The Evolving Use of PI3 Kinase Inhibitors for the Treatment of Chronic Lymphocytic Leukemia

Abstract

B cells express four PI3K isoforms (p110α, p110β, p110γ, and p110δ) but have a unique dependence on p110δ for survival. The design of isoform-selective inhibitors is possible, and pharmacologic inhibition of p110δ is toxic to neoplastic CLL cells for both cell-intrinsic and cell-extrinsic reasons. Idelalisib is a first-in-class p110δ inhibitor that exhibits efficacy for the treatment of relapsed CLL irrespective of adverse prognostic features. Toxicities of presumed autoimmune origin (hepatitis, enteritis/colitis, and pneumonitis), opportunistic infections, and bacterial infections were more pronounced in treatment naïve patients and when the drug was given in combination with chemotherapy, limiting further development. Duvelisib is a p110γ/δ inhibitor with a similar efficacy and safety profile to idelalisib. Recent data indicate that umbralisib, a p110δ/CK-1ε dual inhibitor, is safe and effective when administered to patients with CLL in either the treatment naïve or relapsed/refractory setting. Future studies are needed to establish where umbralisib should be incorporated into the treatment of CLL, what drug classes are most effective when combined with umbralisib and why umbralisib’s safety profile is superior. Whether other PI3K inhibitors in development can similarly retain efficacy while mitigating toxicity remains to be determined; most are exploring intermittent dosing schedules for that purpose.

Introduction

A remarkable evolution in the approach to the treatment of chronic lymphocytic leukemia (CLL) has occurred over the past decade. Cytotoxic chemoimmunotherapy, established as the gold standard based on multiple trials from the late 1990s to the early 2010s,^1-5 has been challenged – and in certain cases supplanted – by small molecule inhibitors of key signaling pathways within CLL cells, i.e. targeted therapies.⁶ The most commonly used of these targeted therapies fall into three classes: Bruton Tyrosine Kinase (BTK) inhibitors, B-cell lymphoma 2 (BCL2) inhibitors, and phosphatidyl-inositol-3-kinase (PI3K) inhibitors. The story of drug development within each of these classes is itself also a story of evolution, as various challenges specific to each class have had to be overcome to ensure their safe and effective administration to patients. And arguably in no class is this story of evolution more interesting and more informative than in the development of PI3K inhibitors. In this review, we will take a chronological approach to follow the ups and downs of PI3K inhibitor development in CLL over time, an origin story which ranges from initial excitement, to challenging toxicities, to more recent data causing a resurgence of interest in this drug class.

Molecular Mechanisms of PI3 Kinase Signaling

Phosphatidylinositol is a membrane-bound lipid molecule that serves as a signaling intermediate, transducing extracellular signals into intracellular responses. Upon stimulation of cell surface receptors, most notably the B cell receptor (BCR), multiple phosphorylation events occur, leading to activation of PI3 kinase. This activation event occurs on a molecular level when the regulatory PI3K subunit dissociates from the catalytic PI3K subunit. There are four different isoforms of the catalytic subunit, denoted p110α, p110β, p110γ, and p110δ.⁷ The enzymatic reaction catalyzed by the p110 subunit is the addition of a phosphate group to phosphoinositide 4,5-diphosphosphate to create phosphoinositide-3,4,5-triphosphate (PIP₃). PIP₃ is a lipid second messenger that is recognized by other proteins in the cell, such as BTK.⁸ These proteins are thus recruited to the inner leaflet of the plasma membrane and can serve as platforms for activation of additional downstream pathways.

The exact cellular effect of PI3K activation depends on many factors. The dynamic characteristics of the activating stimulus, such as amplitude, duration, and frequency, can affect output from the PI3K pathway.⁹ The spectrum of catalytic subunits present within the cell (each isoform has a specific expression pattern across the various tissues of the body) also affects output.¹⁰ Finally, the range of downstream effectors present that recognize the PIP₃ moiety can affect the outcome of PI3K activation.¹¹ This same complexity applies not just to PI3K activation, but also to pharmacologic PI3K inhibition. So, for example, intermittent versus tonic PI3K inhibition may have differing efficacies and/or different toxicities. Targeting a different spectrum of isoforms also produces different cellular, and clinical, effects. These observations have been exploited in the development of PI3K inhibitors that target different combinations of the p110 isoforms (Figure 1), and also by using various pharmacokinetic strategies to dose these inhibitors, as detailed below.

Figure 1:

Profiles of the isoform-specific PI3K inhibitors discussed in the text, and the toxicities expected to result from inhibition of each isoform.

Physiologic Role of PI3 Kinase Signaling in B Lymphocytes and Rationale for Targeting PI3 Kinase Enzymes in CLL

Expression of the p110α and p110β isoforms is universal across different tissue types, while expression of p110γ and p110δ is primarily restricted to leukocytes.^7,12 Thus, B lymphocytes express all four isoforms of the p110 catalytic subunits at points during their lifetime.^13-15 Studies of mice with genetic ablation of each of the p110 isoforms provide insight into the role of each isoform in the life of a B cell, and thus also may predict the phenotype one could expect with pharmacologic isoform-specific inhibition. For example, deletion of p110β has no discernible effect on murine B cell development or function,¹³ in keeping with the fact that p110β is the least-expressed of the four isoforms in B cells when comparing relative mRNA levels.¹⁴ While knockout of p110γ leads to altered T cell, neutrophil, and macrophage function,^16,17 there is no effect on B cell development.¹⁸ One study has reported that co-ablation of p110γ and p110δ leads to blockade of development at the pre-pro-B cell stage, and mature B cells that do develop in this model have impaired long-term survival.¹⁵ Deletion of p110α alone has no effect on B cell development, but deletion of both p110α and p110δ leads to a block in development at the pre-B-cell stage.¹³ Finally, catalytic inactivation of p110δ alone leads to impaired numbers of three types of mature B cells (follicular [B2] B cells, marginal zone B cells, and peritoneal [B1] B cells).^19,20 Inactivation of p110δ also causes impaired antigen-dependent signaling from the BCR, which results in reduced immunoglobulin levels and humoral responses to antigens in these mice.²⁰ These observations have led to the hypothesis that p110δ is responsible for antigen-dependent signaling from the BCR that is necessary for the survival of mature B cells, while either p110α or p110δ can communicate the tonic signal from the pre-BCR that is necessary for B cell development. Ultimately, these murine genetic studies demonstrate that (1) the isoforms are not redundant, meaning that genetic or pharmacologic inhibition of a single isoform, or a combination of isoforms, can have unique effects on B cell development and function, and (2) p110δ plays a central role in mature B cell function and survival.

