinibbased regimens, compared with chemoimmunotherapy, has not been conclusively demonstrated. Consequently, guidelines differ on the most appropriate first-line regimen for U-CLL.
There is similar ambiguity regarding the most appropriate first-line treatment for unfit U-CLL patients, in which treatment with FCR or BR is contraindicated. In the CLL11 trial, first-line treatment with O-Clb resulted in significantly longer PFS, both compared with Clb alone (HR 0.23; 95%CI, 0.13-0.42) and R-Clb (HR 0.39; 95%CI, 0.29-0.53) in U-CLL (117). In the RESONATE-2 trial, the 18-month PFS of U-CLL patients was 89% after treatment with ibrutinib, compared with 47% after treatment with Clb (25). As both these trials demonstrated an OS benefit over Clb alone, treatment with O-Clb, R-Clb or ibrutinib has become the cornerstone of first-line therapy in unfit CLL patients. In a similar setting, U-CLL patients in the iLLUMINATE trial treated with O-ibrutinib had longer median PFS, compared with O-Clb (NR after 31.3 months    (52). In R/R CLL, venetoclax, ibrutinib and acalabrutinib have not been compared in a head-to-head fashion.
In summary, for refractory or early relapsed U-CLL, novel agent-based regimens, either ibrutinib-, acalabrutinib-or venetoclax-based, are appropriate treatment options. For an overview of clinical trials comparing treatment regimens in R/ R U-CLL, see Table 7.

BCR STEREOTYPED SUBSETS
The molecular composition of the leukemia-specific BCR rearrangement has importance beyond its mutational status. Despite the immense theoretical variety of the BCR repertoire, subgroups of unrelated CLL patients express (quasi)identical leukemia-specific BCRs, a phenomenon known as BCR stereotypy. Although cumulatively, BCR stereotyped subsets are common, encompassing up to 41% of all CLL, the prevalence of each individual subset is low: the largest subset, subset #2 (defined as patients with a BCR comprised of IGHV3-21, IGLV3-21, with a short, stereotypic heavychain complementarity-determining region of 9 amino acids), represents around 2.5% of all patients (128). Certain  (129,130). Due to their low individual prevalence, little is known about the predictive impact of BCR stereotyped subsets. Jaramillo et al. analyzed the pooled results from the CLL8, CLL10 and CLL11 trials, which evaluated the efficacy of chemoimmunotherapy as first-line treatment for CLL (129). In these trials, compared with all other patients with mutated IGHV, patients with subset #2 and mutated IGHV had significantly shorter time to next treatment (TTNT) (HR 2.01; 95%CI, 1.23-3.28), numerically comparable to U-CLL patients. As of yet, there is no available data regarding the predictive impact of BCR stereotypy in the context of novel agent-based treatment regimens. Consequently, the therapeutic consequences of BCR stereotypy remain undefined, and testing for stereotypy has thus far not been embedded in regular CLL care.

NOTCH1 MUTATED CLL
Activating mutations in NOTCH1, most often located in the PEST-domain or 3'-untranslated region (3'-UTR), are present in around 6-12% CLL patients at diagnosis and in 15-20% of patients with relapsed or refractory disease (131). CLL patients harboring a mutation in NOTCH1 (NOTCH1-mut) have shorter OS, compared with their NOTCH1-wildtype (-wt) peers (132)(133)(134). Interestingly, some trials have provided evidence that NOTCH1-mut CLL patients may not benefit from treatment with an anti-CD20 mAb. In the CLL8 trial, NOTCH1-wt patients benefited significantly from inclusion of rituximab in their treatment regimen (FC: median PFS 32.8 months versus FCR: median PFS 57.3 months, P<0.001) (16). Contrastingly, the median PFS of NOTCH1-mut patients did not improve significantly upon the addition of rituximab (FC: 33.9 months versus FCR: 34.2 months, p=0.9). Multivariable survival analysis yielded a statistically significant regression coefficient for the interaction term between NOTCH1 status and treatment arm (HR 1.65; 95%CI, 1.076-2.535), thereby satisfying the formal criterion for a predictive variable (16). Concordantly, Dal Bo et al. demonstrated that NOTCH1-mut patients, in contrast to NOTCH1-wt patients, do not benefit from