Prospective Evaluation of the Lymph Node Proteome in Dogs with Multicentric Lymphoma Supplemented with Sulforaphane

AN ABSTRACT OF THE THESIS OF

Cyril Parachini-Winter for the degree of Master of Sciences in Comparative Health Sciences presented on June 20, 2019.

Title: Prospective Evaluation of the Lymph Node Proteome in Dogs with Multicentric Lymphoma Supplemented with Sulforaphane

Abstract approved: ______

Shay Bracha

Lymphoma (LSA) is one of the most common canine malignancies, and is almost invariably a terminal disease. Epigenetic changes in canine and human lymphomas have been linked to disease progression and poor prognosis, leading to the development of epigenetic-targeted therapies which have shown promise in treating various forms of human LSA. Sulforaphane (SFN), a compound derived from cruciferous vegetables, has recently gained significant interest related to cancer prevention and therapy, in particular due to its effects on the epigenome. However, the use of SFN in tumor-bearing dogs has not been reported. The goal of this study was to examine the impact of SFN supplementation on the lymph node proteome of dogs with multicentric LSA. Seven treatment-naïve dogs with multicentric LSA were prospectively enrolled. Lymph node samples were obtained before and after a week of oral SFN supplementation, analyzed by label-free mass spectrometry and Gene Set Enrichment Analysis followed by validation of the results with Immunoblot analysis. No adverse events directly attributed to SFN supplementation were noted. A total of 915 proteins were detected across all dogs, among which 14 proteins were significantly downregulated, and 10 significantly upregulated post-SFN supplementation. For each individual dog, the expression of several hundred of proteins changed by at least two-fold post-SFN supplementation. Proteome changes post-SFN were not dependent on LSA immunophenotype. Recurrent biological functions for the proteins and gene sets identified included regulation of innate or adaptive immunity, regulation of the transcription machinery, protein transport and ubiquitination, cellular response to oxidative stress, and apoptosis. These results confirm that the impact of SFN in neoplastic cells are manifold, and support further investigation of this compound in canine LSA patients, alone and/or in combination with standard therapies.

by Cyril Parachini-Winter

A THESIS

submitted to

Oregon State University

in partial fulfillment of the requirements for the degree of

Master of Sciences

Presented June 20, 2019 Commencement June 2020 Master of Sciences thesis of Cyril Parachini-Winter presented on June 20, 2019

APPROVED:

Major Professor, representing Comparative Health Sciences

Dean of the College of Veterinary Medicine

Dean of the Graduate School

I understand that my thesis will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my thesis to any reader upon request.

Cyril Parachini-Winter, Author ACKNOWLEDGEMENTS

The author expresses sincere gratitude to the dogs and pets’ owners who participated in this study. Without their unwavering dedication, none of this would have been possible.

I also want to thank profusely Carl E. Ruby for his incredible patience, availability and kindness, for answering all my bench-top experiments related questions, and for his unrelenting faith in immunoblots!!

CONTRIBUTION OF AUTHORS

Dr Shay Bracha assisted with organizing and conducting the immunoblots, interpreting the results and revising this thesis.

Dr Kaitlin Curran had the original idea of investigating sulforaphane in tumor bearing dogs, helped greatly with hypothesis generation, patients’ lymph node sample collection and processing, analyzing the results, and revising this thesis.

Dr Liping Yang was of invaluable help to conduct the mass spectrometry experiments, writing the corresponding material and methods section, and assisted in the interpretation of the raw proteomic data.

Dr Stephen Ramsey performed and assisted in interpretation of the gene set enrichment analysis, and was involved with writing the corresponding materials and methods section.

TABLE OF CONTENTS

Page

1. Introduction ...... 2

1.1 Canine multicentric lymphoma: a drive for new treatment avenues ...... 2

1.2 Introduction to histone deacetylases and DNA methyltransferases ...... 2

2. Literature review ...... 4

2.1 Sulforaphane and BroccoMaxTM: pharmacokinetic data ...... 4

2.2 Sulforaphane and cancer prevention ...... 5

2.3 Anti-cancer effects of sulforaphane ...... 5

2.4 Impact of sulforaphane on cancer cells’ proteome ...... 6

2.5 Targeted epigenetic therapy in human lymphoid neoplasms ...... 7

2.6 Targeted epigenetic therapy in veterinary oncology ...... 9

2.7 Significance and aims of the study ...... 11

3. Materials and methods ...... 13

3.1 Patients recruitment ...... 13

3.2 Study design and sample processing ...... 13

3.3 Mass spectrometry ...... 14

3.4 Immunoblots ...... 15

3.5 Gene Set Enrichment analysis ...... 17

4. Results ...... 18

4.1 Patients population ...... 18

4.2 Clinical response to sulforaphane and outcome ...... 18

4.3 Mass spectrometry results: all dogs at day 0 vs all dogs at day 7 ...... 20

4.4 Mass spectrometry results: individual dogs ...... 29

4.5 Mass spectrometry results: B-cell vs. T-cell lymphoma ...... 31 TABLE OF CONTENTS (Continued)

4.6 Validation of mass spectrometry data with immunoblots ...... 33

4.7 Gene Set Enrichment analysis results ...... 35

5. Discussion ...... 39

5.1 Proteins of interest previously investigated in human malignancies ...... 39

5.2 Proteins of interest previously explored in veterinary oncology ...... 40

5.3 Recurrent biological functions ...... 41

5.4 Oncogenic pathways involved ...... 42

5.5 Marked variability in individual dogs’ proteome response to sulforaphane supplementation ...... 42

5.6 Immunoblots validated the mass spectrometry results ...... 43

5.7 Perspectives from previous canine lymphoma proteomic studies ...... 43

5.8 Differences in the proteome of B-cell vs. T-cell lymphoma ...... 44

5.9 Clinical response to sulforaphane supplementation ...... 44

5.10 Limitations ...... 45

6. Conclusion ...... 46

Bibliography ...... 47

LIST OF FIGURES

Figure Page

1. Venn diagram depicting the amount of proteins detected at both day 0 and day 7 (n=885), detected in at least one dog pre-SFN but none of the dogs post-SNF (7), and detected in none of the dogs pre-SFN but a least one dog post-SFN (21) ...... 26

2. Comparison of the proteins fold changes post-sulforaphane in dogs with T-cell lymphoma (dogs #1 and 2) vs. dogs with B-cell lymphoma (dogs #3-7) ...... 33

3. Immunoblots of lymph node samples for 7 dogs (d) with naïve multicentric lymphoma pre (D0) and post (D7) sulforaphane supplementation ...... 34

4. Bar graph displaying the normalized enrichment score of the 11 non redundant gene sets significantly upregulated (nominal p-value <0.01) at a FDR <0.1, across all dogs at day 7 (post-sulforaphane) ...... 36

5. Enrichment plots of the most upregulated gene set (left, “regulation of protein maturation”) and the most downregulated gene set (right, “tRNA processing”) across all dogs post-sulforaphane ...... 37

LIST OF TABLES

Table Page

1. Characteristics of seven dogs enrolled in a study investigating the impact of oral sulforaphane supplementation in canine multicentric lymphoma ...... 18

2. Summary of adverse events experienced by canine patients with naïve multicentric lymphoma during seven days of oral supplementation with sulforaphane ...... 19

3. Characteristics of twelve proteins significantly downregulated after seven days of oral sulforaphane supplementation in dogs with multicentric lymphoma ...... 21

4. Characteristics of nine proteins significantly upregulated after seven days of oral sulforaphane supplementation in dogs with multicentric lymphoma ...... 24

5. Description and main biological functions of the proteins detected in at least one dog at day 0 but none of the dogs at day 7 ...... 27

6. Description and main biological functions of the proteins detected in none of the dogs at day 0, but at least one dog at day 7 ...... 28

7. Number of proteins upregulated and downregulated for each dog after seven days of oral sulforaphane supplementation ...... 30