With regard to pathophysiologic PI3K signaling in chronic lymphocytic leukemia cells, this has been reviewed in detail elsewhere,¹¹ but a few key points should be highlighted. In CLL cells, PI3K shows heightened activity compared to normal B cells, as measured by rates of conversion of PIP₂ to PIP₃.²¹ Correspondingly, p110δ is abundantly expressed in neoplastic lymphocytes from CLL patients.²¹ What leads to this increased flux through the PI3K pathway, putatively mediated by p110δ? There are at least two hypotheses. First, CLL cells have chronic activation of the BCR, either antigen-induced (by receptors that recognize available auto-antigens) or ligand-independent (by receptors that can signal autonomously).²² Second, chemokines -- provided in a paracrine fashion by the microenvironment and in an autocrine fashion by the CLL cells themselves -- lead to activation of cell surface receptors and subsequent PI3K signaling within the CLL cells (Figure 1).²³ Notably, somatic mutations that activate the PI3K pathway are not commonly seen in CLL. For example, activating mutations in PIK3CA (the gene encoding p110α) and inactivating mutations in PTEN (a negative regulator of the PI3K pathway) are extremely rare in CLL.^24,25 This follows a pattern previously established in CLL where key signal transducers (e.g. BTK and BCL2) are activated by mechanisms other than direct mutation. Thus, despite the absence of frequent oncogenic mutations in the PI3K pathway, the p110δ isoform in particular is an attractive pharmacologic target because it is both hyperactive in CLL cells and necessary for mature B cell survival.

A New Beginning - Early Success in Relapsed Disease

In 2014, the first-in-class p110δ inhibitor idelalisib (CAL-101, GS-1101) was FDA approved for the treatment of relapsed CLL. The road to this landmark accomplishment began almost a decade earlier with the discovery that isoform-specific PI3K inhibition was possible.²⁶ The Shokat lab screened a diverse panel of inhibitors of the PI3K family and found that various compounds inhibited each of the isoforms to varying degrees, despite the nearly identical sequence of the active site of p110α, p110δ, and p110β. This was due to the fact that certain inhibitors induce a conformational rearrangement of the active site and surrounding sequence upon binding, and this conformational rearrangement can only be accommodated if the (poorly conserved) surrounding sequence is flexible.²⁶ Thus, the drug must fit into the active site of the protein, which in turn induces a shift in protein conformation, and this new conformation must be able to accommodate the side chains of the drug. Indeed, the crystal structure of idelalisib bound to p110δ shows that the drug induces a conformational change in areas away from the ATP binding site that do not share sequence identity with other members of the PI3K family.²⁷ Multiple small molecule inhibitors with varying degrees of specificity for members of the PI3K family have now been identified (Table 1).

Table 1:

IC₅₀ Values (nM) of Discussed PI3K Inhibitors

Drug Name	p110α	p110β	p110δ	p110γ	Reference
Idelalisib	820	565	2.5	89	28
Duvelisib	1602	85	2.5	27	40
Zandelisib	>5491	335	0.6	2533	65
Parsaclisib	>20000	>20000	1	>20000	69
Copanlisib	0.5	3.7	0.7	6.4	71
Umbralisib	>10000	1116	22	1065	74

Lannutti et al. initially reported the structure of idelalisib and early in vitro experiments with the drug.²⁸ They used idelalisib to demonstrate that the increased PI3K pathway activity in malignant B cells is largely p110δ-dependent, as chemical inhibition of p110δ decreased phospho-Akt levels (a readout of PI3K pathway activity). Other groups subsequently demonstrated that idelalisib treatment inhibits chemotaxis, blocks BCR-mediated survival signals, and downregulates cytokine secretion of primary CLL cells in culture, ultimately promoting CLL cell apoptosis.^21,29 These findings provided strong preclinical rationale for phase 1 studies of idelalisib in both relapsed/refractory CLL³⁰ and other indolent non-Hodgkin lymphomas (iNHLs).^31,32 Consistent with preclinical studies, the phase 1 study in CLL demonstrated in vivo pharmacodynamic effectiveness of idelalisib at doses greater than or equal to 100mg twice daily (150mg twice daily was ultimately chosen as the recommended phase 2 dose), with reduction in phospho-Akt levels in circulating neoplastic CLL cells after only one week on therapy. Even in this heavily pre-treated population (median number of prior therapies was 5), there were early signs of durable efficacy, as the overall response rate was 72% (39% partial responses, 33% partial responses with lymphocytosis) and the median PFS was 15.8 months. Notable toxicities included opportunistic infections (two cases of Pneumocystis jirovecii pneumonia, one case of cytomegalovirus reactivation, and two cases of fungal pneumonia). The most frequent adverse events were neutropenia (57% any grade, 43% grade ≥3), anemia (37% any grade, 11% grade ≥3), and fatigue (32% any grade, 2% grade ≥3). Transaminitis elevation of any grade was observed in 28% of subjects, but grade ≥3 elevations were rare (2%).

