Duvelisib Eliminates CLL B Cells, Impairs CLL-Supporting Cells, and Overcomes Ibrutinib Resistance in Preclinical Models

Inhibitors of Bruton’s Tyrosine Kinase (BTKi) and phosphoinositide 3-kinase (PI3Ki) have signicantly improved therapy of chronic lymphocytic leukemia (CLL). However, the emergence of resistance to BTKi has introduced an unmet therapeutic need. Here we demonstrate in vitro and in vivo the essential roles of PI3K-δ for CLL B-cell survival and migration and of PI3K-γ in T-cell migration and macrophage polarization; and more ecacious inhibition in CLL-cell burden by dual inhibition of PI3K-δ,γ. We also report an ibrutinib-resistant CLL case, whose clone exhibited BTK and PLCγ2 mutations, responded immediately to single agent duvelisib with a redistribution lymphocytosis followed by a partial clinical remission associated with subsequent modulation of T and myeloid cells. CLL samples from patients progressed on ibrutinib were also responsive to duvelisib in patient-derived xenografts irrespective of BTK mutations. Our data support dual inhibition of PI3K-δ,γ as a valuable approach for therapeutic interventions, including patients refractory to BTKi.

Like other functional B cells, CLL cells are activated by signals delivered through multiple receptors, including the B cell receptor for antigen (BCR) and cytokine and chemokine receptors 5,7,8 .
Phosphatidylinositol 3-kinases (PI3K), which play signi cant roles in malignant B-cell growth 9,10 , signal by four PI3K isoforms (α, β, δ, γ), each with distinct functional roles. PI3K-δ and PI3K-γ are speci cally expressed in leukocytes 11 − 13 , with PI3K-δ to be critical throughout the B-cell lifecycle and essential for survival 11,14 ; in contrast, PI3K-γ signaling is utilized by non-B cells for chemokine receptor signaling, thereby controlling the migration and localization of T cells and the polarization of monocytes to immunosuppressive M2 macrophages 11,12,15,16 . These complementary roles of PI3K-δ and PI3K-γ in leukocytes suggest that dual blockade could provide enhanced therapeutic bene ts in CLL relative to inhibition of PI3K-δ alone 17 .
Duvelisib (IPI-145) is a rst-in-class, oral, dual PI3K-δ,γ inhibitor 13 that has been approved by FDA for the treatment of patients with refractory or relapsed (R/R) CLL or small lymphocytic lymphoma (SLL) after ≥ 2 prior therapies; it is also approved to treat patients with follicular lymphoma after ≥ 2 prior systemic therapies 18 . Duvelisib potently inhibits PI3K-δ and PI3K-γ in CLL B cells 13,19 , offers a novel approach to treat patients with CLL, potentially enhancing e cacy by targeting both CLL cells and CLL-supporting cells. In vitro, duvelisib sensitizes primary CLL cells to apoptosis and abrogates bone marrow (BM) stromal cell-mediated CLL-cell survival by inhibiting BCR-mediated signaling and chemotaxis 17,19 .
Of the several approved targeted agents in the CLL treatment armamentarium, the most widely used currently is ibrutinib, a rst-in-class Bruton's tyrosine kinase inhibitor (BTKi) approved for rst-line and relapsed-disease treatment in CLL and other B-cell malignancies 20 . BTKi is taken continuously until progression of disease or intolerance develops. Resistance to ibrutinib can occur by the development of mutations in the BTK-binding site, e.g., a BTK C481S, and/or in mediators of BCR-signaling downstream of BTK, e.g., PLCG2, which lead to expansion of resistant subclones [21][22][23] . Mutations in BTK and PLCγ2 are found in ~ 80% of CLL patients with acquired resistance to ibrutinib 24 . Upon progression in certain patients, therapy is immediately needed as the disease can progress rapidly. Little is known about the role of dual PI3K-δ,γ inhibition as a salvage approach for patients who develop these mutations.