rituximab consolidation following treatment with fludarabine and rituximab (135). Finally, in the phase III COMPLEMENT1 study, which evaluated the efficacy of ofatumumab-Clb compared with Clb alone in unfit, treatment-naïve CLL patients, NOTCH1-mut patients did not benefit from the incorporation of ofatumumab in their treatment regimen (60). More specifically, the median PFS of NOTCH-wt patients treated with ofatumumab-Clb was longer, compared with those treated with Clb alone (23.8 months versus 13.3 months, HR 0.50; 95% CI, 0.39-0.63), whereas NOTCH1-mut patients had similar PFS, regardless of the treatment arm (17.2 months versus 13.1 months, HR 0.81; 95%CI,0.50-1.31). Again, a statistically significant interaction term between NOTCH1 status and treatment arm provided evidence of the predictive impact (P=0.05) (60). Similar research for obinutuzumab, a type II anti-CD20 mAb, has not been performed and is warranted. PFS of NOTCH1-mut patients was similar to that of NOTCH1wt patients in the RESONATE, CLL14 and MURANO trials, suggesting that NOTCH1 status is not predictive in the context of novel agents (39,46,53). An overview of all clinical trials evaluating NOTCH1 mutations in the context of therapy is given in Table 8.

BIRC3 MUTATED CLL
Deleterious mutations in BIRC3, a negative regulator of noncanonical NF-kB signaling, are present in 3-5% of newly diagnosed CLL patients (84). Though relatively rare, BIRC3 mutations are associated with poor outcome, with a 10-year OS rate in the chemoimmunotherapy era of just 29%, which was comparable to the OS of patients with TP53 aberrations (137). Evidence from several trials suggests that patients with a BIRC3 mutation (BIRC3-mut) respond poorly to chemoimmunotherapy. Firstly, BIRC3 mutations are enriched in patients with fludarabine-refractory disease, with a prevalence of up to 24% (84). Furthermore, Diop et al. demonstrated that after treatment with FCR, BIRC3-mut patients had significantly shorter PFS, compared with patients without BIRC3 mutations (BIRC3-wt) (median PFS 26.4 months versus~54 months, P<0.001) (138). Moreover, in the CLL14 trial, the 24-month PFS rate of BIRC3mut patients after treatment with O-Clb was considerably lower compared with BIRC3-wt patients, and similar to that of patients with del(17p) (BIRC3-mut: 14.3%, del(17p): 23.1%, all patients: 64.1%) (39,40). Indeed, the ORR of BIRC3-mut patients after O-Clb was 38%, considerably lower than the ORR of the overall O-Clb arm (71%, P<0.05). In contrast, the presence of a BIRC3 mutation does not seem to associate with inferior response to novel agents. In the Ven-O arm of the CLL14 trial, the 24-month PFS rate of BIRC3-mut patients was similar to the overall 24month PFS rate in that arm (85.7% versus 88.2%). Similarly, in the RESONATE trial, PFS after treatment with ibrutinib for BIRC3mut and BIRC3-wt patients was not significantly different (46).
Interestingly, BIRC3 is located on chromosome 11q22, and is co-deleted together with ATM in 80% of CLL cases with del(11q) (85). In addition, genomic BIRC3 defects mainly cluster in del (11q) patients, leading to biallelic loss of BIRC3 (85). As such, the clinical significance of trials that report on the impact of mutated BIRC3 need to be interpreted with caution, and ideally stratified for the presence of del(11q) to avoid confounding effects. Monoallelic loss of BIRC3 did not influence survival after chemotherapy in the LRF CLL4 trial, whereas biallelic loss of BIRC3 was associated with shorter OS (median OS 3.3 years versus 4.8 years, P=0.03) (56). Additional research on the predictive impact of monoallelic versus biallelic defects in BIRC3 is warranted. An overview of all clinical trials evaluating BIRC3 in the context of therapy is given in Table 9. Notwithstanding the significant advances in the field of CLL therapy, as detailed above, a proportion of CLL patients will exhaust all currently approved treatment options. In general, these are patients suffering from CLL with high-risk molecular features, such as aberrant TP53 signaling, presence of CK/GC, or both. For these patients, a number of experimental treatments are currently under development (see Table 10).