The promising results of these early phase studies eventually led to a large phase 3 registration trial in relapsed, refractory CLL^33,34 and a single-arm phase 2 trial in other relapsed indolent non-Hodgkin lymphomas (including SLL).³⁵ The phase 3 trial in CLL randomized 220 patients to either idelalisib plus rituximab versus rituximab plus placebo, with the primary endpoint of progression-free survival (PFS). Patients in the rituximab arm could cross over to the idelalisib-containing arm at progression. The trial was stopped early due to overwhelming efficacy of the idelalisib-containing arm. At the final analysis performed after 18 months of follow-up, the PFS in the idelalisib-rituximab arm was 19.4 months compared to 6.5 months in the rituximab monotherapy arm, with an overall response rate (ORR) of 86% in the idelalisib-treated arm.³⁴ Patients with high risk disease features such as TP53 aberrations or IGHV unmutated status had similar PFS to those without such features. Almost all responses were partial responses. Despite the opportunity for crossover, there was a trend toward improved median overall survival (OS) in the idelalisib plus rituximab group (40.6 months versus 34.6 months with rituximab monotherapy); this trend became significant in patients with TP53 aberrant disease. Patients on the idelalisib-rituximab arm experienced rapid improvement in quality-of-life assessments.³⁶ Adverse events were consistent with what was observed in the phase 1 trial, and included neutropenia (65%, 42% grade ≥ 3), transaminitis (ALT elevated 39%, grade ≥ 3 ALT elevation in 12%), and diarrhea or colitis in 36% of patients (13% grade ≥3). Opportunistic infections were again seen; all cases of P. jirovecii occurred in patients who had not received prophylaxis.

The above findings were re-capitulated in a similar, separate phase 3 trial that compared idelalisib plus the anti-CD20 monoclonal antibody ofatumumab to ofatumumab alone for the treatment of relapsed, refractory CLL.³⁷ Again, the primary endpoint of PFS was superior with idelalisib-containing combination therapy compared to ofatumumab alone, 16.4 months versus 8.0 months, respectively. Again, similar PFS was seen regardless of the prognostic markers of the underlying disease, although the study was underpowered to identify subgroup-specific benefits. And again, neutropenia, diarrhea, and transaminase elevations were more common with idelalisib plus ofatumumab versus ofatumumab alone.

There are important caveats to these large phase 3 trials. First, the comparator arm of anti-CD20 monotherapy is now considered obsolete, particularly since most enrolled patients (~90%) in these trials had previously progressed on rituximab-containing therapy. Comparison of PI3K inhibitors in the relapsed setting to other novel agents would come much later with the ASCEND trial.³⁸ Additionally, because these trials enrolled patients largely before ibrutinib and venetoclax were approved, these populations had chemo-resistant disease but not BTK-inhibitor-resistant or BCL2-inhibitor-resistant disease. The efficacy of p110δ after disease progression on other targeted therapies, a more relevant question in today’s environment, is not addressed by these trials. Finally, because these studies did not contain idelalisib monotherapy arms, the benefit of a co-administered anti-CD20 antibody is unclear. Similar durable responses were also seen in the SLL subset of patients enrolled in the iNHL trial of idelalisib monotherapy,³⁹ and duvelisib was effective without a CD20 antibody (see below), suggesting that anti-CD20 agents are not necessary for PI3K inhibitors to be effective. Despite these caveats, the impressive efficacy of idelalisib in a hard-to-treat CLL patient population clearly established it as the first in a new class of drugs to treat CLL. The combination of idelalisib plus rituximab was approved by the U.S. Food and Drug Administration (FDA) to treat relapsed CLL on July 23, 2014 and by the European Medicines Agency (EMA) on September 18, 2014.

The efficacy of idelalisib encouraged development of other PI3K inhibitors, and duvelisib (IPI-145) was the next to be FDA-approved. Duvelisib was developed as an oral, dual p110δ and p110γ inhibitor.⁴⁰ In vitro studies demonstrated that p110γ was important for chemotaxis of supportive stromal cells to lymphoid tissues containing CLL cells.⁴¹ By blocking both p110δ and p110γ, duvelisib was uniquely positioned to target both intracellular and extracellular survival signals,⁴¹ perhaps increasing efficacy compared to p110δ inhibition alone. The initial phase 1 study of duvelisib examined safety and early efficacy in a wide variety of hematologic malignancies, including 55 patients (26% of the total study population) with relapsed CLL/SLL.^42,43 Doses from 8mg twice daily to 100mg twice daily were explored. Including all tested dose levels, adverse events were similar to those reported for idelalisib, with neutropenia (39% any grade, 20% grade ≥ 3), elevated transaminases (ALT increase 39% any grade, 20% grade ≥ 3), and diarrhea (42% any grade, 11% grade ≥ 3) particularly frequent. Two cases of P. jirovecii pneumonia were seen, and when the study protocol was amended to mandate prophylaxis against opportunistic infections, no further P. jirovecii cases occurred. The maximum tolerated dose was 75mg twice daily (due to rash and transaminitis at 100mg twice daily), and based on pharmacodynamic data showing similar efficacy at the lower dose of 25mg twice daily, this latter dose was chosen for later stage development in B cell malignancies. Early efficacy data in CLL were promising, with a 56% ORR and a median PFS of 15.7 months across all dose levels.⁴³

