The biology behind PI3K inhibition in chronic lymphocytic leukaemia

Abstract

Phosphoinositide 3′-kinase (PI3K) is a key component of both chronic active and tonic B-cell receptor-signalling pathways. As such, PI3K inhibitors have emerged as promising therapeutic agents for diverse lymphoid malignancies, particularly chronic lymphocytic leukaemia. Multiple in vitro experiments and clinical trials have shown efficacy of these agents across all prognostic subgroups with a favourable toxicity profile. Moreover, in vitro studies suggest that combinations with monoclonal antibodies and/or other immune strategies could enhance the effect of PI3K inhibition.

Introduction

Treatment for patients with chronic lymphocytic leukaemia (CLL) has remarkably changed in the last few years. For decades, it was based on alkylating agents and similar drugs with modest efficacy and no impact on survival. The introduction of fludarabine and other purine analogues lead on to a significant improvement [Keating et al. 1991; Rai et al. 2000], which was further substantiated with the addition of monoclonal antibodies, particularly those targeting CD20 [Hallek et al. 2010]. Currently, the outlook of patients with CLL is becoming brighter with the advent of modern targeted therapies, including second-generation monoclonal antibodies [Wierda et al. 2010; Goede et al. 2014], immune modulators [James et al. 2014], Bruton tyrosine kinase (BTK) inhibitors [Byrd et al. 2013; Byrd et al. 2014], phosphoinositide 3′-kinase (PI3K) inhibitors [Furman et al. 2014] and BCL2 antagonists [Ng and Davids, 2014]. In this manuscript, we will review the scientific basis for the use of PI3K inhibitors as treatment for CLL.

The role of the B-cell receptor in B-cell development and function

We have known for some time that the B-cell receptor (BCR) is the key molecule for B-cell survival and proliferation. As such, an early event during haematopoietic stem cell commitment to the B-cell lineage involves rearrangement of the gene encoding the immunoglobulin heavy chain, which is then synthesized and expressed on the plasma membrane [Cambier et al. 2007]. This, together with the immunoglobulin light chain, Igα and Igβ (also known as CD79a and CD79b), are the components of the BCR [Rickert, 2013]. Several seminal experiments performed in last decades formally proved that BCR signalling is not only important for B-cell maturation and differentiation [Lam et al. 1997], but also for B-cell maintenance in the absence of the antigen exposure, since in vivo ablation of the surface immunoglobulin and other BCR components leads to apoptosis of already mature B cells. For example, knocking out the phosphorylation motifs of the Igα molecule in mice eradicates mature B cells within 2 weeks [Kraus et al. 2004].

The BCR signalling pathway in normal B cells

When the BCR encounters its specific antigen, the tyrosine residues located in CD79a and CD79b are phosphorylated by LYN and SYK [Yamamoto et al. 1993; Rolli et al. 2002], and this phosphorylation recruits the signalosome, a set of several kinases and adaptor proteins such as BTK and BLNK [Kabak et al. 2002]. This signal is then propagated via multiple pathways, notably phospholipase C-γ2 (PLC-γ2), PI3K and BTK [Takata et al. 1994; Hashimoto et al. 1999]. PLC-γ2 displaces phosphatidylinositol 4,5-bisphophate (PIP2) from the plasma membrane and produces diacylglycerol (DAG), which activates protein kinase C (PKC)-β, and inositol-triphosphate (IP3), thus increasing calcium influx to the cytoplasm [Dal Porto et al. 2004]. PI3K, on the other hand, phosphorylates PIP2 to create phosphatidylinositol 3,4,5-triphosphate (PIP3), which recruits BTK and other kinases maintaining BCR activation [Deane and Fruman, 2004]. Calcium influx activates several transcription factors such as nuclear factor kappa B (NF-κB), Jun and nuclear factor of activated T cells (NFAT), which lead to cytokine production, B-cell proliferation, class switching and B-cell survival [Dolmetsch et al. 1997]. PKC-β, on the other hand, activates MAP kinase and CARD11, a key adaptor molecule that coordinates the activation of NF-κB together with BCL10 and MALT1, in what is called the CBM complex [Shinohara et al. 2005]. PIP3, the product of PI3K, recruits BTK and AKT, thus activating a number of signalling effectors, most notably mTOR [Bellacosa et al. 1998]. So, the net result of BCR signalling is the activation of NF-κB, PI3K, MAPK, NFAT and many others [Dal Porto et al. 2004].

