A double blinded, placebo-controlled pilot study to examine reduction of CD34⁺/CD117⁺/CD133⁺ lymphoma progenitor cells and duration of remission induced by neoadjuvant valspodar in dogs with large B-cell lymphoma

Supplies and reagents

Clinical grade valspodar (PSC-833) was provided by Novartis Pharma AG (Basel, Switzerland). Valspodar was compounded for use in pet dogs by Custom Rx Compounding Pharmacy (Roy D. Katz R. Ph., Richfield, MN). Capsules containing 100 mg valspodar or placebo (compounding materials without valspodar) were formulated with the same method used to compound cyclosporine-A for oral use in dogs, since these compounds share a high degree of structural similarity. Activity of the compounded valspodar was confirmed using the side population assays described below. Research grade valspodar and verapamil were purchased from Sigma-Aldrich (St. Louis, MO) and were diluted in dimethyl sulfoxide (DMSO; Sigma-Aldrich) for use in vitro. Lymphoma cells were maintained in short-term culture as described^6,20,21. COSB hemangiosarcoma cells were maintained as adherent cultures as described²².

Trial design

This was a double blinded, placebo-controlled trial with 10 dogs in each study arm. The main statistical endpoint was a change in LPCs following treatment. The hypothesis was that a significant reduction in the number of LPCs in blood and/or in lymph node cells would occur in dogs treated with valspodar, but not in dogs receiving placebo. The sample size of 10 dogs per group was selected to provide 80% power to establish a difference of ± 2 S.D. in LPCs pre-and post-valspodar or placebo treatment within and between groups. The study was not powered to detect significant differences in duration of remission or overall survival. However, outcomes were recorded to evaluate trends that could be used to design future studies or could be included in meta-analysis. Criteria for inclusion were (1) clinical diagnosis of multicentric lymphoma (WHO stage I-V); (2) confirmed WHO classification of large B-cell lymphoma (DLBCL or MZL in transition)²³; (3) favorable performance status with an expected survival time of ≥ 30 days; (4) body weight more than 15 kg (to allow adequate blood sampling) and less than 40 kg (to ensure dosing feasibility); (5) platelet count ≥100,000/ml and packed cell volume ≥30%; and (6) informed pet owner consent in writing. Criteria for exclusion were (1) disease substage b; (2) any previous therapy for lymphoma, including corticosteroids; (3) lymphomas classified as other than DLBCL or MZL in transition; (4) dogs from herding breeds with high frequency of inactivating MDR-1 polymorphisms^24,25; and (5) significant co-morbidities, such as renal or hepatic failure, congestive heart failure, or clinical coagulopathy. There were no restrictions based on age, gender, neuter status, or other physical parameters.

Treatment costs for eligible participants up to $2500 were paid by study funds through the end of the chemotherapy protocol. The study was conducted with approval and under the oversight of the University of Minnesota Institutional Animal Care and Use Committee (IACUC Protocol 1011A92815 “Ablation of tumor initiating cells by P-glycoprotein inhibition: Proof of principle study in canine diffuse large B-cell lymphoma”). The trial design and implementation conformed to the Standard Protocol Items: Recommendations for Interventional Trials (SPIRIT) guidelines²⁶ where they apply to studies in companion animals. The flow of participants is provided in Figure 1. The demographic composition of the study population after unblinding is provided in Table 1. The timing of each procedure is shown in Table 2.

Figure 1. Enrollment, exclusions, and assessments.

Flow chart with details of dogs enrolled in the study and exclusions from each of the measured endpoints.

Sample handling and pathological classification

Incisional wedge biopsies collected during eligibility screening before treatment (Day 0) and tru-cut biopsies collected on the fourth day of neoadjuvant treatment for enrolled dogs (Day 4) were processed as described²⁷. Briefly, representative sections from each biopsy were fixed in 10% neutral buffered formalin for 24 hours and embedded in paraffin for routine histological analysis. Sample processing, staining, and immunohistochemical stains were done by the Comparative Pathology Shared Resource of the Masonic Cancer Center, University of Minnesota. Samples were classified according to the modified WHO scheme for canine lymphoma based on cell morphology, immunophenotyping using antibodies against human CD3 (AbD Serotec Cat# MCA1477T RRID:AB_10845948), human CD20 (Lab Vision Cat# RB-9013-P0 RRID:AB_149766), and CD79a (clone HM47/A9, Cat# CM 067 C, now supplied by Thermo Scientific as RRID:AB_10981393), and available clinical history by two board certified veterinary pathologists (TDO and DMS). The remainder of the biopsy samples was used to prepare single cell suspensions to support the diagnoses through flow cytometry; these suspensions were cryopreserved in liquid nitrogen storage for the following analyses as described^6,27.