Duvelisib was then advanced to a single arm phase 2 study in relapsed iNHL⁴⁴ and a randomized controlled phase 3 trial in relapsed CLL, the DUO trial.^45,46 In DUO, patients with relapsed CLL were randomized to duvelisib or ofatumumab. The primary endpoint was met, with a statistically superior PFS (as assessed by an independent review committee) in the duvelisib arm (13.3 months) versus ofatumumab (9.9 months). Investigator-assessed PFS was also superior for duvelisib versus ofatumumab, 17.6 months versus 9.7 months, respectively. The ORR was 74% to duvelisib and 45% to ofatumumab. As with idelalisib, PFS and ORR were similar in TP53-aberrant and TP53-intact disease. There was a significant decrease in circulating cytokines thought to support the tumor microenvironment, suggesting that the goal of dual p110δ/γ inhibition was achieved. While adverse events such as neutropenia (33% any grade, 30% grade ≥ 3), diarrhea (51% any grade, 15% grade ≥ 3), transaminitis (3% grade ≥3), and opportunistic infections were seen, the patients in DUO benefited from prior experience with idelalisib. Infectious prophylaxis was mandated and the only cases of P. jirovecii pneumonia occurred in subjects not receiving it. The DUO protocol also recommended duvelisib holds and dose reductions for early signs of diarrhea, pneumonia, or transaminitis. These interventions increased tolerability and did not seem to impact efficacy,⁴⁷ providing an early hint that regular dose interruptions could be exploited as a way to avoid toxicities of PI3K inhibitors. As with idelalisib, patients on this trial had progressed on chemoimmunotherapy and not on targeted therapies, and the comparator arm of single agent ofatumumab would now be considered insufficient. While cross-trial comparison is difficult, the response rate and PFS did not greatly differ from idelalisib, calling into question whether the added p110γ inhibition significantly contributed to efficacy. Nevertheless, this trial demonstrated that a second PI3K inhibitor could elicit a durable response in a majority of relapsed CLL patients, including those with adverse prognostic features, and duvelisib was approved as monotherapy for relapsed CLL by the FDA on September 24, 2018; it has not been approved by the EMA for CLL.

The Mid-life Crisis – Challenges with Toxicities Limit Clinical Use and Development

With success in the relapsed/refractory setting, development of idelalisib and duvelisib in CLL then focused on (1) establishing the efficacy of these agents in the front-line setting and (2) establishing the efficacy of these agents as part of combination regimens. The initial enthusiasm surrounding these agents was tempered as these efforts to expand their use were limited by toxicities. These toxicities fell into two broad categories: immune-mediated toxicities and infectious toxicities.

Three common categories of immune-mediated toxicities that have been reported in trials of idelalisib and duvelisib are transaminitis, diarrhea/colitis/enteritis, and pneumonitis. Preclinical data first suggested that these toxicities were immune-mediated and also hinted at a mechanism –toxic effects of p110δ inhibition on regulatory T cells (Tregs). Compared to other types of T cells, Tregs uniquely require p110δ for normal immunosuppressive function.⁴⁸ Genetic loss of p110δ⁴⁹ or pharmacologic inhibition of p110δ in vitro with idelalisib⁵⁰ is specifically toxic to Tregs (as compared to CD4+ and CD8+ T cells). On a systemic level, the net effect of p110δ inhibition is pro-inflammatory, as mice with genetic inactivation of p110δ develop an autoimmune colitis characterized by activated intraepithelial lymphocytes^19,51 and are also protected from tumorigenesis due to a potent anti-tumor immune response.⁴⁸ The clinical finding that transaminitis, colitis, and pneumonitis were steroid-responsive further supported an autoimmune origin.^52-55 Studies of PI3K inhibitors in the front-line setting were particularly informative. For example, a phase 2 study of idelalisib plus ofatumumab in previously untreated CLL reported an extremely high rate of transaminitis (63% any grade, 52% grade ≥ 3).^52,53 Liver biopsies of patients with active hepatitis on this study showed an activated CD8+ T cell infiltrate, mirroring similar findings that had been previously reported in patients with idelalisib-related colitis.^56,57 The transaminitis responded to steroids and subjects were more likely to tolerate resumption of idelalisib if they were on steroids at the time of drug re-initiation. Other studies replicated this finding of higher rates of transaminitis in the treatment naïve setting. For example, when idelalisib was combined with rituximab for previously untreated TP53-aberrant CLL, there was a 41% rate of grade ≥3 transaminitis,⁵⁸ and when duvelisib was combined with FCR for young patients with treatment naïve CLL, there was a 28% rate of grade ≥ 3 transaminitis.⁵⁹ Finally, in a comprehensive analysis of 853 CLL patients treated across multiple phase 2 and 3 studies of idelalisib in the treatment naïve and relapsed setting, only previously untreated disease and younger age (as stratified by decile, <55, ≥55 to <65, ≥65 to <75, and ≥75 years of age) were identified on multivariate analysis as predictors of grade ≥ 3 transaminase elevation on drug.⁶⁰ All of these observations are consistent with an autoimmune origin for these adverse events, as patients who are less heavily pretreated and who are younger have a more intact immune system that can mount a robust autoimmune response. Thus, despite the fact that these small trials did demonstrate clinical efficacy of PI3K inhibitors in treatment-naïve CLL,^52,59 the risk-benefit ratio is unfavorable in this population compared to relapsed patients, and development of idelalisib and duvelisib in this space stalled.