8. Characteristics of the proteins most downregulated and most upregulated for each dog after seven days of oral SFN supplementation ...... 31

Prospective Evaluation of the Lymph Node Proteome in Dogs with Multicentric Lymphoma Supplemented with Sulforaphane

1. Introduction

1.1 Canine multicentric lymphoma: a drive for new treatment avenues

Lymphoma (LSA) is one of the most frequently diagnosed malignancies in the dog, accounting for 7% to 24% of all diagnosed canine neoplasia. The multicentric form is by far the most common clinical presentation, with over 80% of dogs with LSA presenting with peripheral lymphadenomegaly.1,2 Dogs are most commonly affected by high grade LSA, while less than 30% of canine LSA are considered low grade.1,2 The most common subtype of LSA in dogs, similar to the disease in people, is diffuse large B-cell LSA (DLBCL), which accounts for 36-58% of all cases.3 Despite years of research and a large number of studies investigating various combinations of chemotherapeutic agents, the prognosis for canine multicentric LSA has not improved significantly in the last two decades. The median survival times reported are 9-14 months, and only 20-25% of dogs are still alive 2 years after the diagnosis.4-7 Surgery is rarely a reasonable treatment option. A moderate improvement in the remission length and survival times have been noted in some studies with the addition of half-body radiation therapy during or immediately after conventional chemotherapy protocols. For example, some authors have reported a median survival time of 23 months and 1-year survival rate of 66% with the combination of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP)-based chemotherapy and low dose rate irradiation.8 Other studies however showed no significant improvement in survival time with the addition of radiation therapy.9-11 Overall, long-term remission remains rare and most dogs with multicentric LSA eventually die from the disease. In this regard, there is a significant need for new treatment options in order to improve the prognosis of dogs with LSA.

1.2 Introduction to histone deacetylases and DNA methyltransferases

The nucleosome is the building block of chromatin, consisting of helical DNA looped around a core histone octamer. Addition of acetyl groups to histone tails by histone acetyl transferases (HATs) creates an open chromatin conformation and promotes gene expression.12 Conversely, the removal of acetyl groups by histone deacetylases (HDAC) results in a “closed” chromatin conformation, expulsion of transcription activators, and repression of transcription.12 Of particular importance, removal of acetyl groups from histone in tumor suppressor genes decreases their transcription, and has been linked to increased risk, severity and recurrence rate of various cancers.12 Four classes of HDAC have been described. Class I (HDAC 1, 2, 3, and 8) are expressed in all tissues. Class IIa (HDAC 4, 5, 7, and 9) are found predominantly in tissues such as muscle, brain, heart, endothelial cells, and thymocytes. Class IIb includes HDAC 6 and 3

10, with HDAC 6 also targeting various non-histone substrates. Class III (Sirtuins 1-7) and class IV (HDAC 11) have also been described.12,13 The process of DNA methylation involves the addition of a methyl group to the cytosine pyrimidine ring to form 5-methylcytosine. Methylation commonly occurs on dinucleotides where cytosines precede guanines (CpGs).14 This event is especially common in the promoter region of numerous genes where many CpGs dinucleotides (called CPGs islands) are present. This methylation interferes with the binding of transcription factors and attracts proteins named methyl-CpG-binding domain (MBDs) to initiate chromatin compaction and gene silencing.14 A variety of enzymes, called DNA methyltransferases (DNMTs), are responsible for addition of these methyl groups and maintenance of DNA methylation patterns. 14 Aberrant methylation in the promoter region of tumor suppressors can result in tumor suppressor transcriptional silencing, loss of cell cycle checkpoints, inactivation of apoptotic pathways and decreased sensitivity to chemotherapy.15 Moreover, DNA methylation and histone acetylation pathways can crosstalk to influence gene expression. For example, hypermethylated CpG islands attract HDAC enzymes, and DNMT1 recruits class 1 or 2 HDAC to function as co-repressors.16

Unlike most stable mutations in the DNA sequence, a key feature of epigenetic events is their reversibility. This provides a unique opportunity for the introduction of new therapeutic interventions in various malignancies.14 Several drugs have been developed that are able to thwart tumor-promoting epigenetic profiles. These drugs include histone deacetylase inhibitors (HDACi) such as suberoylanilide hydroxamic acid (SAHA; also known as vorinostat), trichostatin A (TSA), valproic acid, romidepsin, belinostat, abexinostat or panobinostat, as well as DNMT inhibitors (DNMTi) such as azacitidine, decitabine or zebularine.13

Epigenetic events at the DNA level can influence the expression of numerous proteins.17 Such events may change the chromatin conformation, or transcription factors accessibility to gene promoter regions. Moreover, epigenetic events can occur directly at the protein level. For example, class IIb HDAC can remove acetyl groups from non-histone targets such as α-tubulin, aggresomes, heat shock protein 90 (HSP90), nuclear factor kappa beta (NFκB), P53, hypoxia inducible factor 1 (HIF1-α), c-Myc and estrogen receptors.12,18,19 Several protein substrates of HDAC6 interact with tubulin and HSP90 and enhance protein stability.12 Therefore, epigenetic-targeted treatments such as HDACi or DNMTi have the potential to directly affect the proteome content of cells. As an example, vorinostat led to a decrease in the expression level of various anti-apoptotic proteins such as XIAP, Mcl-1, survivin and c-FLIP, and to an increase in the expression of the negative cell cycle regulator protein p21.17

2. Literature review

2.1 Sulforaphane and BroccoMax™: pharmacokinetic data

2.1.1 Sulforaphane pharmacokinetic data in human

Sulforaphane (SFN) is a compound that has recently gained significant interest related to cancer prevention and therapy.20,21 Sulforaphane is a natural isothiocyanate derived from cruciferous vegetables such as broccoli, Brussel sprouts, cauliflower and cabbage. The highest concentration of SFN is found in the seed and sprouts.20,21 Hydrolytic conversion of the precursor glucoraphanin (GFN) to SFN is catalyzed by the enzyme myrosinase. Mammalian cells do not have endogenous myrosinase activity, and in the plant form GFN and myrosinase are physically separated.20 Hence, the conversion of GFN to SFN occurs only after physical damage to the plant (for example when it is chewed) or in vivo from myrosinase produced by the colon microbiota.20,22 As myrosinase is heat labile, the bioavailability of SFN is significantly reduced if the plant is cooked. Bioavailability is also reduced if the plant is not chewed or blended before ingestion, because the endogenous plant myrosinase and GFN remain physically separated.20,22

After hydrolytic conversion, SFN diffuses rapidly across the intestinal epithelium, and is metabolized through the mercapturic acid pathway, which involves conjugation to glutathione (GSH) followed by urinary excretion in the form of various metabolites such as cysteine-SFN and N-acetyl- cysteine-SFN.20,21 Studies in humans showed a peak plasma concentration 1-3 hours after feeding, a half- life time in plasma of 1.8 hours, and extensive tissue distribution (breast, prostate, intestines). Most of the SFN is cleared from the body within 72 hours.18,20-24

BroccoMax™ (Jarrow Formulas®, Los Angeles, CA, USA) is a commercially available oral broccoli supplement containing a standardized amout of GFN and myrosinase. It is commercialized as a capsule containing 30 mg of GFN, each yielding approximately 8 mg of SFN. This formulation enables intake of large amount of SFN with few capsules, which would otherwise only be achieved after ingestion of a substantial amount of broccoli sprouts, and an even larger amount of raw broccoli. As an example, in a recent study in people, the same amount of GFN contained in 6 BroccoMax™ capsules was obtained with the ingestion of 40 grams of broccoli sprouts.22

2.1.2 Sulforaphane pharmacokinetic data in dogs

Our team has recently conducted the first pilot study to characterize the bioavailability of SFN in dogs.25 In that study, ten healthy dogs over 1 year of age were administered three capsules of BroccoMax™ 5

once orally, to achieve 90 mg total of GFN (approximately 24 mg of SFN). Sulforaphane metabolites in plasma and urine were analyzed using liquid chromatography and mass spectrometry (LC-MS/MS). Histone deacetylase activity was also measured from peripheral blood mononuclear cells (PBMC) lysates. We found a bioavailability profile comparable to what is noted in humans. Plasma total SFN metabolite levels peaked at 4 hours post-consumption and were cleared by 24 hours. Sulforaphane-GSH was the major metabolite detected in plasma. Urinary SFN metabolites peaked at 4 hours post-consumption, and remained detectable at 24 and 48 hours. Sulforaphane and cysteine-SFN were the two major SFN metabolites detected in urine samples. A trend for decreased HDAC activity was observed at 1h post-consumption (11.3% decrease vs. baseline) and a significant decrease was observed at 24h post-consumption (24.4% decrease vs. baseline). These data are the first to indicate that SFN is absorbed and metabolized in dogs, but also that SFN exerts some HDACi activity following a single dose. Moreover, no adverse events attributed to the BroccoMax™ were noted during the 48-hour study period.