Blood samples were collected in evacuated EDTA tubes at Day 0, Day 4, and Day 11 to monitor toxicity and to evaluate blood LPCs. Adverse events were recorded and classified according to the Veterinary Cooperative Oncology Group (VCOG) criteria²⁸.

Flow cytometry

Tumor tissues - Flow cytometry analysis was performed as described^6,20. Briefly, 5 × 10⁵ tumor cells were incubated with dog immunoglobulin G (IgG; Jackson ImmunoResearch, West Grove, PA) to prevent non-specific binding of antibodies to Fc receptors. Cells were stained using fluorescein isothiocyanate (FITC), phycoerythrin (PE), or allophycocyanin (APC) and conjugated antibodies against dog CD3 (clone CA17.2A12, AbD Serotec Cat# MCA1774F RRID:AB_2291174), dog CD4 (clone YKIX302.9, AbD Serotec Cat# MCA1038F RRID:AB_321271), dog CD5 (clone YKIX322.3, AbD Serotec Cat# MCA1037F RRID:AB_322643), dog CD8 (clone YCATE55.9, AbD Serotec Cat# MCA1039PE RRID:AB_322646), dog CD45 (clone YKIX716.13, AbD Serotec Cat# MCA1042F RRID:AB_324047, Cat# MCA1042PE RRID:AB_322644, and AbD Serotec Cat# MCA1042APC RRID:AB_324810), dog CD21 (clone CA2.1D6, AbD Serotec Cat# MCA1781PE RRID:AB_323238), human ABCB1 (clone UIC2, eBioscience Cat# 17-2439-42 RRID:AB_10736477), and human ABCG2 (clone 5D3, eBioscience Cat# 12-8888-82 RRID:AB_466219). Anti-human CD22 antibody (clone RFB4, Abcam Cat# ab23620 RRID:AB_447570) was labeled using the Zenon anti-mouse IgG1 Alexa-Fluor 647 labeling kit (Invitrogen-Molecular Probes, Carlsbad, CA). LPCs were detected by a cocktail of antibodies directed against canine CD34 (clone 1H6, BD Biosciences Cat# 559369 RRID:AB_397238), human CD117 (clone YB5.B8, BD Biosciences Cat# 555714 RRID:AB_396058), and mouse CD133 (clone 13A4, eBioscience Cat# 12-1331-80 RRID:AB_465848), where the mix was designated as “Progenitor”⁶. The antibodies directed against human and mouse antigens have been shown to recognize the canine homologs^6,21,29. Cells were gated based on their light scatter properties, and dead cells were excluded using 7-amino-actinomycin D (7-AAD; eBioscience) staining. Flow cytometry was performed using an LSRII cytometer (BD Immunocytometry Systems, San Jose, CA), and results were analyzed using FlowJo software (Tree Star, RRID:nif-0000-30575).