Infectious complications had been seen in early trials of the PI3K inhibitors, but their frequency and severity became more apparent as experience with the drugs grew and trial populations became larger and less heavily pretreated. Infections included opportunistic infections, likely attributable to effects of PI3K inhibition on both the adaptive and innate immune system, and bacterial infections, likely attributable to effects on the innate immune system and treatment-related neutropenia. Opportunistic infections can be mitigated by rigorous use of prophylactic antibiotics and surveillance for fungal and viral infections. The second category of infections, however, was not easily remedied and was further exacerbated by combination with chemoimmunotherapy. For example, a large phase 3 trial randomized 416 patients with relapsed CLL to either ibrutinib plus bendamustine-rituximab (BR) or BR alone.⁶¹ While addition of idelalisib improved the primary endpoint of PFS (median 20.8 months with idelalisib-BR versus 11.1 months with BR alone) and ORR (70% vs. 45%, respectively), this benefit came at a cost. There were more serious adverse events in the idelalisib-BR group (68% of patients versus 44% of patients with BR alone), and these events were primarily infection-related: febrile neutropenia (20% versus 5%), pneumonia (14% versus 7%), pyrexia (12% versus 5%), and sepsis (5% versus 1%). Cytomegalovirus reactivation was seen in 6% of patients in the idelalisib-containing arm versus 1% in the BR arm. Even more troubling, treatment-emergent adverse events leading to death occurred in 11% of patients in the idelalisib-BR group compared to 7% of patients in the BR group. In early 2016, an internal analysis of six trials of idelalisib in combination with other anti-neoplastic agents in non-approved front-line or second-line indications was conducted by Gilead Pharmaceuticals, the manufacturer of idelalisib. While never published, this analysis identified an increased rate of serious infectious complications and death in the idelalisib-containing arms in these trials. All ongoing clinical trials examining idelalisib for the front-line treatment of CLL and iNHL were immediately stopped.

A string of challenging results with idelalisib was capped by the ASCEND trial.³⁸ This large phase 3 study in 398 patients with relapsed CLL compared acalabrutinib to the investigator’s choice of idelalisib plus rituximab or BR. This was the first trial in CLL to pit two novel agents against one another, although it was not powered for direct comparison of idelalisib-rituximab to acalabrutinib. Acalabrutinib bested the investigator’s choice control arm; after a median follow-up of 16.1 months, the median PFS in the acalabrutinib arm was not reached, while the median PFS in the control arm was 16.5 months. When broken down by therapy received, idelalisib-rituximab and BR behaved similarly in this study, with a median PFS of 15.8 months and 16.9 months, respectively. Intriguingly, the ORR to acalabrutinib and the control arm were similar, 81% vs 75%, implying that differences in durability of response were responsible for the differences seen in PFS. The high rate of treatment discontinuation in patients that received idelalisib-rituximab may have limited their exposure to the drug and thus potentially impacted the durability of any response. Supporting this hypothesis, 47% percent of patients receiving idelalisib-rituximab discontinued due to adverse events (most commonly diarrhea), compared to only 11% of patients in the acalabrutinib arm and 17% of patients receiving BR. The median duration of exposure to idelalisib-rituximab was only 11.5 months.

The median duration of exposure to acalabrutinib was 15.7 months, closely approximating the median follow-up on the study of 16.1 months, suggesting that most patients on acalabrutinib continued on the drug at time of publication. Infections were also a frequent reason for idelalisib discontinuation; the rate of grade ≥ 3 infection in the idelalisib-rituximab group was 28% compared to 11% in the BR group and 15% in the acalabrutinib arm. On subgroup analysis, the favorable hazard ratio for acalabrutinib shifted closer to 1 in patients over age 65 and patients who had received ≥ 4 lines of therapy, exactly the groups who are expected to have fewer adverse effects from idelalisib and thus better tolerability.

In summary, this era in the development of PI3K inhibitors was defined by a growing appreciation of their autoimmune and infectious toxicities. At the same time, when patients could be maintained on the drugs, efficacy was encouraging. Efforts were made to identify ways to retain the efficacy of this class in CLL while reducing toxicities, leading to the third act in the development of PI3K inhibitors.

Resurrection - Recent Approaches to Retain Efficacy While Mitigating Toxicity

One challenge to maintain efficacy while reducing toxicity is that, at least for the autoimmune toxicities, the mechanism of toxicity is likely due to on-target inhibition of p110δ. However, since the cellular contexts in which these effects are happening are different (a neoplastic cell versus a Treg), a pharmacokinetics-driven approach to p110δ inhibition may be able to elicit effects in one cell type but not the other. For example, if p110δ inhibition for a sustained time is required to block Treg function, while brief p110δ inhibition still successfully blocks neoplastic cell proliferation, then intermittent dosing could retain efficacy while reducing toxicity. Alternatively, Tregs in humans recover to normal levels about 2 weeks after treatment with a Treg-specific cytotoxic therapy,⁶² meaning that even if Treg loss cannot be circumvented, regular treatment breaks of two weeks or longer will allow for recovery of Tregs and potential avoidance of p110δ-specific immune-mediated toxicities. Clinical experience with idelalisib and duvelisib demonstrated that brief drug holds could allow resolution of toxicities without significantly impacting efficacy.^47,63 Many efforts at development of the next generation of PI3K inhibitors have therefore focused on intermittent dosing schedules.

The TEMPO trial is a phase 2 study exploring two intermittent dosing strategies for duvelisib in iNHL, including SLL.⁶⁴ In arm 1, patients receive an up-front “loading” dose of duvelisib at 25mg twice daily for 10 continuous weeks. This is followed by 4-week cycles consisting of 2 weeks off, 2 weeks on duvelisib 25mg twice daily. In arm 2, the loading dose is omitted and subjects receive 25mg twice daily for two weeks out of every four week cycle. A phase 2 trial specifically focusing on patients with relapsed CLL is exploring a strategy where duvelisib is given in an induction phase for 3 months at 25mg twice daily, and then reduced to a maintenance phase where the drug is taken only two out of every seven days at 25mg twice daily (NCT03961672).