Here we aimed to understand how duvelisib targets TME and treat CLL and BTK-resistant disease as shown in C481S mutant XLA B cell line 25 . We rst performed in vitro assays and utilized a patient-derived xenograft (PDX) model of CLL [26][27][28] to explore the mechanisms of action of duvelisib and to characterize the distinct functions of PI3K-δ and PI3K-γ in CLL B cells and CLL-supporting cells. We documented that the two PI3K isoforms played distinct roles in sustaining the survival of CLL cells and found in the PDX model that the inhibitory activity of duvelisib was effective against CLL B cells from patients at various diseases stages and from those whose disease had progressed while receiving ibrutinib, even in the setting of BTK and/or PLCG2 mutations. To substantiate these ndings, we also describe the clinical course of a patient who underwent a clinical remission with dual PI3K-δ,γ inhibition after disease progression on ibrutinib.
To evaluate the effects of PI3K-i on CLL B-cell survival in vitro, CLL B cells from four patients was incubated with either 0.1 to 1 µM PI3K-δi, PI3K-γi, or duvelisib and the absolute numbers of live CLL B cells were quanti ed after 96 hours. This revealed a signi cant reduction in viable CLL cells by duvelisib and PI3K-δi with IC50 values at 100nM and 1uM, both doses being achievable and tolerable in patients. However, PI3K-γi did not cause signi cant changes in CLL cell viability (Fig. 1A).
Finally, we assessed the effects of individual or combined PI3K-γi or PI3K-δi on CLL-cell migration in vivo. CLL cells were pre-incubated with 0.1 to 1µM of the PI3K-γi or PI3K-δi or duvelisib for 48 hours, and then injected intravenously (iv) into NSG mice. Twenty-four hours later, the numbers of CLL cells in murine spleens were quanti ed. Although the PI3K-γi did not signi cantly affect the migration of CLL B cells into the spleen relative to control, both the PI3K-δi and duvelisib signi cantly did so in a dose-dependent manner relative to PI3K-γi (Fig. 1C).
Overall, these data indicate the direct effects of PI3K-δi on CLL B-cell survival, proliferation, and migration.
Inhibition of PI3K-γ impairs the ability of T cells and macrophages in the TME to support CLL B-cell growth and survival. We next investigated the impact of the single versus the dual inhibitory agents on two key components of CLL TME, T cells and myeloid cells.
Next, migration of CLL-derived T cells was assessed in vivo. CLL-derived T cells were stimulated in vitro by anti-CD3/28 Dynabeads + IL-2 for 7 days in the presence of duvelisib, PI3K-γi or PI3K-δi, at doses ranging from 0-1 µM. At the end of culture, duvelisib, PI3K-γi or PI3K-δi treatment didn't change the growth of T-cells (Fig. S1) or surface levels of CXCR4, CXCR5, CCR6 and CCR7 (data not shown).
However, when equal number (5x10 6 ) of T cells from each culture condition were injected iv into NSG mice, T-cell migration to the spleen within 24 hours was blocked by PI3K-γi and duvelisib, but not by PI3Kδi pre-treatment. Notably duvelisib and PI3K-γi alone were signi cantly better than PI3K-δi alone, while PI3K-γi and duvelisib were not signi cantly different from each other (Fig. 2B). Thus, PI3K-γ function is critical for the migration of CLL-derived T cells.
In CLL, macrophages provide survival signals for the leukemic B cells through PI3K-dependent AKT activation 31,32 . Increasing the M1 to M2 ratio alters the ability of a CLL-cell line to engraft and grow in NSG mice 33 . Therefore, we investigated whether PI3K-γ inhibition impeded macrophage-facilitated CLL Bcell survival (Fig. 3C). Co-culturing CLL B cells with M2-polarized murine macrophages for 5 days increased CLL B-cell survival compared with cultures of CLL B cells alone, consistent with M2 macrophages playing a supportive role for CLL B cells (Fig. 3C). Notably, addition of duvelisib to the CLL B cell/M2 macrophage co-culture signi cantly reduced CLL B-cell survival at doses > 500nM (Fig. 3C).