Zanubrutinib is a second-generation Btk inhibitor, which, compared with ibrutinib and acalabrutinib, has fewer off-target effects and a longer half-life. In a phase I trial in previously untreated or R/R CLL patients, zanubrutinib monotherapy resulted in an ORR in the overall cohort of 96.2%, and an ORR of 100% in patients with del(17p) (140). Moreover, in another phase I trial, the efficacy and safety of zanubrutinib plus obinutuzumab (O-zanubrutinib) was evaluated (143).
This combination yielded an ORR of 100% in previously untreated CLL patients and 92% in R/R CLL patients. The ORR was 100% and 80% in treatment-naïve and R/R CLL patients with del(17p), respectively. The currently ongoing phase III SEQUOIA trial will evaluate the efficacy of zanubrutinib compared with BR in treatment-naïve CLL patients (148). Considering the poor outcome associated with any chemoimmunotherapy in patients with del(17p), those patients were not randomized in this trial but assigned to receive single-agent zanubrutinib, analyzed separately. Tam et al. published the primary report on safety and efficacy of the latter cohort (141). The ORR was 94.5%, with estimated 18months PFS and OS of 88.6% (95%CI, 79%-94%) and 95.1% (95%CI, 88%-94%), respectively. The currently ongoing, phase III ALPINE trial will evaluate whether in R/R CLL, the efficacy of zanubrutinib is non-inferior, compared with ibrutinib (149).
To overcome treatment resistance against covalent Btk inhibitors, pirtobrutinib, previously known as LOXO-305, a noncovalent Btk inhibitor, has been developed. In the phase I/II  BRUIN study, the safety and efficacy of pirtobrutinib were evaluated in R/R B-cell malignancies, including CLL (147). The ORR was 63% (95%CI, 55%-71%) for patients with CLL or SLL, irrespective of whether patients previously discontinued a covalent BTK inhibitor for progression, toxicity or other reasons. In patients with del(17p) and/or TP53 mutation the ORR was 79%, compared with 60% in patients with del(11q) and 68% in patients with U-CLL. The currently ongoing phase III BRUIN CLL-321 trial will compare the efficacy of pirtobrutinib to either R-IDELA or BR in R/R-CLL patients that have been previously treated with a covalent Btk inhibitor (150). Whereas IDELA only targets the delta subunit of PI3K, the novel agent duvelisib is a dual inhibitor of both the delta and gamma isoforms of PI3K. In the phase III DUO trial, duvelisib treatment was evaluated in patients who were progressive on ofatumumab (144,145). Duvelisib treatment resulted in an ORR of 77%, with and median PFS of 15.7 months. Comparable responses were seen in patients with del(17p) and/or TP53 mutation (ORR, 77%, median PFS 14.7 months). Based on these results, the Food and Drug Administration (FDA) approved duvelisib for the treatment of R/R CLL and SLL in September 2018 (151). Duvelisib in combination with FCR was recently evaluated in young treatment-naïve CLL patients. The overall response was 88%. Hematological toxicity and infectious complications were common. Only three patients with TP53 aberrations were identified, of whom two responded to the treatment (146).