Zandelisib (ME-401) is an oral, once daily p110δ inhibitor.⁶⁵ In the initial design of its phase 1 study in follicular lymphoma and CLL, the drug was administered in a continuous dosing schedule, but the rate of grade ≥ 3 adverse events was high and the protocol was modified to explore the efficacy of intermittent dosing.^66,67 When the drug was given for two continuous 28-day cycles followed indefinitely by cycles of 1 week on, 3 weeks off drug, tolerability improved. After a median follow-up of 9.7 months, 3.5% of patients experienced grade ≥ 3 diarrhea and 2% of patients experienced transaminitis. The ORR for the 10 relapsed CLL patients enrolled on the trial was 100% in the continuous dosing group and 83% in the intermittent dosing group, supporting the conclusion that efficacy is maintained with intermittent dosing. Based on this, the design of the phase 2 TIDAL study of zandelisib in iNHL was amended – instead of randomizing patients to continuous versus intermittent dosing, all subjects will receive the drug intermittently.⁶⁸ Studies of the drug in CLL in combination with rituximab or zanubrutinib (NCT02914938) are ongoing.

Parsaclisib (INCB50465) is a potent, specific p110δ-specific inhibitor that has taken a similar path to development.⁶⁹ The phase 1/2 dose-escalation and dose-expansion study had multiple cohorts that explored parsaclisib monotherapy and combination therapy in a wide variety of B cell malignancies.⁷⁰ An initial approach at continuous dosing was limited by the adverse events of diarrhea/colitis and rash, and a modified dosing regimen of 20mg once daily for 9 weeks followed by 20mg once weekly improved tolerability. No patients that received this latter dosing schedule had to discontinue due to treatment-emergent adverse events. Six patients with CLL received parsaclisib monotherapy in this study, and their ORR was 33%.

Copanlisib (BAY 80-6946), an inhibitor of p110α, p110β, p110γ, and p110δ (see Table 1) that is FDA-approved for relapsed follicular lymphoma after at least two prior systemic therapies, also utilizes an intermittent dosing schedule.⁷¹ The drug is given intravenously on days 1, 8, and 15 of a 28-day treatment cycle. The inhibition of multiple PI3K isoforms, the intravenous route of administration (avoiding direct exposure to the gastrointestinal tract and first-pass metabolism in the liver), and the intermittent dosing schedule may all contribute to the unique adverse effect profile of this drug compared to other PI3K inhibitors. After two years of follow-up in patients who received the drug for indolent lymphomas, the most common adverse events included hyperglycemia (50% any grade, 40% grade ≥ 3), diarrhea (35% any grade, 9% grade ≥ 3), and hypertension (30% any grade, 24% grade ≥ 3), and neutropenia (29% any grade, 24% grade ≥ 3).⁷² Hyperglycemia in particular is thought to be due to p110α inhibition. The immune-mediated toxicities of pneumonitis (6% any grade, 1% grade ≥ 3) and transaminitis (ALT increased 4.2% any grade, 0.7% grade ≥ 3) were infrequent. Use of the drug in CLL/SLL has been limited, but it appears efficacious, with response rates in the 40-75% range as monotherapy in relapsed patients.^72,73 A study exploring its use to deepen responses to ibrutinib in relapsed CLL is planned (NCT04685915).

Another approach to mitigate autoimmune toxicities is to combine the PI3K inhibitor with a second entity that blocks cellular pathways needed for the autoimmune response. This entity could be a second drug or could be a second enzymatic target of the PI3K inhibitor itself. This is one hypothesis for why the drug umbralisib (TGR-1202), a combination p110δ and casein kinase-1ε (CK-1ε) inhibitor, has a favorable toxicity profile compared to other drugs within this class. This drug was initially billed as a p110δ inhibitor,⁷⁴ although phenotypic differences from idelalisib in in vitro assays prompted a search for another kinase target. CK-1ε was identified as a second target based on kinome profiling data.⁷⁵ The most common toxicity in the phase 1 first-in-human study was diarrhea (40%) but it was rarely severe (grade ≥ 3 3%).⁷⁶ Transaminitis elevations were rare (4% any grade, 3% grade ≥ 3). This favorable toxicity profile was also seen in the phase 1 study of umbralisib in combination with the anti-CD20 monoclonal antibody ublituximab.⁷⁷ Neutropenia remained the most common hematologic toxicity (32% any grade, 28% grade ≥ 3) but infections, particularly opportunistic infections, were less common with umbralisib. There were no reports of P. jirovecii pneumonia or symptomatic CMV reactivation in either the umbralisib or umbralisib/ublituximab phase 1 studies, despite low rates of prophylaxis. In both trials, treatment holds because of adverse events were still common, required for 43% of patients on umbralisib monotherapy⁷⁶ and 39% of patients on umbralisib plus ublituximab.⁷⁷ The most common reasons for dose reductions were fatigue, neutropenia, and transaminitis, and diarrhea.