Apart from this active, antigen-dependent form of BCR signalling, mature B cells use the BCR in a different manner known as ‘tonic’ BCR signalling. Such tonic signalling is likely to be antigen independent as it affects all B cells, regardless of their immunoglobulin variable region. Moreover, a very elegant study published in 2009 proved that PI3K is the key component of tonic BCR signalling [Srinivasan et al. 2009]. In this study, the authors performed in vivo BCR ablation, which lead to rapid death of mature B cells as previously reported by the same research group, but also showed that constitutive PI3K activation prevented these cells from dying. In contrast, other components of the BCR signalling pathway, such as constitutive activation of the canonical NF-κB pathway were ineffective for salvaging these cells from apoptosis [Srinivasan et al. 2009]. In summary, PI3K signalling is not only an important component of chronic active BCR signalling pathway, but also the key component of tonic BCR signalling pathway.

The BCR signalling pathway in neoplastic B cells

Most B-cell lymphoid malignancies maintain surface expression of the BCR, which is surprising since their immunoglobulin loci are often disrupted by translocations, inversions, etc. [Kuppers, 2005; Nussenzweig and Nussenzweig, 2010]. However, lymphomas retain one intact set of IgH and IgL alleles, allowing the lymphoid neoplasm to form a BCR. This fact suggests that malignant B cells clearly benefit from the proliferation and survival signals triggered by the BCR. Furthermore, many lymphoid malignancies, even those derived from the germinal centre, use IgM constant regions to form the BCR [Vaandrager et al. 1998]. This is also surprising because normal B cells usually change their class from IgM to IgG when they pass through the germinal centre (Figure 1), and the reason is that IgM-BCR signalling promotes survival and proliferation, whereas IgG-BCR signalling favours plasmacytic differentiation [Horikawa et al. 2007]. As an example, in follicular lymphoma, the productive IgH allele is never translocated to BCL2 and remains as IgM, while the nonproductive allele is translocated and undergoes class switch recombination to IgG, demonstrating the selective pressure to maintain a productive IgM expression on the surface of follicular lymphoma cells [Staudt, 2007; Young and Staudt, 2013].

Figure 1.

Immunoglobulin structure and class switching. Class switching is the substitution of an expressed heavy-chain constant-region gene (mu constant region for IgM or delta constant region for IgD) by another downstream constant-region gene (gamma for IgG, alpha for IgA and epsilon for IgE), allowing the synthesis of its corresponding antibody. This process involves the deletion of the upstream constant-region genes and leads to the switch of the effector function of one type of antibody (IgM or IgD) for another (IgG, IgA or IgE) without changing the antibody’s specificity.

An example of a lymphoid malignancy that depends mostly on chronic active BCR signalling is the activated B-cell type of diffuse large B-cell lymphoma (ABC-DLBCL) [Alizadeh et al. 2000]. These tumours mostly rely on the NF-κB pathway, which, as already mentioned, is activated by the CARD11/BCL10/MALT1 complex [Lenz et al. 2007]. As such, activating mutations of CARD11 and other molecules further upstream, such as CD79b and CD79a are observed in a third of the patients with ABC-DLBCL, and the presence or absence of these mutations determine the sensitivity of these tumours to ibrutinib [Davis et al. 2010], a drug that blocks specifically the chronic active BCR signalling pathway (see below). In contrast, Burkitt’s lymphoma and germinal centre B-cell-type diffuse large B-cell lymphoma (GCB-DLBCL) rely mostly on tonic BCR signalling, and we know this because these tumours do not need CARD11 or BTK for survival, which results in a lower response rate to ibrutinib [Younes et al. 2014]. In these malignancies, BCR signalling activates PI3K and AKT, while mTOR inhibition effectively kills cell lines derived from Burkitt’s lymphoma [Schmitz et al. 2012].