In summary, these experiments indicate that inhibition of PI3K-γ alone or in combination with PI3K-δ, modulates the numbers and functions of leukemia-supporting T cells and macrophages in the tumor microenvironment.
Dual inhibition of PI3K-δ,γ reduces the numbers of CLL B cells, CLL-supporting T cells, and myeloid cells in vivo. To assess the impact of duvelisib on transplanted CLL B cells and CLL-supporting T cells and endogenous murine myeloid cells in lymphoid tissues, we used a CLL PDX model that allows CLL B-and T-cell engraftment and growth in recipient alymphoid mice [26][27][28] . Brie y, CLL patient PBMCs and patientderived T cells, previously activated in vitro, were injected retro-orbitally (ro) into NSG mice and allowed to engraft for 2 weeks prior to administering 3 weeks of duvelisib treatment (70 or 100 mg/kg via oral gavage once daily) (Fig. 4A). At a dose of 100 mg/kg, duvelisib signi cantly reduced splenic CLL B-cell numbers relative to vehicle control ( Next, we examined the distinct functions of PI3K-δ and PI3K-γ in CLL-cell engraftment and growth using the PDX model, but this time treating engrafted CLL cells and activated autologous T cells with PI3K-γispeci c (15 mg/kg) or PI3K-δi-speci c (10 mg/kg) agents, or duvelisib (100 mg/kg) once a day for 3 weeks. Both PI3K-δi-and duvelisib treatments signi cantly decreased the number of leukemic B cells from both U-CLL and M-CLL patients in murine spleens ( Fig. 5A and Fig. S3).
We then examined the distinct functions of PI3K-δ and PI3K-γ on murine splenic macrophage levels in the PDX model. Consistent with in vitro results, duvelisib and PI3K-γi monotherapy signi cantly reduced the number of murine M2 macrophages in the spleens after 3 weeks of treatment, whereas the PI3K-δi did not induce a signi cant change (Fig. 5C).
Because of the difference in IC 50 for duvelisib versus the isoform-speci c agents (Table S1), we also used the same patient's cells to test the combination of the PI3K-γi-and PI3K-δi-speci c agents. This indicated that in vivo treatment with the combination of PI3K-δ and PI3K-γ inhibitors led to a signi cant decrease in splenic CLL B cells compared with the PI3K-δi agent alone (Fig. 5D), supporting the advantage of dual PI3K-δ,γ inhibition compared with PI3K-δi treatment alone.
Case study of a CLL patient salvaged with duvelisib after progression on ibrutinib. Intolerance and resistance to ibrutinib are emerging therapeutic problems, with the former developing in 50 ~ 63% of patients 36 and the latter in 13 ~ 30% 37-39 , often by selection for intraclonal variants harboring BTK (C481S) or PLCG2 mutations 22 .
In this context, we report a patient (CLL 1570), who had failed treatment with udarabine, FCR ( udarabine, cyclophosphamide, rituximab), tonsillar radiation, and BR (Bendamustine and Rituximab), before achieving a partial response to ibrutinib. Six years later, the patient became refractory to ibrutinib, and near the time of ibrutinib relapse, the leukemic clone contained BTK C481S and PLCG2 L845F mutations, along with mutations in BRCA2 (truncation in intron 26); MAP2K1 (MEK1; Q56P, subclonal); SF3B1 (K700E); TP53 (F113C and L330R, splice site 559 + 1G > A). Immediately after discontinuing ibrutinib (D0, Fig. 6), the patient was started on duvelisib (25 mg orally, twice daily) and continued for > 266 days, with only 2 short (13 and 3 days duration) treatment holds; the rst, at day 50, due to grade 2 elevations of ALT/AST and low level cytomegalovirus reactivation, and the second, at day 179, because of a temporary elevation of serum creatinine.