Based on these promising results in early phase studies, the phase 3 UNITY-CLL was launched. This four-arm study compared ublituximab, umbralisib, umblituximab plus umbralisib, and chlorambucil plus obinutuzumab in both treatment naïve and relapsed/refractory CLL patients. Results from the umbralisib plus ublituximab (U2) versus chlorambucil plus obinutuzumab arms (a total of 421 patients) were presented in abstract form in late 2020, representing the first time that any data from a randomized trial of PI3K inhibition in the CLL front-line setting has been publicly reported.⁷⁸ At a median follow-up of 36.2 months, the primary endpoint of the trial was met, as U2 compared to chlorambucil plus obinutuzumab had a superior PFS in both the previously untreated (38.5 months versus 26.1 months, respectively) and relapsed (19.5 months versus 12.5 months, respectively) CLL populations. The ORR was higher with the PI3K inhibitor as well, 84% versus 78% in the treatment naïve setting and 82% versus 57% in the relapsed setting. Subgroup analysis confirmed that the PFS benefit was present whether del(17p) was present or absent. The benefit of U2 over chlorambucil-obinutuzumab was greater in IGHV-unmutated compared to IGHV-mutated patients, a finding which has been seen in many studies comparing targeted therapies to chemoimmunotherapy. Perhaps the most interesting finding of the study was the tolerability of the U2 regimen. Median exposure to the U2 combination was a prolonged 21 months, but despite this, rates of grade ≥ 3 toxicities were lower than might be expected in a patient population composed of both treatment naïve and previously treated patients, although the types of adverse events seen were similar to other drugs within this class. Events of particular interest with respect to PI3K inhibition included diarrhea (56% any grade, 12% grade ≥ 3), neutropenia (34% any grade, 31% grade ≥ 3), ALT elevation (17% any grade, 8% grade ≥ 3), and pneumonitis (3% any grade 0.5%, grade ≥ 3).

UNITY-CLL breathes new life into the PI3K inhibitor class by demonstrating that a p110δ inhibitor can be safely administered to both previously untreated and relapsed CLL patients for a prolonged time in a multicenter setting. It also provides preliminary prospective evidence that PI3K inhibition can be effective in patients who progress on BTK inhibitors, as 15% (n=14) of the enrolled patients had done. This group had a 57% ORR compared to 82% in other previously treated patients. However, the study has significant limitations. In the front-line setting, indefinite therapy with U2 is unlikely to supplant fixed duration venetoclax-obinutuzumab^79,80 or indefinite therapy with a BTK inhibitor^81,82 as the treatment of choice in the majority of patients since these other treatments appear to have a longer PFS (Table 2). In the relapsed setting, chlorambucil-obinutuzumab was an unusual choice as a comparator arm; the combination is not FDA approved in this setting and in the era of targeted therapies, chemoimmunotherapy is rarely used as salvage therapy in patients who have relapsed after prior chemoimmunotherapy (as >85% of the previously treated patients in UNITY-CLL had). The question of whether an anti-CD20 monoclonal antibody adds to the effectiveness of a PI3K inhibitor also remains unresolved.

Table 2:

Randomized Phase 3 Trials in Untreated CLL with a Chlorambucil-Obinutuzumab Arm

Trial	Median Follow-up , months	Control Arm	Duration of Clb, months	Control Arm Median PFS, months	Comparator Arm	Comparator Arm Median PFS, months	Reference
UNITY-CLL, treatment naïve cohort	36.2	Obin + Clb	6	26.1	U2	38.5	78
CLL14	39.6	Obin + Clb	12	35.6	Obin + Ven	Not reached	79,80
iLLUMINATE	31.3	Obin + Clb	6	19.0	Obin + Ibr	Not reached	81
ELEVATE-TN	28.3	Obin + Clb	6	22.6	Obin + Acal	Not reached	82
CLL11	59.4	Obin + Clb	6	28.9	Ritux + Clb	15.7	2,5

Clb = chlorambucil, Obin = obinutuzumab, Ibr = ibrutinib, Acal = Acalabrutinib, Ven = venetoclax, U2 = umbralisib plus ublituximab, Ritux = rituximab

Perhaps the most interesting question that UNITY-CLL raises for the field is why the drug’s toxicity profile appears similar but significantly less severe than the PI3K inhibitors that have come before it. Multiple hypotheses have been put forward, but none is clearly convincing. As noted above, umbralisib also inhibits CK-1ε. CK-1ε silencing suppresses WNT pathway activity.⁸³ In Tregs, WNT pathway activation leads to decreased Treg function,⁸⁴ suggesting that this added feature of umbralisib may protect Tregs. Alternatively, umbralisib is less potent than idelalisib and duvelisib (K_d 6.2nM vs. 1.2 nM vs. 0.047 nM, respectively, and see Table 1 for IC₅₀ values),⁷⁶ and the drug may have found a “sweet spot” with enough p110δ inhibition for clinical effectiveness but not enough for toxicity. In keeping with either of these hypotheses, umbralisib treatment of Eμ-TCL1 mice does not cause as large a decrease in Treg number and function as idelalisib and duvelisib do.⁸⁵ Simply put, umbralisib “spares” Tregs when compared to idelalisib and duvelisib. Finally, perhaps the assumption that some of the toxicities shared by duvelisib and idelalisib are due to on-target p110δ inhibition is incorrect, and it is in fact the side effect profile of these older drugs that we do not completely understand. Arguing against this, however, are the similar side effect profiles of the newer, more potent, and more specific drugs in this class, zandelisib and parsaclisib. A better understanding of the reasons behind the differing side effect profiles of the various PI3K inhibitors is needed to aid drug development.

Future Directions

With the demonstration in UNITY-CLL that it is possible to safely administer a PI3K inhibitor for prolonged periods of time, combination therapy can be re-addressed. As with BTK inhibitors and venetoclax, the goal of combination therapy is to achieve deeper remissions and potentially time-limited regimens. Umbralisib and ibrutinib were safely combined at standard doses in a phase 1b study in relapsed CLL.⁸⁶ The rate of complete responses with combination therapy (29%) was higher than expected from ibrutinib alone (typically around 2% with similar length of follow-up), hinting that depth of response might be improved by combining the two. The addition of ibrutinib to U2 to create a triplet was also safe and all agents could be used at standard doses.⁸⁷ Triplets of pembrolizumab / umbralisib / ublituximab, (relapsed disease),⁸⁸ TG-1501 (anti-PD-L1 antibody) / umbralisib / ublituximab (relapsed, NCT02535286), venetoclax / umbralisib / ublituximab (relapsed),⁸⁹ TG-1701 (BTK inhibitor) / umbralisib / ublituximab (relapsed)⁹⁰ and acalabrutinib / umbralisib / ublituximab (treatment naïve and relapsed, NCT04624633) are all now being studied in CLL.