The BCR signalling pathway in CLL

CLL is the most common leukaemia in the Western countries and has several features that make it an ideal target for BCR signalling inhibition. First, the immunoglobulin repertoire of CLL is biased or skewed, meaning that the proportion of the different immunoglobulin families is different to what we would expect in normal B cells [Fais et al. 1998]. Moreover, almost a third of CLL patients have distinctive, almost identical, antigen-binding sites among unrelated cases, also known as stereotyped immunoglobulin genes, something that would be very unlikely by chance alone [Agathangelidis et al. 2012]. These two features point towards a possible role for specific antigens in the pathogenesis of CLL. However, and unlike DLBCL, whole genome and exome sequencing analyses have failed to identify genomic aberrations affecting the BCR signalling pathway [Puente et al. 2011; Quesada et al. 2012]. The third important feature of CLL is the fact that the mutational status of the rearranged IGHV genes directly correlates with patient outcome [Damle et al. 1999; Hamblin et al. 1999] so that the disease can be divided into two subtypes: a more indolent subtype with a mutated variable region, more dependent on tonic BCR signalling, and a more aggressive subtype with unmutated variable region, which is more dependent on chronic active BCR signalling [Packham et al. 2014].

If BCR signalling is so important for the biology of CLL cells, where does it occur anatomically? This is important because we know that different anatomic compartments are histologically different in CLL [Matutes et al. 2010]. As such, when the CLL infiltrates the lymph nodes (LN) or the spleen we observe not only the typical small mature lymphocytes, but also collections of larger cells forming proliferation centres. These proliferation centres or ‘pseudofollicles’ are far more frequent in LN or spleen than in the bone marrow or peripheral blood [Matutes et al. 2010]. It was not until many years later, though, that investigators from the NIH proved that CLL cells from different anatomic sites have different gene signatures, with the LN being the key site for CLL pathogenesis [Herishanu et al. 2011]. In this study, CLL cells from the lymph nodes generally demonstrated upregulation of BCR and NF-κB pathways, whereas peripheral blood cells from the same patients did not have the same upregulation [Herishanu et al. 2011]. Moreover, the proliferation activity of CLL cells was also different across these anatomic sites, and correlated with the prognosis of patients, which has also been observed by ourselves and others [Giné et al. 2010; Falchi et al. 2014]. In our study, we found that the size of the proliferation centres, the number of mitoses and the proliferation rate all predicted patient survival [Giné et al. 2010]. Another study also evaluating the histological characteristics of CLL patients found that the proliferation activity in the LN not only predicted for patient survival, but was also highly correlated with positron emission tomography (PET) positivity, a test that is less invasive than a LN biopsy [Falchi et al. 2014].

The role of PI3K in cancer

As shown in Figure 2, the PI3K-AKT-mTOR pathway controls most hallmarks of cancer, including cell cycle, survival, metabolism, motility and genomic instability [Hanahan and Weinberg, 2000; Fruman and Rommel, 2014]. Of all classes of PI3K enzymes, only the four class I isoforms are clearly implicated in cancer: PI3Kα is a frequent genetic driver [Vadas et al. 2011], while PI3Kβ is usually implicated in tumours with loss of PTEN [Jia et al. 2008], an enzyme that does the opposite as PI3K (it dephosphorylates PIP3). PI3Kδ is almost specific of leukocytes and has a fundamental role in the survival of normal B cells, whereas PI3Kγ is also found in leukocytes, but it is mostly involved in inflammation [Baracho et al. 2011]. The PI3K pathway is so important that it is the most frequently altered pathway in human tumours: the gene encoding PI3Kα is the second most frequently mutated oncogene [Lui et al. 2013], and PTEN is among the most frequently mutated tumour suppressor genes [Wee et al. 2008].

Figure 2.

The role of PI3K in the biology of B cells. The PI3K/AKT/mTOR pathway controls most hallmarks of cancer, including cell cycle, survival, metabolism, motility and genomic instability. Abbreviations: IL-6, interleukin 6; BCR, B-cell receptor; CXCR5, CXC chemokine receptor type 5; JAK, Janus kinase; LYN, Lck/Yes novel tyrosine kinase; SYK, spleen tyrosine kinase; BTK, Bruton tyrosine kinase; PI3K, phosphoinositide 3′-kinase; NF-κB, nuclear factor κB, PKC, protein kinase C; PLC-γ2, phospholipase C-γ2; mTOR, mammalian target of rapamycin.