2.2 Sulforaphane and cancer prevention

The importance of SFN in cancer lies not only in its “suppressive” properties (inhibition of cancer growth), but also in its “blocking” activity (cancer prevention).20,21 The latter was first suspected when researchers detected an inverse correlation between cruciferous vegetable intake, and risk of colon and prostate cancer in people.20 Sulforaphane was subsequently shown to inhibit phase 1 enzymes such as those of the cytochrome P450 family, which are involved in the conversion of pro-carcinogens to carcinogens.20 Sulforaphane also induces phase 2 enzymes such as glutathione transferase (GST), which detoxifies oxidants and carcinogens, and facilitates their excretion from the body. Sulforaphane is involved in degradation of the Keap1 protein, which leads to disruption of the nuclear factor erythroid 2-related factor 2 (Nrf2)-Keap1 complex, nuclear translocation of Nrf2, binding of Nrf2 to anti-oxidant response element, and modulation of the transcription of target genes including many phase 2 enzymes. 20,21,24 Beaver et al. has also shown that SFN can influence the expression of long non-coding RNAs (decrease expression of LINC01116) which could play a role in prostate cancer prevention.26

2.3 Anti-cancer effects of sulforaphane

The effects of SFN on cancer cells are numerous and include inhibition of HDAC, decreased DNA methylation, disruption of various oncogenic pathway, and increased sensitivity to chemotherapeutic agents.16,18-20,27-31 These effects have been demonstrated both in solid tumors (small intestine, lung, colon, 6

prostate, ovaries) and hematopoietic malignancies (B and T-cell leukemia, multiple myeloma) in rodents and people.16,27,30-32

One of the most widely recognized properties of SFN is the ability to act as a natural HDACi. Oral supplementation with SFN leads to increased levels of acetylated histones (H3 and H4), reactivation of P21 and bax transcription, cell cycle arrest, and apoptosis in the colonic mucosa of APCmin mice.27 Polyp development decreased by 50% in all intestinal regions of these mice. It was suspected that the HDACi properties of SFN allowed acetyl groups to be added to histone tails, leading to loosening of the DNA. Transcription factors’ access to P21 and bax genes promoters was then facilitated, and the corresponding proteins re-expressed, mediating cell cycle arrest and apoptosis.27 In a murine xenograft model of human PC-3 prostate cancer, SFN was more effective than TSA in inhibiting tumor growth.19

Silencing cyclin D2 expression by promoter methylation has been associated with progression of breast, lung, pancreatic, and gastric cancers.16 In human prostate cancer cells, exposure to SFN led to down regulation of DNMT 1 and 3a, decreased methylation of the CpG island of cyclin D2 promoter, increased cyclin D2 mRNA expression, and decreased cell proliferation.16 In ovarian cancer cells, SFN blocked the degradation of the retinoblastoma protein (Rb), which can in turn sequester DNMT1 and inhibit its interaction with DNA.33

In vitro studies have also shown that SFN can directly disrupt various pathways in colon and prostate cancer, acute leukemia, and multiple myeloma. These pathways include phosphatidylinositol (PI3K/AKT), mitogen associated protein kinase (MAPK/MEK/ERK), NF-κB, and cJUN.20,29-31,34 Sulforaphane can also activate both mitochondrial and death receptor apoptotic pathways. In prostate and colon cancer cells, SFN activates caspase 3, 7, 8, 9, Bax and Bak, and decreases Bcl-Xl, Bcl-2, MCL-1 and IAP levels.20,29-31,34 Moreover, since binding of SFN with GSH is a necessary step in SFN metabolism, a high level of SFN supplementation can deplete intracellular GSH, leading to increase reactive oxygen species (ROS) production, mitochondrial membrane disruption, cytochrome C release, and apoptosis.20

2.4 Impact of sulforaphane on cancer cells’ proteome

The expression of more than 100 proteins were up or down regulated with >15 fold change in breast cancer cells after treatment with SFN.35 Members of aldo-keto reductase family were amongst the most strongly upregulated (303 fold increase for aldo–keto reductase family 1, member B10).35 In rat cardiomyocytes, 64 proteins were differentially expressed between cells treated with SFN and cells treated with DMSO, including macrophage migration inhibitory factor and glyoxalase 1.36 Sulforaphane can covalently bind cysteine residues on various proteins, and in human non-small cell lung cancer cells, 7

researchers have found that more than 30 proteins were directly targeted by SFN, including cytoskeleton proteins (actin, tubulin), redox regulating proteins (thioredoxin-1, glutaredoxin-1), proteasome subunits, heat shock proteins, mitochondrial proteins (NADH dehydrogenase, cytochrome c oxidase, ATP synthase) and proteins of the 14-3-3 family.37

Overall, these results show that SFN can influence the expression level of many individual proteins, and also have a significant effect on the cancer cell proteome as a whole.

2.5 Targeted epigenetic therapy in human lymphoid neoplasms

Unlike most genetic lesions, epigenetic alterations are reversible, and targeting the epigenome has been the focus of numerous recent studies.15 Mutations affecting HAT have been identified in 16-39% of human DLBCL and follicular lymphoma (FL) cases. DNA methyl-transferase 1, 3A and 3B were overexpressed in 13-48% of DLBCL, and correlated with advanced clinical stage.15 Aberrant DNA methylation correlated with the aggressiveness and chemoresistance of these forms lymphoma.15 Frequent methylation of the p15 and p16 tumor suppressor genes was detected in both B- and T-cell lymphomas, but was not appreciated in the reactive lymph nodes analyzed.38 Moreover, p15 and p16 genes were methylated more often in high- grade vs. low grade lymphomas.38 Others have shown aberrant DNA methylation of tumor suppressors such as p57 (KIP2) or the DNA repair enzymes O6-methylguanine-DNA methyltransferase (MGMT) in various subtypes of human B-cell lymphomas.39,40 Overall, many epigenetic aberrations have been detected in human lymphoid neoplasms, and epigenetic-targeted therapies are appealing new options for the treatment of various forms of LSA.