Peripheral blood – All flow cytometric experiments on peripheral blood were performed at the Flow Cytometry and Cell Separation Facility within the Bindley Bioscience Center at Purdue University using an iCyt Reflection HAPS cell sorter (Sony Biotechnology, Inc., San Jose, CA). Results were analyzed using FlowJo software (Tree Star, RRID:nif-0000-30575). Data to derive the frequency of LPCs in blood were also analyzed using FCS Express (De Novo Software, Los Angeles, CA). Briefly, approximately 50 ml peripheral blood was collected via jugular venipuncture into EDTA tubes from each study dog on Days 1, 4, and 11. Blood samples collected at the University of Minnesota and the University of Pennsylvania were mixed in a 1:1 ratio with RPMI-1640 (Mediatech, Inc., Manassas, VA) and shipped on ice to Purdue University for flow cytometric analysis. Samples collected at Purdue University were processed immediately for analysis. All blood samples were centrifuged at 1500 × g for 20 minutes at 4°C. Plasma was removed by vacuum suction, and the buffy coat was manually harvested from each sample, then transferred to microcentrifuge tubes. Buffy coats were re-centrifuged at 1500 × g for 15 minutes at 4°C, then re-harvested. Cells were stained using FITC, PE, or APC-conjugated antibodies against human CD22 (clone RFB4, Abcam Cat# ab23620 RRID:AB_447570), canine CD34 (clone 1H6, BD Biosciences Cat# 559369 RRID:AB_397238), human CD117 (clone YB5.B8, BD Biosciences Cat# 555714 RRID:AB_396058), and mouse CD133 (clone 13A4, eBioscience Cat# 12-1331-80 RRID:AB_465848). Isotype control antibodies (mouse IgG1 (eBioscience Cat#12-4714-82) and rat IgG2b (eBioscience Cat#11-4031-81) conjugated to APC were used to exclude dead or irrelevant cells, while LPCs were detected by dual staining with FITC-CD22 and PE-“Progenitor mix” (CD34, CD117, CD133). Assuming that circulating LPCs would be very rare in the peripheral blood, approximately 10⁸ cells were sorted at each sampling time point for each dog to provide a reasonable likelihood of identifying this population.

Side population assays

Side populations were measured as described³⁰. Briefly, DyeCycle Violet (DCV) (Life Technologies, Eugene, OR) was added to a final concentration of 10 μM, and 5 × 10⁵ cells were incubated for an additional 60 minutes at 37°C with intermittent mixing. Cells were washed, filtered, and maintained on ice until analysis. To exclude dead cells from analysis, 7-AAD was added to each sample immediately before collection. DCV emission was detected using a BD LSRII flow cytometer (BD Biosciences). Valspodar and verapamil were diluted in DMSO for use in this assay. Equivalent amounts of DMSO were added to control samples, and verapamil was used to determine the side population gates. Data were analyzed using FlowJo software (Tree Star, RRID:nif-0000-30575).

RNA preparation and RNA sequencing

RNA prepared from biopsies obtained at diagnosis (Day 0) and on the fourth day of neoadjuvant treatment for enrolled dogs (Day 4) was quantified and assessed for quality as described^11,22. Briefly, total RNA was quantified using a fluorimetric RiboGreen assay and the total RNA integrity was assessed using capillary electrophoresis in the Agilent BioAnalyzer 2100 to generate RNA Integrity Numbers (RIN). Samples passed a QC step if they contained >1 µg with a RIN >8. Next-generation RNA sequencing (RNAseq, 50-bp paired-end) with HiSeq 2000 (Illumina) was done at the University of Minnesota Genomics Center (UMGC) in 14-paired (pre- and post-treatment) samples and two additional pre-treatment samples as described²². A minimum of ten million read-pairs was generated for each sample. Illumina’s CASAVA software 1.8.2 was used for verifying the quality of the sequence data and for FASTQ file generation and de-multiplexing. FASTQ files and processed data files are available through the National Center for Bioinformatics (GSE93516).

Bioinformatic analyses

Initial quality control analysis of the FASTQ files was performed using the FastQC software (v0.11.5; http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Sequenced reads were trimmed for adaptor sequence, and masked for low-complexity or low-quality sequence using Trimmomatic (v0.33) enabled with the optional "- - quality control" option set as: 3bp sliding-window trimming from 3’ end requiring minimum Q16³¹. Paired-end sequences were mapped to the CanFam3 reference genome with the HISAT2 aligner (2.0.2-beta) using default parameters³². Metrics for insert size distribution of each paired-end library were calculated using Picard software (version 1.126) (http://picard.sourceforge.net.). Samtools (v1.2 using htslib-1.2.1) was used to sort and index each bam file³³. For each sample, a transcript abundance estimation file was generated with Cuffquant (Cufflinks (v2.2.1)) using default settings and the “–multi-read-correct” option enabled³⁴. Cuffnorm was used to generate a table of fragments per kilobase of exon per million fragments mapped (FPKM) expression data using the classic-fpkm normalization method and the “-p/–num-threads” option set to 16. Cuffdiff with default parameters was used to test differences in the summed FPKM of transcripts sharing each gene_id to get gene level differential expression between two groups (i.e., pre- and post-treatment). Expression differences with a false discovery rate (FDR)- adjusted p-value (q-value) of less than 0.05 were considered significant. Principal Component Analysis (PCA) was performed on FPKM values with the built-in R function prcomp and these results were visualized with the ‘ggplot2’ and ‘ggfortify’ packages in Rstudio (Version 0.99.491).