Duvelisib therapy led to a favorable response, designated as a partial response according to the International Workshop on Chronic Lymphocytic Leukemia (iwCLL) criteria 40 , at day 27 on duvelisib treatment. While thrombocytopenia and elevated absolute lymphocyte counts had developed on ibrutinib, these measures normalized following duvelisib treatment, with absolute values of lymphocytes decreasing by ≥ 50% on days 62-76, 168-236, 266 and beyond (Fig. 6A). Since the patient rapidly developed a lymphocytosis (> 5 x 10 9 cells/L), the fall in CLL cells presumably occurred by leukemic B cells exiting tissue niches, including the BM which was the site of the most dramatic changes upon disease relapse. Speci cally, the CLL cells in the BM fell from 60% before initiating duvelisib to 10% after 7 months on treatment (Fig. 6B). Additionally, other immune populations involved in leukemic B-cell control were either reduced (M2-like macrophages), increased (M1-like macrophages), or unchanged (T cells) compared with baseline values, for > 150 days of duvelisib treatment (Fig. 6B).
CLL B cells from patients relapsing with ibrutinib are e ciently eliminated by duvelisib in a PDX model. Next, we attempted to model in vivo, using the PDX approach, the bene cial effects of duvelisib in this patient and in two others who had also developed ibrutinib-resistant disease. Like CLL1570 above, leukemic B cells from patient CLL1273 had at least a C481S BTK mutation; in contrast, neither BTK nor PLCG2 mutations were found in the leukemic cells from patient CLL1782.
NSG mice, engrafted with leukemic B and T cells from each patient, were treated with ibrutinib (25mg/kg) or duvelisib (100mg/kg) daily for 3 weeks. As expected, based on the patients' clinical histories, in vivo ibrutinib treatment did not signi cantly affect CLL B-cell numbers in the spleens of recipient mice (Fig. 7A,  7B). However, after treating mice engrafted with each ibrutinib-resistant patient with duvelisib, splenic CLL B-cell numbers were signi cantly reduced regardless of BTK mutation status. Additionally, duvelisib, but not ibrutinib, signi cantly reduced the numbers of Ki67 + CLL B cells (Fig. 7C) and murine M2 macrophages (Fig. 7D) in the spleens. Interestingly, in these animals, no difference in the number of total T, Th1, or Th2 cells was observed with either duvelisib or ibrutinib treatment (Fig. S3). Collectively, ndings in Figs. 6 and 7 indicate that the duvelisib responsiveness in ibrutinib-resistant patients, without or with BTKi resistance, is recapitulated in these PDX models. To our knowledge this is the rst in vivo demonstration of ibrutinib-refractory CLL that are treated with duvelisib in clinical practice and PDX models, where effective suppression of BTK-mutated leukemic clones is achieved.

Discussion
BTK inhibitors, the current mainstays of treatment for patients with CLL, are administered continuously until disease progression or intolerance occurs. Despite excellent responses and progression-free survival, relapses occur and tumor cell acquired mutations 41 . Clinical outcomes after ibrutinib relapse are poor, and such patients often progress rapidly 21,22,42 . Since signals from T cells and myeloid cells support the survival and growth of leukemic cells in tissue niches and can also thereby allow and advance disease progression and therapeutic refractoriness 43,44 , there is an ongoing need for therapies that circumvent BTKi drug resistance by directly targeting leukemic B cells directly and tumor-supportive cells, the latter thereby forestalling indirectly impeding leukemia-promoting pathways in the TME.
In this study, we dissected, in vitro and in vivo using a PDX model, the functions of dual inhibition of PI3Kδ and PI3K-γ by duvelisib in primary CLL B, T and myeloid cells with the aim of determining if blocking another BCR-mediated pathway in CLL B cells and other functionally-relevant pathways in T cell and monocytes/macrophages by dual PI3K inhibition would avert BTK resistance. Our ndings support previous observations of duvelisib's effects in B-cell malignancies 17,25,45 and provide novel understandings of the drug's mechanisms of action by characterizing the individual contributions of PI3K-δi and PI3K-γi, integral functions of duvelisib, in various cell types.