If we focus now on the two isoforms most commonly expressed in lymphocytes, PI3Kδ is a second messenger of many cell receptors, including the BCR, CD40, IL-6 and CXCR5, playing a crucial role in the survival, proliferation, chemokine secretion, motility and adhesion to endothelial and stromal cells (Figure 2) [Marshall et al. 2000; Davies et al. 2004; Bilancio et al. 2006; Henley et al. 2008]. In addition, PI3Kγ is also important in supporting the growth and survival of lymphoid malignancies, particularly in response to chemokines [Liu et al. 2007; Konrad et al. 2008]. Since PI3Kδ has a unique role in the migration, proliferation, survival and differentiation of B cells, this pathway is frequently affected in CLL and non-Hodgkin lymphoma (NHL), even though genome and exome sequencing studies have consistently failed to detect mutations in patients with these diseases. In contrast, hyperactivity of the different elements of the pathway is very common in patients with lymphoid malignancies [Herman et al. 2010].

With all of these facts in mind, targeting the PI3K pathway is a rational approach in patients with cancer, but more specifically in patients with lymphoid malignancies since two class I isoforms are relatively specific of lymphoid cells. The three main approaches tested so far are: (1) mTOR inhibitors; (2) pan-PI3K inhibitors; and (3) selective PI3K inhibitors (Figure 3).

Figure 3.

Signalling of the PI3K/AKT/mTOR pathway and relevant drugs that target each of the components of the pathway. PI3K generates the lipid messenger PIP3, which mediates activation of several protein kinases, including AKT. AKT stimulates glycolysis and drives tumour cells to consume glucose and also promotes survival and cell-cycle progression. Protein synthesis, cell growth and proliferation, and metabolic functions downstream of AKT are regulated by the mTORC1 axis. Negative regulation of this pathway is conferred by PTEN, which cleave a phosphate group in PIP3. Abbreviations: GFR, growth factor receptor; EGFR, epidermal growth factor receptor; HER2, human epidermal growth factor receptor 2; BCR, B-cell receptor; mTORC, mammalian target of rapamycin complex; PIP2, phosphoinositide (4,5)-biphosphate; PIP3, phosphoinositide (3,4,5)-triphosphate; PTEN, phosphatase and tensin homologue; PI3K, phosphoinositide 3′-kinase.

mTOR inhibitors (rapalogues) have been available for quite some time, and are effective in some neoplasms, such as renal cell carcinoma [Hudes et al. 2007], mantle-cell lymphoma [Hess et al. 2009] and breast carcinoma (in combination with anti-oestrogen treatment) [Baselga et al. 2012], but most single-agent trials have not demonstrated a therapeutic benefit, and these agents are not devoid of toxicity. In particular, a pilot trial evaluating the use of everolimus in patients with CLL was prematurely stopped because of toxicity although some degree of efficacy was also observed [Decker et al. 2009].

Pan-PI3K inhibitors also make sense, because most cancer cells overexpress multiple PI3K isoforms [Foukas et al. 2010]. However, the doses needed to block all class I isoforms for extended periods of time might not be tolerable since PI3K is important for many cellular functions, including glucose metabolism, cardiovascular system, platelet function, fertility and bone formation [Rodon et al. 2013]. For instance, one of the most common side effects of pan-PI3K inhibitors is hyperglycaemia [Busaidy et al. 2012]. In contrast, PI3K inhibitors that are selective for the δ, with or without the γ isoform, are truly effective in lymphoid malignancies in general, and in CLL in particular.

The use of selective PI3K inhibitors in CLL

The PI3Kδ inhibitor that is more advanced in its clinical development is idelalisib, formerly known as GS-1101 or CAL-101. It is a selective inhibitor of the δ isoform that promotes apoptosis in primary CLL cells ex vivo regardless of well-established prognostic factors such as 17p or 11q deletions. Moreover, it is effective in CLL cells with mutated IGHV genes, the subtype that is mostly dependent on tonic BCR signalling, as well as in CLL cells with unmutated IGHV genes, the subtype that relies primarily on chronic active BCR signalling pathways [Herman et al. 2010]. In addition, idelalisib does not promote apoptosis in normal T or natural killer (NK) cells and it does not impair antibody-dependent cell-mediated cytotoxity (ADCC) [Herman et al. 2010], which is a key mechanism of action of monoclonal antibodies such as rituximab and obinutuzumab [Herter et al. 2013].