2.5.1 Pharmacological HDACi and DNMTi

Vorinostat, belinostat and TSA are some of the HDACi that have been investigated in human lymphoid malignancies. Treatment with vorinostat led to increased expression of p21, acetylation of histone H3, cell cycle arrest and dose dependent cell death in rituximab-resistant DLBCL cell lines and cells isolated from patients with refractory DLBCL.17 Vorinostat also re-sensitized rituximab-resistant cell lines to the cytotoxic effects of cisplatin, gemcitabine and etoposide.17 Trichostatin A and vorinostat induced apoptosis in 7 of 8 human DLBCL cell lines.41 Additionally, over expression of several anti-apoptotic proteins (BCL-2, BCL-XL, MCL-1) decreased the sensitivity to HDACi, while overexpression of pro- apoptotic proteins (BIM) increased the sensitivity to HDACi.41 Another study identified 9 genes recurrently hypermethylated in chemoresistant DLBCL in people.42 The prolonged exposure to low dose decitabine, a 8

DNMTi, reversed this chemoresistance, and cells regained sensitivity to doxorubicin. Reactivation of the gene encoding for SMAD1 (an intermediate of the transforming growth factor β pathway), appeared critical for this chemosensitization, and an association between SMAD1 protein abundance and sensitivity to doxorubicin was noted.42 Belinostat showed synergism with decitabine in cells derived from patients with T-cell lymphoma, both in vitro and in vivo.12 Alisertib, a selective Aurora A kinase inhibitor, showed a synergistic cytotoxicity with romidepsin in T-cell lymphoma both in vitro and in xenografts models, likely secondary to cytokinesis failure.43 The anti-metabolite 6-thioguanine caused DNMT1 protein degradation in the ubiquitin-proteasome pathway, subsequent genomic demethylation, and reactivation of various epigenetically silenced gene in human T-cell leukemia lines.44 Therefore, HDACi and DNMTi have the potential to exert tumor suppressor effects in neoplastic lymphoid cells, reverse chemoresistance, and be synergistic with various chemotherapy protocols.

These encouraging pre-clinical results logically led to numerous clinical trials. A recent phase 2 trial using the oral HDACi abexinostat in refractory non-Hodgkin’s lymphoma and chronic lymphocytic leukemia patient showed an overall response rate of 28%, with the highest response rate noted in FL (56%), T-cell lymphoma (40%) and DLBCL (31%).45 In another study, azacitidine was administered before each cycle of R-CHOP (Rituximab + CHOP) in human naïve DLBCL.15 and a complete response was noted in 11 of 12 patients. Samples taken pre- and post-azacitidine treatment revealed global DNA demethylation, reactivation of the TGF-β pathway, and increased sensitivity to chemotherapy. The recent approval of vorinostat and romidepsin for the treatment of peripheral and cutaneous T-cell lymphoma is another proof that pharmacological HDACi and DNMTi are promising new treatments for various types of lymphoid neoplasms in people.12

2.5.2 Sulforaphane

In addition to the previously described pharmacological HDACi and DNMTi, the use of natural compounds as epigenetic-targeted therapies has also been investigated. In particular, SFN has been subjected to multiple recent studies in human lymphoid malignancies. T-cell leukemia cells treated with SFN exhibited cell cycle arrest, p53-mediated apoptosis, and necrosis.32 When PBMC and bone marrow samples from children with B or T-cell acute lymphoid leukemia (ALL) were exposed to SFN, dose- dependent apoptosis was observed including activation of caspases 3, 8, and 9, and inactivation of Poly (ADP-ribose) polymerase.30 A decrease in total and phosphorylated AKT and m-TOR proteins were also noted. Interestingly, the non-malignant PBMCs were much less sensitive to the effect of SFN, suggesting a differential effect of SFN in neoplastic and non-neoplastic cells.30 In the same study, oral supplementation 9

with SFN hampered tumor development in mice with ALL xenografts when compared to mice treated with the control vehicle.30 Treatment of human multiple myeloma cell lines and cells from primary tumor samples with SFN increased phosphorylation of histone H3, altered the phosphorylation status of c-jun, HSP27, Akt, and p53, cleaved caspase 3 and 9, down regulated anti-apoptotic proteins (Mcl-1, survivin), and induced apoptotic cell death.31 Moreover, SFN enhanced in vitro cytotoxicity of several anti-cancer agents commonly used to treat multiple myeloma such as bortezomib, lenalidomide, doxorubicin and melphalan. 31 In the same study, oral supplementation with SFN also significantly decreased tumor volume and increased overall survival of murine xenograft models. This research indicates that “natural” epigenetic targeted therapies such as SFN, as well as pharmacological HDACi and DNMTi, are exciting new treatment considerations for various lymphoid neoplasias.

2.6 Targeted epigenetic therapy in veterinary oncology

Similar to research available in human LSA, there is evidence for epigenetic aberrations in canine LSA. Hypermethylation of CpG islands in the promoter of Deleted in Liver Cancer 1 (DLC1, a tumor suppressor gene) has been noted in canine LSA.46 Sato et al demonstrated that in 45% of canine high grade B-LSA, the CpG island of the death-associated protein kinase (DAPK), a tumor suppressor involved in various apoptotic pathways, was hypermethylated. This change was associated with a significant decrease of progression-free survival and overall survival time (all dogs were treated with a CHOP protocol).47 Interestingly, hypermethylation of DAPK promoter was had previously been noted in B-LSA cells but not in T-LSA cells or in lymphocytes from healthy Beagles.48 Other investigators reported hypermethylation and repression of the gene encoding for the p16 protein in a canine B-LSA cell line.49 Another study also documented significant changes in the DNA methylation pattern of canine LSA cell lines. Specifically, the CpG sites within the promoters of hundreds of genes (for example TWIST2 and TLX3) were hypermethylated, while the CpG sites outside promoter CpG islands were hypomethylated.50 Ferraresso et al. showed that in 77% of dogs with DLBCL the promoter region of Tissue Factor Pathway Inhibitor 2 (TFPI2, a tumor suppressor gene) was hypermethylated.51 In the promoter of this gene, 23 CpG sites were evaluated and 82% were hypermethylated at a level 2-120 fold higher in dogs with DLBCL compared to control dogs. 51 These authors also showed a positive correlation between increasing TFPI2 promoter methylation and increased age.51 Another study demonstrated a global genomic hypomethylation in the majority of canine LSA samples, and in approximately one third of canine leukemia samples.52 Additionally, recent work has demonstrated that drug resistant lymphoid tumor cell lines expressed a higher level of H3 acetylation and decreased methylation in the CpG island of the ABCB1 gene promoter. This 10

leads to increased transcription of ABCB1 mRNA, when compared to drug sensitive lymphoid tumor cell lines.53 In summary, many oncogenes and tumors suppressor genes display aberrant methylation and acetylation patterns in canine LSA, which paves the way for the exploration of the use of epigenetic therapy in this common cancer.

2.6.1 Pharmacological HDACi and DNMTi

Although fewer investigations have been performed in veterinary oncology as compared to people, various pharmacological HDACi or DNMTi such as vorinostat, TSA, and some experimental drugs, have recently been studied in relation to several canine cancers. Nine different canine cancer cells lines were treated with vorinostat or OSU-HDAC42 (an investigational HDACi), and apoptosis was noted in all.54 Cell lines of mast cell tumor and T-cell lymphoma were the most sensitive. Trichostatin A induced a significant reduction in cell number (96% at 72 hours), and induced apoptosis in cells from a dog with a grade III mast cell tumor.55 AR-42 (a pan HDACi) induced downregulation of the KIT receptor, and death of normal and malignant mast cells in canine and mouse mast cell tumor cell lines.56 AR-42 also led to hyper-acetylation of H3/H4 and alpha tubulin, upregulation of p21, and downregulation of p-AKT and p-STAT3/5.56 Another recent in vitro study evaluating the nucleoside analogues 6-thioguanine and zebularine showed that these drugs cause a significant downregulation in the enzyme DNMT1, decrease in global DNA methylation, and induced caspase-mediated apoptosis in canine T-cell leukemia and B-cell chronic lymphocytic leukemia lines.57 Finally, Sato et al. showed that the demethylating agent 5-aza-2’-deoxycytidine was able to reverse hypermethylation and allow re-expression of the tumor suppressor DAPK in a B-LSA cell line.48 Together, these studies demonstrate that drugs targeting the epigenome can elicit in vitro anti-cancer effects in various canine neoplastic cells, including B- and T-LSA cell lines.