Treatment

Eligible dogs were randomized into an experimental treatment group that was given encapsulated valspodar (7.5 mg/kg orally every 12 hours for 5 days) or a control group that was given the equivalent encapsulated placebo over the same schedule. Starting on Day 4, every dog received five doses of doxorubicin 21 days apart using a dosing schedule based on a previous study using valspodar in the neoadjuvant setting with single agent doxorubicin chemotherapy in dogs with osteosarcoma¹⁴. The first dose was reduced by 30% from the standard (from 30 mg/m² to 21 mg/m²) to mitigate potential side effects of ABCB1 inhibition by the neoadjuvant valspodar. If no serious toxic effects of combined doxorubicin/valspodar were observed, subsequent doxorubicin treatments were dosed at 30 mg/m². If toxic effects were observed, the dose remained at 21 mg/m² and subsequent dose escalation to 30 mg/m² only occurred if no serious adverse events were recorded following the previous dose. An overview of the treatment and collection of blood and tissue samples is provided in Table 2. The treatment responses were evaluated based on the VCOG criteria for lymphoma in dogs³⁵. The last treatment was given at 111 days; dogs were examined once more at 180 days, which was near the expected median survival for single agent doxorubicin protocol³⁶, and then released to their attending veterinarian. The status for each dog was ascertained by telephone or electronic mail communication with the attending veterinarians and/or the owners periodically thereafter until a death event was recorded or >500 days had elapsed. Relapse was determined using clinical parameters (generalized lymphadenopathy on physical exam) with conventional testing as needed (routine radiographs or ultrasound imaging, fine needle aspirate). Dogs were considered off-study at relapse and were then eligible to undergo rescue therapy (N=11) or enter other clinical studies (N=4).

Measurement of valspodar concentration in canine serum

Serum samples collected on the fourth day of neoadjuvant treatment (Day 4) were stored at -80°C until analysis. Valspodar was quantified by liquid chromatography/mass spectroscopy (LC-MS/MS) using a high-performance liquid chromatograph (Agilent 1200 Series, Santa Clara CA) coupled with a TSQ Quantum triple stage quadrupole mass spectrometer (Thermo-Electron, San Jose, CA) as described³⁷.

Statistics

Descriptive statistics (mean, median, minimum, maximum) were recorded for age, gender, breed, and disease stage; for each variable, differences between groups were determined using Fisher’s exact test. Time to remission, duration of remission, and overall survival were recorded in days starting on the date that the dogs first received a clinical diagnosis. The percentage of LPCs in lymph node samples was calculated based on expression of relevant cell surface markers (CD34/CD117/CD133) as a proportion of live, large, CD22⁺ B cells⁶. The ΔLPC was calculated as the ratio of LPCs at Day 4 over LPCs at Day 0. The Mann-Whitney Test (Prism 5, GraphPad Software, Inc., La Jolla, CA) was used to determine significance between LPC numbers in lymph nodes from dogs in the experimental treatment and in the control groups. The associations between variables were determined using the Pearson correlation. Differences between groups in duration of remission and overall survival were determined using Kaplan-Meier probability and log-rank tests. The proportion of LPCs in peripheral blood was calculated by conservatively gating on CD22/“Progenitor mix” double positive cells identified in two-dimensional scatter plots. However, the very low numbers of LPCs detected in peripheral blood and the presence of random effects (significant intra- and inter-dog variation in light scatter parameters characterizing LPC cells shape and size) made a direct comparison of the identified proportions between dogs challenging. The effect of the therapy on the percentage of cells was expressed therefore as a summary of Cohen’s h values computed for every dog, and for every pair of time points. The summary h and accompanying 95% confidence intervals were calculated assuming a random-effects model.

Inhibition of drug efflux by valspodar in vitro

Valspodar is a potent, selective inhibitor of the ABCB1 efflux transporter^12,13. To confirm that the clinical grade compound retained potency after compounding, we examined its effect to inhibit DCV efflux using the flow cytometric side population assay. COSB canine hemangiosarcoma cells contain a subpopulation of cells that shows robust dye efflux in this assay³⁰ (Figure 2, Dataset 1). The compounded, clinical grade valspodar was as effective as the research grade valspodar in this assay, eliminating >90% of the side population (i.e., it inhibited dye efflux) at concentrations as low as 30 ng/ml (Figure 2, Dataset 1). The effect of valspodar was comparable to that observed in verapamil (Figure 2, Dataset 1), which inhibits both ABCB1 and ABCG2 at the 50–100 µM concentrations used in this assay.