A phase I trial was performed in 54 patients with relapsed or refractory CLL, 24% of them with either 17p deletion of TP53 mutations. Idelalisib achieved a 39% overall response rate according to International Workshop and Chronic Lymphocytic Leukaemia (IWCLL) criteria, although 81% of patients benefited from treatment in terms of LN reduction [Brown et al. 2014]. The reason for this discrepancy lies on the fact that patients receiving this drug commonly experience a LN reduction together with a progressive lymphocytosis that usually lasts for many months. Since lymphocytosis is historically considered a sign of progressive CLL by the IWCLL guidelines [Hallek et al. 2008], it is important to not categorize these patients as such, particularly because their symptoms improve and their lymphadenopathy or splenomegaly also improve. This discrepancy has prompted the creation of a new category, called partial response with lymphocytosis to account with this new clinical situation [Cheson et al. 2012].

This lymphocytosis can be considered as a class effect since it also happens in patients receiving BTK or SYK inhibitors. Recent in vitro data have shown that idelalisib reduces the adhesion of CLL cells to endothelial and marrow stromal cells, and this effect is particularly evident in those CLL cells with a high expression of VLA-4, also known as CD49d [Fiorcari et al. 2013]. This is particularly interesting because a high CD49d expression is associated with a more aggressive form of the disease [Bulian et al. 2014]. As a result of this lymphocyte redistribution, idelalisib is thought to overcome the protection mediated by endothelial and marrow stromal cells, rendering the malignant cells more vulnerable to cell death.

These responses were durable considering how refractory these patients were, with a median progression-free survival of almost 16 months and a 75% overall survival at 3 years [Brown et al. 2014]. The safety profile of the drug was relatively favourable, and the majority of adverse events were grade 1 or 2. The most common serious adverse events were pneumonia in 20% and neutropenic fever in 11%, which may have been in part due to prior chemotherapy. Less frequent side effects, although probably drug-related, were pneumonitis, increased transaminases and diarrhoea [Brown et al. 2014]. Idelalisib-induced colitis can be severe and a cause for drug discontinuation, as seen in clinical trials and murine models [Furman et al. 2014; Steinbach et al. 2014].

This phase I study was quickly followed by a variety of trials, including this double-blind phase III trial, also in patients with relapsed-refractory CLL [Furman et al. 2014]. A total of 220 patients were recruited and one half were allocated to the idelalisib + rituximab arm while the other half received placebo + rituximab. Median age was over 70 years, 85% of patients had a comorbidity score higher than 6 and over 40% of patients had either 17p deletion or TP53 mutations. The overall response rate was significantly higher in the idelalisib group (77% versus 15% in the second interim analysis), which translated into a significantly prolonged progression-free and overall survival [Furman et al. 2014]. These results led to FDA and EMA approval of idelalisib for patients with relapsed/refractory CLL in combination with rituximab.

Furthermore, the fact that patients from this trial achieved a significantly higher response rate compared to the previous trial, even considering that the percentage of patients with TP53 disruption was higher, points to a synergistic effect between idelalisib and rituximab. In addition, the beneficial effect of idelalisib was observed across all prognostic subgroups, including patients with 17p deletion, TP53 mutations and both mutated and unmutated IGHV genes, highlighting the importance of PI3K signalling in both CLL subtypes [Furman et al. 2014]. This contrasts with the results obtained with the BTK inhibitor ibrutinib, where patients with unmutated IGHV genes have a much faster response compared with patients with mutated IGHV genes [Byrd et al. 2013], perhaps because CLL cells with mutated IGHV genes are more dependent on tonic BCR signals, and the role of BTK is less clear in this pathway.

The second PI3K inhibitor currently in development is duvelisib (IPI-145), a drug that blocks the δ and γ isoform of PI3K. A phase I trial performed in patients with CLL was presented in December 2013. It included patients with relapsed/refractory disease but also a small cohort of elderly patients with previously untreated disease. Over 50% of patients had TP53 disruption and a small group of patients had already received BTK inhibitors. The response rate was 47%, with no significant differences between patients with and without TP53 disruption [Flinn et al. 2013]. The most common severe adverse event was neutropenia, but this complication was probably associated with prior therapy and advanced CLL since it was rarely seen in patients with therapy-naïve disease [Flinn et al. 2013]. Finally, a third PI3Kδ inhibitor called AMG319 is also in development, and first-in-human results were also presented last year at ASH [Lanasa et al. 2013]. Results of these and other clinical trials evaluating PI3K inhibitors in CLL are summarized in Table 1.