Only few studies have reported the use of HDACi or DNMTi in cancer-bearing dogs. In a recent study, the oral administration of zebularine to purpose-bred laboratory and tumor bearing dogs was well tolerated when given once every 3 weeks. When given daily on the other hand, dogs developed grade 3 and 4 neutropenia, and various cutaneous toxicities. Two dogs with bladder carcinoma experienced stable disease for 3-4 months.58 Hahn et al. showed that daily subcutaneous injection of 5-azacytidine in dogs with urothelial carcinoma was well tolerated, with partial response and stable disease noted in 22% and 50% of dogs, respectively.59 One dog with splenic hemangiosarcoma was treated with vorinostat post- splenectomy, and lived > 1,000 days with no signs of disease recurrence.60 In addition, the role of valproic acid combined with doxorubicin was investigated in 21 dogs with various tumor types. Two patients with lymphoma had a complete response, and one patient with lymphoma had a partial response.61 Hyper- 11

acetylation of histone H3 was identified in PBMCs, and was positively associated with valproic acid dose.61 Therefore, similar to people, pharmacological HDACi and DNMTi appear promising for the treatment of both solid and hematopoietic tumors in dogs.

2.6.2 Sulforaphane

To our knowledge, published research focusing on SFN in veterinary medicine consist of only one bioavailability study in healthy dogs25 and one in vitro study in osteosarcoma cells.34 In the later, SFN significantly reduced cell invasion and decreased focal adhesion kinase (FAK) expression and phosphorylation in three canine osteosarcoma cell lines.34 No effects were noted on extracellular regulated kinase (ERK), jun-n terminal kinase (JNK), AKT, caspase 3 and p53.34 The same authors evaluated the impact of SFN on osteosarcoma and mast cell tumor cell lines,62 and noted reduction in cell viability in all cell lines, and increased apoptosis in one mast cell tumor and one osteosarcoma cell line at higher dose of SFN. Sulforaphane did not promote further cell death when combined with doxorubicin or toceranib. Further investigation of the use of SFN in cancer bearing dogs is warranted.

2.7 Significance and aims of the study

We described in the previous sections that:

1. Histone deacetylase inhibitors and DNMTi show promising results in the treatment of human lymphoid neoplasia, and some HDACi have been approved for the treatment of various forms of LSA.

2. Studies investigating pharmacological HDACi such as valproic acid or DNMTi such as decitabine, have also reported promising results in canine neoplasia, including LSA.

3. Sulforaphane possesses broad cancer preventive and cancer suppressive effects due to HDACi and DNMTi activity, as well as interference with many oncogenic pathways.

4. Sulforaphane shows encouraging results specifically in various human lymphoid neoplasms. It induces cell cycle arrest and apoptosis in lymphoma, lymphoid leukemia and multiple myeloma cells in vitro. In mice xenograft models of ALL or multiple myeloma, SFN slowed tumor progression, enhanced the efficacy of chemotherapy, and prolonged survival.

5. Sulforaphane is absorbed in healthy dogs, metabolites are detectable in plasma and urine after a single dose, and some HDACi activity has been detected in PBMCs. 12

However, the work related to pharmacological HDACi/DNMTi remains limited in veterinary oncology. To our knowledge, only three studies and one case report in relation to this topic have been reported in tumor-bearing dogs 58-61. Moreover, the impact of SFN in cancer-bearing dogs has not been previously explored. In addition, only 5 proteomic profiling studies have so far been performed in canine LSA.63-67 In most of these studies, only serum proteins were analyzed, while lymph nodes samples were used in only one study.65 Moreover, proteomic analyses were performed with either surface-enhanced laser desorption ionization time-of-flight (SELDI-TOF) mass spectrometry (MS) or matrix-assisted laser desorption/ ionization (MALDI)-TOF MS. The combination of electrospray ionization (ESI) and liquid chromatography-tandem mass spectrometry (LS-MS/MS) was not used in any study. Advantages of ESI include a soft ionization process with very little fragmentation of macromolecules, the possibility to analyze large peptides (up to several hundreds of kilodaltons), and a high detection limit (in the range of 10-15 to 10- 18 mole) due to a low chemical background.68

Therefore, this study was designed to investigate the effects of SFN in a small cohort of cancer-bearing dogs using ESI and LC-MS/MS. Our primary objective was to analyze the changes in the lymph node proteome of dogs with multicentric LSA pre- and post-SFN supplementation. A secondary objective was to determine if a difference in the proteomic profile of dogs with B-cell vs. T-cell LSA could be detected. We hypothesized that SFN supplementation would result in a decrease in the expression level of various oncoproteins, while an increase in the expression level of some tumor-suppressor proteins would be noted. We also hypothesized that many proteins would be differentially expressed in the lymph nodes of dogs with B-LSA when compared to T-LSA at the time of initial diagnosis.

3. Materials and methods

3.1 Patients recruitment

This study was carried out with approval of the Oregon State University (OSU) Institutional Animal Care and Use Committee, Clinical Review Board (OSU IACUC office number 4893). Dogs were screened for inclusion at the OSU Veterinary Teaching Hospital (OSU-VTH) based on physical examination, complete blood count, serum chemistry, urinalysis, lymph node cytology, and immunophenotyping. All screening tests occurred within 1 week of enrollment. To be enrolled, each dog had to be newly diagnosed with LSA (via cytology or histopathology), and have flow cytometry or immunocytochemistry performed for immunophenotyping. Other inclusion criteria included a body weight >15 kg, clinically healthy at diagnosis (substage a, no hypercalcemia), adequate organ function (absolute neutrophil count >1500 cells/L; hematocrit >25%; platelets >75,000/L; creatinine <2x the upper limit of reference interval (RI); bilirubin 1.5x the upper limit of RI; ALT  3 X the upper limit of RI), a Veterinary Cooperative Oncology Group (VCOG) performance status < 2 69 , and at least 2 lymph nodes > 2 cm in length. Exclusion criteria included patient that had received any previous treatment for lymphoma (including corticosteroids), patients with any concurrent malignancy or other serious systemic disorder, and patients that had been taking homeopathic/alternative therapies within 3 days of enrollment (in particular any food, vegetable treats or supplements containing cruciferous vegetables such as broccoli, Brussel sprouts, cauliflower and cabbage, were excluded). Supplements such as chondroitin sulfate, glucosamine, vitamins, and essential fatty acids were permitted. Owners were directed to feed their pet their usual dog food only, and avoid any treats (including vegetables) for 3 days prior to, and continuing for the duration of the trial. Signed informed consent was obtained from all owners before study enrollment.

3.2 Study design and sample processing

On day 0 (D0), pre-SFN supplementation lymph node fine needle aspirates (FNA) were collected. A total of 3 FNA were obtained from at least 2 different peripheral lymph nodes from each dog. When possible, collection from mandibular lymph nodes were avoided. All patients subsequently received 90 mg (3 capsules) of BroccoMaxTM twice daily by mouth for seven days. This dose was based on the bioavailability study performed by our research team in healthy dogs25 and on a similar dosing regimen utilized in a human breast cancer trial.70 Any form of treatment for lymphoma (including prednisone) was postponed until the end of the study. However, if LSA progression or deterioration of the patients’ clinical status were noted during the course of SFN supplementation, owners were given the option to remove their 14

pet from the study and start prednisone +/- chemotherapy. On day 7 (D7), post-SFN supplementation, duplicate lymph node samples were obtained. After sample collection on D7, patients were deemed off- study and were free to pursue mainstay chemotherapy treatment.

At each time point, material obtained from the three FNA samples were mixed together with 1 ml of 0.9 % sterile saline in a cryovial. Then 400 µl of the sample and saline mixture were removed and were added into another cryovial with 100 µl of RIPA protein lysis buffer (150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholic acid (NaDOC), 0.1% SDS, 20 mM TRIS pH 8.0). Two cryovials with 400 µl lymph node samples + 100 µl Ripa were prepared, flash frozen in liquid nitrogen, and stored at -80oC until further analysis.