The four class I isoforms of PI3K -α, β, γ and δ -have overlapping, yet unique patterns of expression and biological functions. In contrast to the ubiquitously expressed PI3K-α and PI3K-β, the PI3K-γ and PI3K-δ isoforms are restricted to hematopoietic cells, with PI3K-δ contributing ~ 50% of the total PI3K activity in B lymphocytes and PI3K-γ being more functionally restricted to T and myeloid cells 46 . Here we show that duvelisib targets both malignant B cells and the TME, consistent with achieving clinical improvement and impinging on trophic changes in the TME of R/R CLL patients 47,48 .
For CLL B cells, we found that duvelisib (0.1-1 µM) inhibits PI3K-δ signaling, thereby reducing leukemic B-cell proliferation and survival in vitro. Also, in vivo using a PDX model, duvelisib inhibited the migration of CLL B cells to the spleen through PI3K-δi, resulting in decreased leukemic B-cell numbers and indicating that preventing CLL cells from entering survival niches has a bene cial therapeutic effect. Moreover, since PI3K-δ modulates the IL-4R-induced phosphorylation of stromal cells which normally promotes B-cell adhesion and growth 49,50 , this TME effect likely affected the survival of those leukemic B cells that did reach the spleen and other solid lymphoid tissues.
Both PI3K-δi and PI3K-γi modulate cytokine synthesis in activated T cells 46,51 , and hence dual inhibition by duvelisib decreases production of IL-2, IFNγ, and TNFα 25 . However, in vitro, we did not nd a reduction in activated T-cell numbers with PI3K-δi nor PI3K-γi after stimulation with anti-CD3/28 beads and IL-2, whereas in vivo we did observe that pre-treatment with duvelisib and PI3K-γi blocked migration of T cells. These ndings are compatible with PI3K-γi having minimal direct impact on CLL B-cell proliferative capacity, but ultimately leading to a reduction in T-cell numbers and ampli cation by reducing the ability of these cells to reach the spleen where growth can occur. Thus, like the direct tra cking inhibition of CLL B cells by PI3K-δi seen above, direct tra cking inhibition of T cells by PI3K-γi leads to a fall in T cells.
Since the majority of T cells that expand and support CLL B-cell growth in this PDX model are CD4 + their diminution indirectly leads to a reduction in CLL B cells by eliminating of their trophic effects. This was especially well exhibited in our PDX model that requires the support of T cells for CLL cell growth in tissue niches 52 . Thus, the enhanced treatment e cacy provided by PI3K-γi might be due to an effect on the TME, such as on autologous T cells which are essential for CLL-cell survival in patients 53 and in our PDX model 52 . Although PI3K-δi has weak effects on T cells, PI3K-γi had a signi cantly greater effect on T-cell migration, and subsequent engraftment and growth in tissue niches. Dual inhibition of PI3K-δ,γ by duvelisib mirrored these effects on T-cell functions.
Finally, macrophages, which can dampen immune responses by producing cytokines that suppress cytotoxic T cell activation and function, were altered through PI3K-γi, and not PI3K-δι. For example, the PI3K-γ inhibitor, IPI-549, reduced the immune suppressive function of M2 macrophages 15,16 , whereas the PI3K-δi idelalisib has minimal impact on macrophage functions in vitro 54 . Consistent with this, when we assessed the role of PI3K-δ and PI3K-γ in the endogenous macrophages of NSG mice in vitro and in vivo, duvelisib decreased M2 murine macrophage polarization and the support that such macrophages provided for CLL survival. These ndings are consistent with previous ndings in CLL 33 and T-cell lymphoma 55 . In addition, since T-cell subset distribution was changed in the TCL-1 murine model of CLL 33 after macrophage-targeted treatment, the decreased T-cell growth that we observed in the PDX experiments might be contributed by the impaired macrophage subsets post duvelisib and PI3K-γi treatments. We similarly found that M2 macrophages were reduced in patient the CLL1570 upon initiation of duvelisib treatment for ibrutinib resistance, consistent with loss of support for CLL cell survival and promotion of an immune-stimulatory environment that activates effector T cells 56 . Thus, the reduced numbers of CLL B cells after duvelisib treatment could result directly by PI3K-δi and indirectly by altering the supporting effects of T cells and macrophages via PI3K-γi.