Figure 2. Valspodar inhibits dye excluding side population in canine hemangiosarcoma cells at clinically achievable concentrations.

Side population analyses were done as described in Materials and methods using cultured COSB canine hemangiosarcoma cells. (A) Live cells were gated based on light scatter properties and exclusion of 7-AAD, and (B) the side populations were determined based on DyeCycle Violet (DCV) efflux. Verapamil was used to inhibit ABCB1 and ABCG2 at 50–100 µM concentrations. Clinical grade and research grade valspodar was used at concentrations that were achieved in the plasma of dogs in the study (30 – 600 ng/ml) as well as at the saturating dose of 1 µg/ml. The Y-axis is DCV-blue (450+/-50 nm) emission while the X-axis is DCV-red (660 +/- 40 nm) on the LSR-II. Data were analyzed and dot plots were created in FlowJo.

Recruitment and randomization

Excluding dogs that had received previous chemotherapy, 40 dogs were screened for eligibility. Twenty dogs were eligible and enrolled in the trial. Of the 20 dogs that were excluded, 5 dogs had lymphomas that were classified as other than DLBCL or MZL in transition (specifically, three had T-cell lymphoma, one had an indolent type of lymphoma, and one had disease largely confined to spleen with minimal peripheral lymphadenopathy that precluded biopsy) and 15 dogs had hypercalcemia (N=2), lymphoma in substage b or a ongoing co-morbidity (N=8), exceeded the maximum allowable body weight (N=1), or the owners declined participation (N=4).

Of the twenty dogs enrolled, 10 were randomized to each group. The distribution of dogs according to demographic characteristics is shown in Table 1. The composition of the study population was predictable^38,39, and there were no statistically significant differences in any category between the experimental treatment group and the control group. One dog in the placebo group did not receive doxorubicin chemotherapy after the neoadjuvant period per its owner’s decision. This dog was censored in the outcome assessments.

Toxicity attributable to valspodar

Six dogs, including three in the placebo group and three in the experimental (valspodar) group, had reportable events during the study (Table 3). The most common toxicities observed in both groups were grade-1 and grade-2 inappetence, lethargy, vomiting, and diarrhea. No grade-4 or grade-5 toxicities were observed, although one event was potentially dose limiting. One dog had grade-2 hematological toxicity (neutropenia and thrombocytopenia) after the first administration of doxorubicin. The doxorubicin dose for the second administration was maintained at 21 mg/m² and no toxicity was observed. However, the owner only permitted subsequent doxorubicin doses to be escalated to 24 mg/m². The dog that was withdrawn after neoadjuvant placebo had grade-2 gastrointestinal toxicity and grade-1 lethargy.

Quantification of LPCs in blood and lymph nodes from dogs with lymphoma

Blood and lymph node LPCs were quantified for each dog at diagnosis (Day 0) and on the fourth day of neoadjuvant treatment (Day 4) as described in Materials and Methods. Blood LPCs were also quantified for each dog 7 days after doxorubicin treatment (Day 11). The distribution of lymph node LPCs at diagnosis was narrower in the dogs that received valspodar than in the control dogs (Figure 3A), but the two groups were not significantly different, and neither group showed a statistically significant reduction in LPCs on the fourth day of the neoadjuvant period (ΔLPC). Table 4 shows that LPCs were detectable in every peripheral blood sample; however, at a frequency lower than what was previously reported in lymph nodes⁶. The effects of treatment on peripheral blood LPCs, as described by Cohen’s h, were extremely small and characterized by wide confidence intervals (Figure 4A). There was an intriguing reversal of trends in the frequency of blood LPCs in dogs treated with valspodar and dogs treated with placebo at day 11 (one week after administration of doxorubicin, Figure 4B). However, this must be interpreted cautiously. First, we must note that the number of blood LPCs were not significantly different (p=0.18) using analysis of variance (ANOVA). Second, large variances were present within each group, as showed by the Cohen’s h (Figure 4A). And finally, the number of LPCs in all the samples remained within the relatively wide range observed in pre-treatment samples. The low abundance of LPC cells means that even a decrease in frequency by an order of magnitude would result in an effect size no larger than h=0.0013 per dog. Representative dot plots illustrating gating and presence of LPCs in peripheral blood are depicted in Supplemental Figures 1A – 1D.