All adverse events (AEs) occurring during SFN supplementation were recorded and graded according to the Veterinary Cooperative Oncology Group (VCOG) common terminology criteria for adverse events.69 Although this was not the main aim of the study, we also assessed the clinical response to SFN at D7 via caliper measurement of peripheral lymph nodes. According to the VCOG response evaluation criteria for peripheral nodal lymphoma in dogs71, a complete response (CR) was defined as the disappearance of all pathologically enlarged lymph nodes. A partial response (PR) was defined as at least a 30% reduction in the mean sum of the longest diameters of target lymph nodes. Progressive disease (PD) was defined as at least a 20% increase in the mean sum of the longest diameters of target lymph nodes, or the appearance of one or more new lesions. Stable disease (SD) was defined as neither a sufficient decrease to qualify as PR, not a sufficient increase to qualify as PD.

3.3 Mass spectrometry

The material from the matched lymph nodes samples of 7 dogs at D0 and D7 (14 samples total) were analyzed by mass spectrometry (LC-MS/MS) at the Oregon State University Mass Spectrometry Center. The protein concentration of each lymph node sample was determined using PierceTM BCA protein assay kit (Thermo Fisher Scientific, Rockford, IL, USA) according to the manufacturer’ s protocol. The proteins were digested by sequencing grade modified trypsin following a protocol provided by Promega (Promega Corporation, Madison, WI). Peptide analysis was achieved using an Orbitrap Fusion Lumos mass spectrometer with a Nano ESI source (Thermo Scientific, Waltham, MA) coupled with a Waters nanoAcquity UPLC system (Waters, Milford, MA). The proteolytic products were desalted and loaded on a nano Acquity UPLC 2 G Trap Column (180 μm × 20 mm, 5 μm) for 5 min with solvent 0.1% formic acid in 3% ACN at a flow rate of 5 μL/min. An nanoAcquity UPLC RPeptide BEH C18 column (100 μm × 100 mm, 1.7 μm) was applied to separate peptides following by a 120-min gradient consisting of 0.1% formic 15

acid in H2O (mobile phase A) and 0.1% formic acid in ACN (mobile phase B), where B was increased from 3% to 10% at 3 min, 10%  30% at 105 min, 30%  90% at 108 min and held 4 min, and then decreased to 3% at 113 min and held until 120 min. The LC flow rate was set at 500nL/min. All mass spectral data were acquired in the positive ion mode. The spray voltage was 2400 V and the ion transfer tube temperature was 300 °C. MS and MS/MS spectra were acquired by the Orbitrap analyzer (resolution 120 K at m/z 200) and Ion Trap (collision induced dissociation CID) respectively. Automatic gain control target was set to 4.0 × 105 for precursor ions and 104 for product ions. Mass tolerances were set at ± 10 ppm for precursor ions and 0.6 Da for fragment ions.

All raw data files were analyzed with Thermo Scientific Proteome Discoverer 2.2 software and searched initially against the Uniprot Canis database using Sequest HT as search engine. A maximum of two missed cleavage sites was allowed. Carbamidomethylation of cysteine and oxidation of methionine were specified as static modification and dynamic modification, respectively. The overall false discovery rate (FDR) at the protein level was less than 1%. To allow GO annotation analysis the datasets were also searched against the Uniprot Homo sapiens protein database. The proteins of interests still uncharacterized after searching against both canine and human database were submitted to a Basic Local Alignment Search Tool (BLAST) analysis in Uniprot website (https://www.uniprot.org/blast/). Only canine proteins with more than 99% similarities with its human ortholog were retained.

The fold change (FC) of a given protein was defined as the ratio of abundance between 2 time points (D7 vs. D0) or between 2 groups (B vs. T-cell lymphoma). To calculate the FC of a given protein, each peptide group ratio was first calculated as the geometric median of all combinations of ratios from all the replicates in the same group. Secondly, the protein ratio was subsequently calculated as the geometric median of the peptide group ratios. Overall, the FC D7/D0 for protein X reflects the ratio of abundance of protein X across all dogs at D7, relative to the abundance of protein X in dogs at D0. The proteins FC between the following groups was investigated: (1) all dogs at D0 vs. all dogs at D7; (2) each dog individually at D0 vs. D7; and (3) B-cell lymphoma at D0 vs. T-cell lymphoma at D0. Only proteins with a significant FC between groups (p-value <0.05) and identified in at least half of the biological replicate (i.e. at least 4 out of 7 dogs at D0 and D7) were selected for further analysis.

3.4 Immunoblots Immunoblot analyses were performed for selected proteins in order to validate the LC-MS/MS results. The 14-3-3- θ and HIP proteins were chosen as they had shown differential expression between D0 and D7 according to the LC-MS/MS results, and because the antibodies targeting these proteins had either 16

previously been validated in dogs, or were predicted to be cross-reactive with the canine protein according to the manufacturer. The protein concentration of each lymph node sample was assessed using PierceTM BCA protein assay kit (Thermo Fisher Scientific). The samples were mixed at a volume ratio of 4:1 with NuPAGETM LDS sample buffer (Thermo Fisher Scientific) and 2.5% β-mercaptoethanol. Proteins were denatured by boiling at 95oC for 10 minutes, and immediately placed on ice. The samples volume was calculated to load 50 µg of proteins per well. Proteins were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS–PAGE) on a 10% SDS-PAGE gel at 100 Volts for 2 hours. The separated proteins were then transferred onto a nitrocellulose membrane at 100 Volts for 1.5 hour. Membranes were washed with TBS (20 mM Tris–HCl pH 7.4 and 150 mM NaCl) and blocked overnight at 4 °C with Odyssey blocking buffer (LI-COR, Lincoln, NE, USA) mixed 1:1 with TBS. Blots were incubated overnight at 4 °C with a goat polyclonal anti-14-3-3-θ antibody (Thermo Fisher Scientific, Rockford, IL, USA) diluted 1:5000 with TBST (TBS with 0.1 % Tween-20). Replicate blots were probed with a rabbit polyclonal anti-HIP (Heat Shock 70 interacting protein) antibody (Novus Biologicals, Centennial, CO, USA) diluted 1:5000 with TBST. Blots were then washed three times for 10 minutes in TBST followed by TBS, and incubated for 1 hour at room temperature with the corresponding secondary antibodies: a horseradish peroxidase (HRP)-labeled donkey anti-goat polyclonal antibody (Santa Cruz Biotechnology, Dallas, TX, USA) diluted 1:1000 in TBST, and a biotin-SP-conjugated goat anti- rabbit polyclonal antibody (Jackson ImmunoResearch, West Grove, PA, USA) diluted 1:20000 in TBST followed by a HRP-conjugated streptavidin (Jackson ImmunoResearch) diluted 1:1000 in TBST. All blots were washed three times for 10 minutes at room temperature with TBST followed by TBS, and exposed to luminol and peroxide (ECLTM Prime Western Blotting Detection Reagent; GE Healthcare Life Sciences, Marlborough, MA, USA) mixed at a volume ratio of 1:1, and visualized using the ImageQuantTM LAS 4000 scanning system (GE Healthcare Life Sciences). Membranes were subsequently washed in TBS and incubated for 15 minutes at room temperature with RestorTM PLUS Western Blot stripping buffer (Thermo Fisher Scientific), blocked overnight at 4°C with Odyssey buffer mixed 1:1 with TBS, and incubated at room temperature for 2 hours with a mouse monoclonal anti-β-actin primary antibody (Santa Cruz Biotechnology) diluted 1:200 with TBST. After washing in TBST and TBS, membrane were incubated at room temperature for 1 hour with a HRP-labelled goat anti-mouse secondary antibody (Santa Cruz Biotechnology) diluted 1:5000 with TBST, and revealed with luminol as previously described. The β-actin was used as a standard to normalize the expression level of 14-3-3-θ and HIP. Briefly, after subtracting the background intensity, the volume of each band of the protein of interest and β-actin was 17

determined using the ImageQuant TL 8.2 software (GE Healthcare Life Sciences). A normalized expression of the protein of interest was calculated by dividing the volume of the protein of interest band by the volume of the corresponding actin band. The normalized expression ratio (NER) D7/D0, defined as the normalized expression of the protein of interest at D7 divided by the normalized expression of the protein of interest at D0, was then calculated for each dog.