Our PDX ndings, demonstrating that duvelisib inhibited and ibrutinib did not the growth of CLL B cells from patients whose disease had progressed during ibrutinib therapy, regardless of the detection of BTK and/or PLCG2 mutations, could have substantial clinical relevance. Notably, patient CLL1570, who also previously progressed on ibrutinib, achieved a partial clinical response on duvelisib. Conspicuously, since, in this patient, the majority of ibrutinib-resistant cells resided in the BM, the duvelisib-induced redistribution lymphocytosis likely affected these cells, potentially by inhibiting the adhesion of CLL cells to BM mesenchymal stromal cells which requires the action of PI3K-δ 49,50 . Finally, the patient's response to duvelisib was manifested by an altered myeloid-cell phenotype that could abrogate support for CLL cell growth.
Regarding T cells, our PDX ndings indicated that the reduction of T cells after duvelisib treatment could be due to the loss of B-CLL cells, impaired T-cell migration, and/or altered macrophage distributions. Differently, T-cell inhibition was not observed in mice receiving samples from patients with ibrutinibresistant disease, or clinically for patient CLL1570, who had progressed on ibrutinib. Since ibrutinib can affect T cell function [57][58][59][60]  T-cell recruitment and expansion, particularly within the CD8 + cytotoxic subpopulation. The effects on T cells in the latter context are thought to be mediated by reducing the immune suppressive function of myeloid cells, rather than a direct effect on T-cell proliferation or activation 16 . This underscores the importance of context and may reveal differences between the tumor microenvironment of hematologic malignancies and solid tumors.
Overall, our study provides the mechanistic rationale for using duvelisib in R/R CLL patients, although a larger number of patients will need to be studied to determine the scope of the clinical relevance of these initial observations. In this regard, although duvelisib was FDA-approved based on the phase 3 DUO trial 18 , patients who were previously exposed to ibrutinib or any other BTK inhibitors were excluded from that trial 48 , and hence data on post-BTK inhibitors are limited. Since duvelisib can cause serious toxicities (colitis, pneumonitis, transaminases, and severe infections) with the potential for rare fatal outcomes, further studies are warranted in patients refractory to BTKi and who are unable to receive or have failed BCL2 inhibition. To our knowledge, this is the rst pre-clinical study supporting the use of duvelisib in patients who have progressed or are refractory to ibrutinib 61 , a current unmet need, especially if such patients are not candidates for BCL2-inhibitor therapy In summary, by studying primary CLL PBMCs in vitro and in vivo, we have delineated the mechanistic contributions of the individual and dual inhibitions of PI3K-δ and PI3K-γ in CLL. The multiple mechanistic inhibitory actions of duvelisib, which directly and indirectly affect leukemic and ancillary cells involved in the disease, support its use in R/R CLL.

Materials And Methods
CLL patient samples. Studies were approved by the Institutional Review Board of Northwell Health.
Treatment-naïve and ibrutinib-resistant CLL subjects were selected using current iwCLL guidelines 40 . All patients provided written informed consent to participate. CLL clonal IGHV-IGHD-IGHJ sequences were determined by direct sequencing of PCR products from these samples as described 62 ; cases with germline identity ≥98% were considered U-CLL and those with germline identity < 98% were designated as M-CLL 62 . Targeted DNA sequencing for BTK C481S was performed in 2 CLL patients (CLL1782 and CLL1273) who exhibited clinical resistance to ibrutinib therapy. For a third ibrutinib-resistant patient (CLL1570), genomic pro ling was conducted from a BM biopsy within 1 month of discontinuing ibrutinib.
Therapeutic response was assessed according to the iwCLL criteria 40 . Response assessments included evaluation of disease-related constitutional symptoms; focused physical examinations; complete blood counts and differential white blood cell counts; and bone marrow biopsy and computed tomography scans, when indicated. CLL PDX model. Alymphoid NOD-scid IL-2Rγ null (NSG) mice were housed 5 or less per cage, allowed free access to food and water, and maintained on a 12:12 hour light/dark schedule. All experiments were conducted in accordance with the Institutional Animal Care and Use Committee (IACUC) guidelines.