Ublituximab is a next-generation, glyco-engineered, type I, anti-CD20 mAb that binds to a unique CD20 epitope which is different from the target site of rituximab, obinutuzumab and ofatumumab. In a phase II trial, ublituximab plus ibrutinib yielded an ORR of 88% in all R/R patients and 95% of patients with either TP53 aberrations or del(11q) (139). In a recent, multicenter, phase III trial, ublituximab plus ibrutinib was compared with ibrutinib alone in R/R CLL patients (139). In the overall cohort, the ORR was 83% for ublituximab plus ibrutinib and 65% for ibrutinib alone (P=0.02

CELLULAR IMMUNOTHERAPY AS NEW PERSPECTIVE FOR CLL PATIENTS WITH HIGH-RISK MOLECULAR FEATURES
Although novel-based agents have revolutionized CLL therapy, they remain incapable of complete disease eradication. To this day, alloHSCT remains, in the context of CLL, the sole treatment with curative intent. Notwithstanding the availability of highly effective, highly tolerable agents, alloHSCT remains a relevant treatment option in several specified situations (152,153). Currently, alloHSCT can be considered for patients with a relapse after chemoimmunotherapy, either in the presence of TP53 aberrations (high-risk category 1) or with additional failure to BTK inhibitors and/or BCL2 inhibitors, irrespective of the presence of TP53 aberrations (high-risk category 2) (153)(154)(155). In the high-risk category 1, the long-term benefits and risks of alloHSCT should be carefully balanced on an individualized basis. Specifically, a younger patient age (<65 years), absence of comorbidities and the availability of a suitable stem cell donor would argue in favor of an alloHSCT, whereas in the converse situation, novel-based agents would be more suitable. Contrastingly, alloHSCT is a more proportional treatment option for patients in high-risk category 2, considering their markedly poor prognosis, due to the limited availability of alternative options (154).
However, the field of immunotherapy has in recent years been revolutionized by the generation of chimeric antigen receptor (CAR) cytotoxic cells. These cells, most often T cells, are molecularly modified to express a single-chain antibody-variable fragment (scFab) with specificity for a marker that is ubiquitously expressed on the malignant target cells, fused to an intracellular CD3z domain. CAR efficacy has been further enhanced by the inclusion of a costimulatory domain, most often either CD28 or 4-1BB. CAR-T cells have been approved for the treatment of acute lymphoblastic leukemia, certain types of large B-cell lymphoma and multiple myeloma (156)(157)(158). Several investigators have evaluated the efficacy of CAR-T cells, most often directed against CD19, in the setting of R/R CLL (see Table 11). While the reported efficacy differs from study to study, the ORR, CR and median PFS reported in the larger studies (n≥10) are markedly lower (ORR: weighted mean 53%, range 38-71%, CR: weighted mean 26%, range 21%-29%, median PFS 3.1-7 months), compared with the impressive efficacy of CAR-T cell treatment in other lymphatic cancers (165,169,171,174). One possible explanation for the unexpectedly low efficacy or CAR-T cell treatment in CLL is T cell exhaustion. In CLL patients, T cells express markers associated with T cell exhaustion, and the CAR-T cells generated from these source cells may have an impaired ability to kill malignant cells (177). Interestingly, ibrutinib has been found to boost T cell numbers and function in CLL, possibly through offtarget effects on interleukin-2 inducible T cell kinase (ITK) or ZAP70, providing a rationale for concurrent treatment with CAR-T cells and ibrutinib (178). Indeed, Gauthier et al. treated 19 heavily-pretreated R/R CLL patients with anti-CD19 CAR-T cells and ibrutinib, achieving an ORR and CR of 83% and 22%, resulting in 1-year PFS and OS rate of 59% and 86%, respectively (172). An alternative approach to circumvent the problem posed by T cell exhaustion is using CAR-transduced natural killer (NK) cells. Allogeneic NK cells can be safely transfused irrespective of a full human leukocyte antigen (HLA) match, allowing the generation of an off-the-shelf, non-patientspecific CAR-construct generated from healthy donor cord blood. In a pilot study, Liu et al. treated 5 R/R CLL patients with anti-CD19 CAR-NK cells from HLA-mismatched donors (173). Of these five patients, three had a complete response, one had a partial response, and one patient did not respond and went on to receive a stem-cell transplantation. Although these results seem promising, both CAR-T cell therapy combined with ibrutinib treatment and CAR-NK cell therapy require more comprehensive evaluation in a larger cohort.