Table 1.

Results of clinical trials evaluating PI3K inhibitors in patients with chronic lymphocytic leukaemia.

Phase	N	Therapy	Disease status	ORR	Median PFS	Reference
I	54	Idela	Relapsed/refractory	72%	15.8 months	Brown et al. [2014]
I	52	Idela + R/B/RB	Relapsed/refractory	83%	28 months	Barrientos et al. [2013a]
I	40	Idela + R/O	Relapsed/refractory	83%	19 months	Furman et al. [2013]
I	45	Idela + B/F/Clb	Relapsed/refractory	78%	28.5 months	De Vos et al. [2013]
I	29	Idela + RB/ClbR	Relapsed/refractory	87–93%	Not reached	Barrientos et al. [2013b]
II	64	Idela + R	Therapy naïve	97%	93% at 24 months	O’Brien et al. [2013]
III	110	Idela + R	Relapsed/refractory	81%	93% at 24 weeks	Furman et al. [2014]
I	47	Duvelisib	Relapsed/refractory	47%	75% at 12 months	Flinn et al. [2013]

Abbreviations: N, number of patients; ORR, overall response rate; PFS, progression-free survival; Idela, idelalisib; R, rituximab; B, bendamustine; O, ofatumumab; F, fludarabine; Clb, chlorambucil.

Combinations of PI3K inhibitors with other agents

In vitro studies suggest that PI3K inhibitors do not impair NK-mediated ADCC and, therefore, are ideal partners for monoclonal antibodies such as rituximab or obinutuzumab. In contrast, the BTK inhibitor ibrutinib also blocks other kinases, such as interleukin-2-induced T-cell kinase (ITK), that are required for ADCC [Dubovsky et al. 2013; Kohrt et al. 2014]. Indeed, even though there is no phase III trial to formally prove this statement, the results obtained with idelalisib + rituximab [Furman et al. 2014] appear significantly better compared with those obtained with idelalisib monotherapy [Brown et al. 2014], whereas ibrutinib does not appear to benefit clearly from the addition of rituximab [Byrd et al. 2013; Burger et al. 2014]. Finally, in the phase III trial previously mentioned, rituximab-induced infusion reactions were significantly reduced in patients who also received idelalisib, and this clearly enhances the tolerability of the combination [Furman et al. 2014]. Combined treatment with idelalisib and otlertuzumab (an anti-CD37 therapeutic protein) has also demonstrated synergy in vitro, providing a rationale for future clinical trials [Lapalombella et al. 2012].

The opposite is true for the potential combination with lenalidomide, an immune modulator with significant activity in CLL [James et al. 2014]. This drug increases costimulatory molecule expression, CLL cell activation as well as vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) gene expression, all of which are blocked by PI3K inhibition. Consequently, the combination of idelalisib and lenalidomide does not appear synergistic, at least from a theoretical point of view [Herman et al. 2011]. In contrast, ibrutinib and lenalidomide potentially have complementary immune-modulating potential and their combination could be beneficial [Dubovsky et al. 2013].

Finally, a recent study has shown that mice with constitutive PI3Kδ inactivation have fewer tumours compared with wild-type mice because they have a reduced number of regulatory T cells compared with wild-type mice, thus breaking immune tolerance [Ali et al. 2014]. Consequently, PI3Kδ inactivation could be potentially used to enhance the efficacy of other immune strategies, such as cancer vaccines, adoptive cell therapy, PD-1 blockade and even PD-L1 inhibition [Porter et al. 2011; Westin et al. 2014]

Conclusion

PI3Kδ is a key component of both chronic active and tonic BCR signalling pathways in lymphoid malignancies. As such, PI3Kδ inhibition represents a promising strategy for treating diverse haematological and nonhaematological malignancies through several mechanisms of action. So far, results have been impressive in patients with CLL with a favourable toxicity profile. Several lines of evidence suggest that the effect of PI3Kδ inhibition could be enhanced by combination with monoclonal antibodies and other immune strategies.