Combination therapy approaches can also be explored for duvelisib and idelalisib if the trials stay within populations where the drugs are known to be safe and effective, i.e. relapsed CLL. For example, a trial of duvelisib and venetoclax is ongoing in relapsed CLL and Richters syndrome, and early results are promising.⁹¹ No grade ≥ 3 transaminitis has been seen and diarrhea is manageable (50% any grade, 5% grade ≥ 3). The ORR in CLL is 94% and 56% of patients achieved undetectable minimal residual disease in the bone marrow, higher than expected from venetoclax monotherapy in this patient population, 32% of whom had progressed on prior BTK inhibitors.

Finally, PI3K inhibitors may combine particularly well with CAR T cells given the immunomodulatory effects of these drugs. Ex vivo treatment of T cells with duvelisib during CAR T manufacture reduced expression of exhaustion markers on the T cells and led to greater CAR T cell expansion and faster elimination of CLL cells when the CAR T cells were engrafted into a mouse model of CLL.⁹² In vitro and in vivo use of duvelisib decreased IL-6 levels secreted by immune cells, which may protect against cytokine release syndrome.⁹³

Future studies will also hopefully elucidate the mechanisms behind disease resistance to PI3K inhibitors. Identification in patients of drug-resistant mutations in p110δ would definitively confirm that the mechanism of action is on target, as has been seen for BTK inhibitors⁹⁴ and BCL2 inhibitors.⁹⁵ To date, however, resistance mutations in PIK3CD have not been described. Whole exome sequencing on a cohort of patients who had progressed on PI3K inhibitors identified activating mutations in MAP kinase pathway genes (e.g. MAP2K1, BRAF, and KRAS) more frequently in patients who did not respond to the drugs than in patients who did.⁹⁶ The abundance of these mutations also increased over time in patients taking PI3K inhibitors. Introduction of these mutations to CLL cell lines prevented idelalisib-mediated abrogation of MEK/ERK signaling. In a separate study, pharmacologic p110δ inhibition increased the expression of activation-induced cytidine deaminase (AID), a mutagenic enzyme.⁹⁷ This led to increased somatic hypermutation and more frequent chromosomal translocations, suggesting that long term treatment with PI3K inhibitors could lead to genetically unstable disease. However, this has yet to be validated in the clinic and likely requires a long duration of therapy. If anything, studies have shown lower than expected rates of progression with Richter’s transformation in subjects on PI3K inhibitors; the opposite would be predicted if the CLL cells were genetically unstable.

Summary

PI3K inhibitors are an effective treatment for CLL, but the toxicities of the drugs have limited their use to the relapsed setting. Even though idelalisib and duvelisib have been primarily studied in relapsed disease after chemoimmunotherapy, they are now most commonly employed after BTK inhibitors and BCL2 inhibitors because of the favorable toxicity profile of these more widely used agents. Alternatively, they are employed in patients who have an intolerance and/or contraindication to the other targeted therapies. Treating clinicians should use prophylaxis against P. jirovecii and herpesviruses when using duvelisib and idelalisib and monitor for neutropenia. If autoimmune toxicities are suspected, it is important to have a low threshold for drug holds, and once infection can be ruled out, for initiation of steroids. Recent data with umbralisib has reinvigorated interest in this drug class and will ultimately require a reassessment of how this class should be integrated into the treatment paradigm for CLL patients. This renewed sense of enthusiasm will also encourage further research to answer some of the most challenging questions in the field. Why do certain toxicities occur more with some PI3K inhibitors than others? Which therapies are best to use in combination with PI3K inhibitors? How does CLL become resistant to PI3K inhibitors? As this chronological review has detailed, however, this is not the first time in the field that challenging questions have been raised and – hopefully – will be answered.

Key Points:

1. Idelalisib and duvelisib are oral small molecule PI3K inhibitors that are highly effective for the treatment of patients with relapsed CLL, including high-risk disease.

2. The major toxicities seen with idelalisib and duvelisib include adverse events of presumed autoimmune origin (hepatitis, enteritis/colitis, and pneumonitis), opportunistic infections, and bacterial infections. These toxicities limited development and widespread adoption of the drugs.

3. Umbralisib is a combination p110δ/CK-1ε inhibitor that has a more favorable toxicity profile and efficacy in treatment naïve and relapsed/refractory CLL, requiring a reassessment of where this drug class stands in the CLL treatment paradigm. Intermittent dosing schedules under investigation may also allow for the safe and effective administration of other newer PI3K inhibitors.

Clinical Care Points.

• Duvelisib and idelalisib should be considered for the treatment of patients with relapsed CLL, particularly older patients and more heavily pretreated patients, who are at lower risk for toxicities from the drugs.

• Since duvelisib and idelalisib do not cause significant cardiotoxicity and are predominantly metabolized by the liver, they are also useful in patients who may have contraindications to BTK inhibitor use (e.g. cardiac comorbidities) or BCL2 inhibitor use (e.g. kidney disease).

• Prophylaxis against P. jirovecii and herpesviruses should be used when duvelisib and idelalisib are prescribed, and patients with a positive baseline CMV serology should be monitored for cytomegalovirus reactivation on a regular basis.

• Treating clinicians should have a low threshold to hold drug for toxicities. Steroids are also useful to treat autoimmune toxicities such as hepatitis, enteritis/colitis, and pneumonitis.