Figure 3. LPCs in lymph nodes from dogs with large B-cell lymphoma at diagnosis and on the fourth day of neoadjuvant treatment with valspodar.

Left: Box plots showing median (white line), 75% confidence intervals, and outliers of the percent LPCs in lymph nodes at diagnosis, and Right: relative change in LPCs from the time of diagnosis (Day 0) to the fourth day of the neoadjuvant period (Day 4) in each group of dogs. A ΔLPC = 1.0 means no change in the percent LPCs measured at both time points. Data were analyzed and graphs were assembled using MS Excel.

Figure 4. Effect of valspodar treatment on depletion of peripheral blood LPCs in dogs with diffuse large B-cell lymphoma.

(A) Plots of effect sizes (Cohen’s h), with corresponding 95% confidence intervals, associated with valspodar treatment on LPC frequency in peripheral blood. Effect sizes of valspodar treatment on changes in LPC frequency between all sampling time points are depicted. All calculated effect sizes were small, suggesting no effect of therapy on changes in peripheral blood LPC proportion between time points. However, the 95% confidence intervals are wide due to high variability in sample quality. t1=Day 0; t2=Day 4; t3=Day 11. (B) Mean frequency of blood LPCs from Day 0 to Day 11 in dogs receiving placebo (red lines, squares) and in dogs receiving valspodar (green lines, triangles). Dotted lines denote the 95% lower and upper boundaries of LPC frequencies for each group.

Alterations in ABCB1 and ABCG2 expression by LPCs in lymph nodes from dogs with lymphoma

The absence of a treatment effect on total LPCs suggested that we should not reject the null hypothesis that neoadjuvant valspodar did not enhance chemosensitivity of LPCs, and could reflect variable expression of ABC transporters by these cells. Samples from 15 dogs in the study (six in the placebo group and nine in the valspodar group) had sufficient material for analysis of ABCB1 and ABCG2 expression in LPCs at diagnosis. The proportion of ABCB1⁺ LPCs and ABCG2⁺ LPCs was variable. In the placebo group, between 1.6% and 52.4% of lymph node LPCs expressed these proteins at the time of diagnosis; in the valspodar group, the range of ABCB1 and ABCG2 transporter expression in lymph node LPCs at the time of diagnosis was 10.0% to 72.7% (Table 5). When we examined the proportion of ABCB1⁺ LPCs and ABCG2⁺ LPCs in dogs from each treatment group, we saw an intriguing reversal in the trends with regard to event-free survival (Figure 5), although neither group showed a significant correlation between the number of ABCB1⁺ or ABCG2⁺ cells at diagnosis and survival (all the R² values were less than or equal to 0.42).

Figure 5. Event-free survival of dogs as a function of ABCB1 and ABCG2 expression in lymph node LPCs from dogs with large B-cell lymphoma at diagnosis.

Dot plots showing the relationship between ABCB1 expression and event-free survival (EFS) in days (top) and between ABCG2 expression and EFS in days (bottom) in dogs treated with placebo (N = 9) or with neoadjuvant valspodar (N = 9) where samples were available for these measurements. The dashed lines represent linear regressions and their R² values are indicated on each graph. The Y-axis represents the % of ABC⁺/Progenitor ⁺lymph node B cells. Data were analyzed and graphs were assembled using MS Excel.

Samples from four dogs (two in the placebo group and two in the valspodar group) had sufficient material for analysis of ABCB1 and ABCG2 to determine if valspodar specifically reduced the number of ABCB1⁺ and ABCG2⁺ LPCs in paired pre-and post-treatment samples. There was a quantifiable decrease in the frequency of ABCB1⁺ and ABCG2⁺ LPCs, but this change was comparable between the two dogs that received valspodar and the two dogs that received placebo (Table 5 and Supplemental Figures 2A–2D, Dataset 1b).