MEASURING PREDICTIVE BIOMARKERS IN CLINICAL CARE AND RESEARCH: AUTHOR RECOMMENDATIONS
In this paragraph, based on the evidence presented above and summarized in the provided tables, we outline recommendations regarding when to measure the previously discussed predictive biomarkers. The powerful predictive impact of TP53 aberrations is universally recognized and is presently incorporated in treatment decision algorithms in routine care. Consequently, TP53 status should be both cytogenetically and molecularly assessed in all clinical trials and in all CLL patients with an indication for treatment. Although ambiguity remains whether the IGHV mutational status should influence the choice of firstline therapy, it can differentiate between patients with potential long-term remission and patients with a risk of earlier relapse after a time-limited highly effective first-line treatment regimen such as FCR. As such, we recommend assessment of the IGHV mutational status in all clinical trials and in all CLL patients with active disease. Accumulating evidence suggests that CK/GC could function as a predictive marker and is associated with poorer prognosis after chemoimmunotherapy. Although routine measurement of CK/GC may be desirable in clinical care, more evidence is required of the impact of CK in the context of novel agents before it can be incorporated in therapeutic decisionmaking. Consequently, we strongly recommend to perform CBA or CMA in all clinical trials, especially in trials evaluating novel agents. Additionally, measurement by CBA/CMA will detect the presence of more classical cytogenetic abnormalities, including del(17p) and del(11q), although the predictive relevance of the latter has significantly diminished since the introduction of chemoimmunotherapy and novel agents. The proposed resistance of NOTCH1-mutated CLL to treatment with rituximab and ofatumumab is intriguing and requires further research, especially focusing on the predictive impact of NOTCH1 mutations in the context of regimens containing obininuzumab or ublituximab. Although the incorporation of NOTCH1-mutations in clinical care is not yet warranted, we recommend the assessment of NOTCH1 mutational status in all trials evaluating regimens that include anti-CD20 mAbs. The predictive impact of BCR stereotypy and BIRC3 mutations is currently unclear. As such, we do not recommend their routine assessment in clinical research, nor in patient care. As the prevalence of individual stereotyped subsets and BIRC3 mutations is relatively low, the predictive impact of these biomarkers should be assessed either retrospectively in several pooled trials, or prospectively in specialized trials that specifically recruit patients with the molecular features of interest.

CONCLUDING REMARKS
In this review, we have discussed treatment approaches to CLL with high-risk molecular features, providing a comprehensive overview of trials on this topic. Catalyzed by the advent of more advanced molecular techniques, our understanding of the pathophysiology of CLL has deepened over the years, leading to the identification of several cytogenetic, immunogenetic and molecular features that can differentiate between patients with low-and high-risk disease. Some of these, most notably TP53 aberrations, have clear predictive impact and are presently incorporated in decision algorithms in routine care. Their presence is strongly associated with inferior response to chemoimmunotherapy and necessitates the use of novel agents. The predictive capability of other molecular features is less clear, especially in the context of novel-based treatment. The predictive importance of del(11q) has significantly diminished since the advent of chemoimmunotherapy and novel agents and is redundant in therapeutic decision-making. Despite accumulating evidence of a predictive impact from clinical trials, the lack of consistent reporting and standardization prohibits the current use of CK/GC in therapeutic decisionmaking. In fit patients, stratification by IGHV mutational status identifies patients who benefit markedly from treatment with first-line FCR, but whether U-CLL always warrants first-line treatment with novel agents remains controversial. Interestingly, some aberrations may be predictive in certain specific contexts only, such as the proposed resistance of NOTCH1-mutated CLL to treatment with rituximab and ofatumumab, but this observation requires further validation before NOTCH1 status should be used to guide treatment choice. The data concerning the predictive impact of BCR stereotypy and BIRC3 mutations are currently immature, and treatment choice should not dependent on the presence of these features. The place of second-generation novel agents and cellular immunotherapy in the treatment of CLL with high risk features is still elusive, but forthcoming data from early-stage trials is promising, necessitating further study. Of note, the vast majority of data concerning the predictive impact of the biomarkers discussed above has been obtained through prespecified or post-hoc subgroup analysis. While informative, these trials have not necessarily been powered to answer such questions, especially in the case of rare features. As such, there is an unmet need for randomized trials that evaluate the efficacy of treatments, especially of novel agent-based regimens, in cohorts of patients with pre-specified high-risk molecular features, to move further towards patient-tailored treatment strategies.

AUTHOR CONTRIBUTIONS
LvdS and PJH performed the collection and interpretations of all relevant literature. LvdS and PJH write the manuscript. APK, AWL and M-DL critically read and revised the manuscript. All authors contributed to the article and approved the submitted version.