3.5 Gene Set Enrichment Analysis

We used the R statistical computing environment (version 3.3.3) to generate ranked lists of genes from the protein abundance ratio data, in three steps. First, we mapped canine proteins (identified by UniProtKB accession numbers) to canine Ensembl protein identifiers using Ensembl BioMart (release 94), and for any identifiers that failed to map by BioMart, we secondarily used the Bioconductor (version 3.4) package org.Cf.eg.db (version 3.7.0) which is based on Entrez Gene (version 2018-Oct11) to map identifiers. Second, we mapped canine Ensembl protein identifiers to human Ensembl protein identifiers using the Bioconductor (version 3.4) package hom.Hs.inp.db (version 3.1.2), which uses the Inparanoid ortholog database (version 8.0). Third, we mapped human Ensembl protein identifiers to Entrez gene identifiers using Ensembl BioMart, and for any identifiers that failed to map by BioMart, we secondarily used the Bioconductor (version 3.4) package org.Hs.eg.db (version 3.7.0) which is based on Entrez Gene (version

2018-Oct11). We log2 transformed the protein expression ratios for each protein and programmatically generated a Gene Set Enrichment Analysis (GSEA) ranked gene (.rnk) file containing human Entrez Gene identifiers.

We analyzed the ranked gene files using GSEA (version 3.0) in GSEAPreranked mode, using gene sets H, C2, C3, and C5 from MSigDB (version 6.2, http://software.broadinstitute.org/gsea/msigdb) and with a minimum set size of 10. Only gene sets with a false discovery rate (FDR) q < 0.1 and a nominal p- value < 0.01 were retained for the final analysis.

4. Results

4.1 Patients population

Seven dogs were prospectively enrolled between March 2017 and March 2018. Details regarding patient demographics are provided in Table 1. Median and mean age at the time of enrollment were 6 and 6.5 year old, respectively. One dog had a history of idiopathic hematuria and sebaceous adenoma. All patients were diagnosed with intermediate or large cell multicentric lymphoma via fine needle aspiration, and all were attributed a clinical stage IIIa at the time of enrollment. All dogs were immunophenotyped, either via flow cytometry (n=6) or immunocytochemistry (n=1). There were four B-cell lymphomas, two T-cell lymphomas and one B-cell lymphoma with possible concomitant emerging T-zone lymphoma.

Table 1. Characteristics of seven dogs enrolled in a study investigating the impact of oral sulforaphane supplementation in canine multicentric lymphoma. M: intact male; MHCII: Major Histocompatibility Complex class II; NM: neutered male; SF: spayed female

Patient # Breed Sex Age Cell size Stage Immunophenotype

1 Husky SF 6 Large IIIa T-cell (CD3+ CD4+)

2 Australian Shepherd NM 6 Intermediate IIIa T-cell (CD5+ CD4- CD8-) B-cell (CD21+, low 3 Pitbull NM 5 Intermediate IIIa MHCII) B-cell (CD21+, high 4 Labrador NM 4 Intermediate IIIa MHCII) 5 Beagle SF 13 Intermediate IIIa B-cell (CD79a+) B-cell (CD21+, high 6 Australian Shepherd M 3 Large IIIa MHCII) B-cell (CD21+, high 7 English Shepherd SF 9 Large IIIa MHCII) +/- TZL (CD5+ CD45- CD4- CD8-)

4.2 Clinical response to sulforaphane and outcome

All but one dog received the planned 14 doses of SFN (90 mg twice a day for 7 days). One dog received 13 doses of 90 mg and only received 30 mg (1 capsule instead of 3) at the last dose due to administration error. There were no other dose delays or modifications in SFN supplementation. There was a total of 24 episodes of AEs, all of them grade 1 or 2. All but one dog experienced AEs, most commonly lethargy, vomiting and anemia. More details regarding the AEs are provided in Table 2. None of the AEs were 19

directly attributed to SFN. Most were deemed most likely related to the patients’ LSA, as 33% of AEs had been identified prior to the study, and the most significant adverse events occurred in dogs experiencing progressive disease while receiving SFN.

Table 2. Summary of adverse events experienced by canine patients with naïve multicentric lymphoma during seven days of oral supplementation with sulforaphane

Adverse Event Grade 1 Grade 2 Grade 3 Grade 4 Grade 5 Total

Lethargy 3 1 4

Anemia 3 3

Vomiting 3 3

Diarrhea 1 1 2

Anorexia 1 1 2

Thrombocytopenia 2 2

Increase ALT 2 2

Increase ALP 2 2

Lameness 1 1

Seroma (lymph 1 1 node biopsy site)

Pain (lymph node 1 1 biopsy site)

Polyuria 1 1

No dogs received any form of LSA-directed treatment before enrollment or while supplemented with SFN. According to the VCOG criteria71, no dogs experienced an objective response (i.e. ≥30% reduction in the lymph nodes’ longest diameters). Four dogs developed progressive disease (+23%, +28%, and +29% in the mean sum of lymph nodes’ longest diameters, as well as one dog with 10% increase and development of edema of the neck and limbs) and three dogs experienced stable diseases (-15%, +5%, and +10%). 20

After the end of the 7 days of SFN supplementation, 6 dogs were started on a CHOP protocol. All dogs eventually relapsed, and were rescued with a variety of chemotherapy protocols including lomustine (n=4 dogs), rabacfosadine (3), mitoxantrone single agent (3), L-asparaginase (3), a protocol of vincristine, melphalan, and cytosine arabinoside (2), and doxorubicin single agent (1). The mean and median number of rescue therapies pursued per dog were 2.7 and 2.5, respectively (range: 1 to 5). One dog was treated with L-asparaginase and prednisone only at the end of SFN supplementation. At the time of writing, 6 dogs had died or were euthanized for lymphoma-related causes at 42, 153, 247, 338, 344 and 387 days post-diagnosis, while dog #3 was still alive and treated with rescue chemotherapy.

4.3 Mass spectrometry results: all dogs at day 7 vs. all dogs at day 0

4.3.1 Proteins differentially expressed between day 0 and day 7

A total of 915 proteins were identified from all lymph nodes samples, and statistical analysis discovered 50 proteins with significant FC between D7 and D0 group (p < 0.05). Twenty-four of the 50 proteins were identified in more than 50% biological replicates (i.e. at least 4 out of 7 dogs at each time point) and were selected for further analysis. The amino acid sequence coverage for these 24 proteins ranged from 3% to 84%, with a mean coverage of 10.8% and a median coverage of 6%.

Among these 24 proteins, 14 (58%) were significantly downregulated at D7 when compared to D0 (mean FC D7/D0 = 0.318; median = 0.328; range = 0.131 to 0.482), and 10 proteins (42%) were significantly upregulated at D7 when compared to D0 (mean FC D7/D0 = 3.80; median = 3.76; range = 2.99 to 5.34). In other words, 14 proteins were expressed at a level ranging from 2.1 (1/0.482) to 7.6 (1/0.131) times lower after a week of SFN supplementation, and 10 proteins were expressed at a level ranging from 3 to 5.3 times higher after a week of SFN supplementation.

Six of the 24 proteins were identified as uncharacterized when search against the Canis lupus familiaris database. We used the Basic Local Alignment Search Tool (BLAST) of the Uniprot website (https://www.uniprot.org/blast/) to search for ortholog proteins of the Homo sapiens database. We identified the human proteins with the highest number of hits for each uncharacterized canine protein, and kept proteins with 99% similarities between the 2 species. In this way, we were able to identify 3 additional canine proteins, for a total of 21 of 24 proteins precisely identified (12 downregulated at D7, 9 upregulated at D7). A description of these proteins is provided in Table 3 and Table 4.