Alterations in genome-wide gene expression in lymph nodes from dogs with lymphoma

We examined if the inhibition of ABCB1 activity with valspodar changed genome-wide patterns of gene expression in lymph nodes from dogs in both groups. Paired pre- (Day 0) and post- (Day 4) treatment samples were available from five dogs in the placebo group and from nine dogs in the valspodar group. One additional pre-treatment sample from dogs in each group was available and included in the analysis, making a total of 16 pre-treatment samples and 14 post-treatment samples. Using Cuffdiff analysis, we identified genes with significantly differently expression between treatment groups and between pre- and post-treatment samples (Supplemental Table 1 and Supplemental Table 2). However, as illustrated by principal component analysis (Supplemental Figure 3), we were unable to achieve a meaningful separation of groups based on the effects of treatment on genome-wide gene expression.

Bioavailability of valspodar

The observation that valspodar treatment did not specifically alter the total blood or lymph node LPCs or the frequency of ABCB1⁺ and ABCG2⁺ LPCs, and that it did not lead to significant changes in genome-wide gene expression of lymph node cells, could be attributed to poor bioavailability. To evaluate this possibility, we examined the purity of the compounded, encapsulated drug and the levels of valspodar in serum samples obtained at Day 4 from seven dogs using LC-MS/MS. Valspodar was undetectable in placebo capsules, and the purity of the compounded capsules was 104% as compared to research grade valspodar.

Valspodar was also undetectable (<5 ng/ml) in dogs that received placebo, but it was present at detectable levels in each of four dogs that received compounded valspodar capsules (34, 63, 375, and 623 ng/ml, respectively). This is equivalent to levels between 0.025 to 0.5 µM on the fourth day of twice-daily administration, which is in the range seen in dogs where valspodar was given at the same dose in an oil-based drinking solution¹⁴.

Clinical responses

Eighteen treated dogs achieved clinical remission, defined as a complete response (disappearance of all evidence of disease with all lymph nodes shrinking to non-pathologic size within the judgment of the evaluator) after the first dose of doxorubicin. One dog in the valspodar group did not achieve clinical remission but survived with stable disease for 428 days. One dog in the placebo group never received doxorubicin and was censored from this analysis. This dog was treated with palliative intent using prednisone only; it failed to achieve remission and died 59 days after diagnosis.

The time to remission after the initial valspodar treatment ranged from 7 to 106 days (after doxorubicin) in the placebo group, and from 7 to 105 days (after doxorubicin) in the valspodar group (excluding the dog that never achieved remission). There were no differences between groups with reference to the median time to remission, the median (or range) duration of remission, the number of dogs alive at the 180-day milestone, or the number of dogs alive at 500 days (Table 6). The event-free survival and overall survival times for each group are shown in Figure 6.

Figure 6. Effect of neoadjuvant valspodar on survival of dogs with large B-cell lymphoma.

Kaplan–Meier analysis of event-free survival (top) and overall survival (bottom) in dogs treated with doxorubicin with the addition of neoadjuvant placebo or valspodar. The table below the graphs shows the median event-free and overall survival for each group. Data were analyzed and graphs were assembled using MS Excel.

Correlation between ΔLPCs and outcome

To test the hypothesis that LPCs contribute to disease progression, we examined if there were direct or inverse correlations between the proportion of LPCs at diagnosis and the ΔLPCs with duration of remission as well as with overall survival for dogs in the valspodar and control groups, individually and for all of the dogs in the study. Figure 7A and 7B show scatterplots illustrating no correlations between the proportion of lymph node LPCs at diagnosis and the ΔLPCs (D4/D0), respectively, and event-free survival (duration of remission) and overall survival. The results were similar when we analyzed correlations between the proportion of blood LPCs at diagnosis or ΔLPCs and survival outcomes (data not shown).

Figure 7. Event-free and overall survival of dogs as a function of lymph node LPCs from dogs with large B-cell lymphoma at diagnosis.

(A) Dot plots showing the relationship between the percent of lymph node LPCs at diagnosis and EFS (N=9), and the relative change in LPCs from the time of diagnosis (Day 0) to the fourth day of the neoadjuvant period (Day 4) and EFS (N=8), in days in dogs treated with placebo (N = 9) or with neoadjuvant valspodar. (B) Dot plots showing the same relationships for overall survival (OS, N=9 and N=10 for LPCs at diagnosis and for ΔLPCs, respectively). Data were analyzed and graphs were assembled using MS Excel.