In vitro CLL viability assessment. For viability assays, patient-derived CLL cells were incubated with PI3Ki (IPI-549) for 72 hours, then stained with the SYTOX Dead Cell Stains (Thermo sher) and analyzed by ow cytometry and FlowJo software (FlowJo, LLC). In the bone marrow-derived monocyte (BMDM)supported CLL-cell survival assay, CLL cells were added to the BMDMs at a ratio of 4:1 (monocytes to CLL cells); cultured 12 hours with M-CSF and IL-4, and then CLL-cell viability assessed via Guava ViaCount assay (MilliporeSigma) on days 0 and 9 of culture.
Percent inhibition of Ki-67 + pAKT + CLL cells was calculated by ow cytometry, and these values were used to generate EC 50 values using GraphPad Prism Software.
In vitro T-cell migration assay. CLL PBMCs were pre-treated with DMSO, duvelisib, PI3K-γi (IPI-549), or PI3K-δi (IPI-3063) for 24 hours and then allowed to migrate across a 6.5-mm polycarbonate insert with a 5-µM pore size in a Transwell (CoStar/Corning) plate in response to a 3-hour exposure at 37 C to CXCL12 (300 ng/ml; R & D systems). Cells were harvested from the lower chamber, and CD3 + CD4 + CD8 + cells quanti ed by ow cytometry. The percent inhibition of T-cell migration was calculated in response to each of the 3 PI3Ki used and compared to the DMSO control. EC 50 values were then calculated using GraphPad Prism Software.
In vivo CLL B-and T-cell migration assay in NSG mice. CLL PBMCs pre-treated with PI3Kis for 48 hours were injected into NSG recipients. Twenty-four hours after injection, mice were sacri ced, and the number of CLL B cells in spleen and blood were measured. The migration of T cells from CLL patients was performed by intravenously injecting 5 x10 6 T cells pre-activated for 7 days by anti-CD3/28 Dynabeads + IL-2 in the presence of PI3Ki.
After 1 hour, M-CSF (50 ng/mL) and IL-4 (20 ng/mL) were added to the culture. 48-hours later, RNA was extracted from cells per manufacturer's protocol (Qiagen) and Arg1 mRNA levels were determined by quantitative RT-PCR in triplicates. The threshold cycle (CT) values for Arg1 and Actb were obtained using an ABI7300 TaqMan instrument. IC50 calculations were obtained using Graphpad Prism by nonlinear regression of log inhibitor concentration versus percent inhibition of Arg1 expression. Intracellular staining for cytokine-producing cells. Samples were stimulated with PMA (10 ng/ml) and ionomycin (1 µM) for 4 hours at 37 o C in the presence of monensin (10 µg/ml), stained intracellularly for cytokine-producing live cells, and then analyzed by ow cytometry.
Statistical analyses. All statistical tests were performed using Prism v8 (GraphPad Software, Inc).
Normality was assessed using the D'Agostino-Pearson Omnibus Test, and appropriate parametric and non-parametric analyses performed thereafter. Mann-Whitney U tests were used to compare the numbers of CLL B and T cells and murine macrophages and the percentage of Ki67 + cells. An unpaired parametric t test was used for the comparisons of CLL B-cell viability at day 5. Figures shown are mean or mean ± S.E.M, P < .05 was considered signi cant for all tests.

Declarations
One Sentence Summary: Duvelisib directly inhibits CLL B cells, reduces support from T cells and macrophages, and is effective against CLL cells from patients whose disease has progressed during ibrutinib therapy, regardless of BTK mutation status.    Effect of the addition of duvelisib to co-cultures of CLL B cells with M2-polarized macrophages. *P < 0.05, ** P ≤ 0.01, ***P ≤ 0.001. (F) Number of murine CD11b+GRlowLY-6Clow macrophages in recipient mouse spleens. * indicates P < .05, ** P ≤ .01.