Table 3. Characteristics of twelve proteins significantly downregulated after seven days of oral sulforaphane supplementation in dogs with multicentric lymphoma. Information related to human neoplasias are identified with (H); information related to canine neoplasia are identified with (d). EBV: Epstein-Barr virus; FC: fold change; HDACi: histone deacetylase inhibitor; LSA: lymphoma; PDGFR: platelet-derived growth factor receptor; PI(3,4)P2: phosphatidylinositol-3,4-bisphosphate; PI3K: phosphatidylinositol 3-kinase; PKC: protein kinase C; MGT: mammary gland tumor; mRNA: messenger ribonucleic acid; RCC: renal cell carcinoma; rRNA: ribosomal ribonucleic acid; tRNA: transport ribonucleic acid; TCC: transitional cell carcinoma; TSA: trichostatin A.

Relevant biological functions / role in canine and human FC p- Protein down- 72-96 regulated at D7 neoplasia D7/D0 value tRNA-splicing  Joins spliced tRNA to mature-sized tRNA 0.131 5.19 x ligase RtcB  Embryonic and placenta development 10 -11 homolog  High expression: negative prognostic factor in RCC (H) (RTCB) Lamin A/C  Nuclear stability, chromatin structure, gene expression 0.162 5.50 x (LMNA)  Protein import into nucleus 10 -9  Negative regulation of apoptosis  Cell migration  Cellular response to hypoxia  Low expression: negative prognostic factor in cervical, colorectal, breast and ovarian carcinoma (H)  Cleavage of LMNA by caspase 6 is crucial to induce apoptosis in B-LSA cells (H)  TSA restored LMNA expression in cervical cancer cells (H)  Positively associated with malignant phenotype of prostate cancer cells though PI3K/AKT/PTEN pathway (H) 14-3-3 protein  14-3-3 isoforms interact with > 200 proteins 0.18 4.29 x theta (14-3-3-θ)  Modulate cell cycle, apoptosis, transcription, stress 10-8 response and many signaling pathways via recognition of phosphoserine/phosphothreonine  High expression: negative prognostic factor in RCC, liver, and endometrial carcinoma (H)  High expression predictive of chemotherapy resistance in breast cancer cells (H)  Aberrant expression of 14-3-3-σ in MGT, RCC, and highly infiltrative bladder TCC (d)  14-3-3-σ: negative prognostic factor in RCC (d)

Table 3. (Continued)

Relevant biological functions / role in canine and human FC p- Protein down- 72-96 regulated at D7 neoplasia D7/D0 value Thioredoxin  Redox homeostasis related  Protein folding 0.283 1.14 x transmembrane  Response to endoplasmic reticulum stress 10-4 protein 1 (TMX1)  High expression: negative prognostic factor in liver carcinoma (H)  Thioredoxin = “adult T-cell leukemia derived factor” o Upregulated in metastatic MGT (d) o Involved in resistance to doxorubicin in T-cell leukemia cell line (H) 60S ribosomal  Initiates and promotes protein translation 0.324 7.25 x protein L12 10-4 (RPL12)  High expression: negative prognostic factor in RCC and head and neck cancer (H)  Viral RPL12 induces LSA in chicken Chromosome 7  Biologic function not characterized 0.332 1.01 x open reading 10-3  High expression: negative prognostic factor in pancreatic frame 50 carcinoma (H) (C7orf50) X-prolyl  Protein homodimerization 0.39 6.73 x aminopeptidase 1  Maturation of kinins, neuropeptides, and peptide 10-3 (XPNPEP1) hormones  High expression: negative prognostic factor in liver carcinoma (H) Pleckstrin (PLEK)  B-cell receptor signaling pathway 0.436 0.021  Binding domains for PI(3,4)P2 and PKC  PI3K signaling  PDGFR signaling  Androgen and estrogen metabolism  Cellular response to hydrogen peroxide  Upregulated in B-lymphocytes transformed by EBV (H)  High expression: negative prognostic factor in RCC (H)

Small nuclear  mRNAs splicing (via spliceosome) 0.459 0.031 ribonucleoprotein  Histone mRNA processing F (SNRPF)  Termination of transcription  High expression: negative prognostic factor in RCC (H) Splicing factor 3a  mRNAs splicing (via spliceosome) 0.462 0.032 subunit 2 (SF3A2)  Microtubule binding  High expression: negative prognostic factor in RCC colorectal and liver carcinoma (H) 23

Table 3. (Continued)

Relevant biological functions / role in canine and human FC p- Protein down- 72-96 regulated at D7 neoplasia D7/D0 value Alanyl-tRNA  Catalyzes the attachment of alanine to tRNA 0.476 0.042 synthetase  Regulation of translational fidelity (AARS)  Negative regulation of apoptosis  No role found in human or dog neoplasia Proteasome 26S,  Degradation of ubiquitinated proteins 0.482 0.047 non-ATPase  Removal of misfolded or damaged proteins regulatory subunit  TNF signaling pathway 2 (PSMD2)  Cell cycle progression  DNA damage repair  High expression: negative prognostic factor in head and neck cancer, liver and lung carcinoma (H)  Drugs targeting PSM 26S (i.e. bortezomib) used alone or combined with HDACi to treat various type of LSA (H)  Low PSM 26S expression associated with increase RT sensitivity and reduced recurrence rate of recurrence in breast cancer (H)

Table 4. Characteristics of nine proteins significantly upregulated after seven days of oral sulforaphane supplementation in dogs with multicentric lymphoma. Information related to human neoplasias are identified with (H); information related to canine neoplasia are identified with (d). CSC: cancer stem cell; EGF: epidermal growth factor; EGFR: epidermal growth factor receptor; EMT: epithelial to mesenchymal transition; FC: fold change; HSP: heat shock proteins; OSA: osteosarcoma; MDR: multidrug resistance; MGT: mammary gland tumor; NADP: nicotinamide adenine dinucleotide phosphate; RCC: renal cell carcinoma; TCC: transitional cell carcinoma; TGF-β: transforming growth factor beta.

Protein up- Relevant biological functions / role in canine and FC p- regulated at D7 human neoplasia 72-77,97-112 D7/D0 value

C-type lectin  Cellular response to TGF-β 5.34 2.88 x domain family 3 10-5  High expression: positive prognostic factor in head and member B neck cancer, RCC, pancreatic and liver carcinoma (H) (CLEC3B) Coatomer subunit  Anterograde, retrograde, and intra-Golgi vesicle- 4.642 2.28 x gamma (COPG) mediated protein transport 10-4  Retrograde transport of EGFR in the nucleus  Lipid homeostasis  High expression: negative prognostic factor in liver carcinoma (H) Aldehyde  Detoxification of endogenous and exogenous 3.988 1.79 x dehydrogenase aldehydes 10-3 family 16 member  Metabolism of chemotherapeutics A1 (ALDH16A1)  Resistance to cyclophosphamide  Maintenance of CSC phenotype  High expression: negative prognostic factor in urothelial carcinoma and neuroblastoma (H) ATPase H+  Cellular response to increased oxygen levels 3.808 3.14 x transporting V1  Acidification of intracellular organelles, zymogen 10-3 subunit A activation (ATP6V1A)  Protein sorting  Receptor-mediated endocytosis  Iron homeostasis

 High expression: positive prognostic factor in RCC (H) Inter-alpha-trypsin  Acute phase protein 3.714 3.99 x inhibitor heavy  Response to cytokines (IL6) 10-3 chain family  Protease inhibitor member 4 (ITIH4)  Platelet degranulation  Liver development and regeneration  ITIH5: tumor suppressor in breast cancer (epigenetic reprogramming of CSC phenotype) (H)  High ITIH2 expression: positive prognostic factor in glioma (H) 25

Table 4. (Continued)

Protein up- Relevant biological functions / role in canine and FC p- regulated at D7 human neoplasia 72-77,97-112 D7/D0 value

Hsc70-interacting  Mediates association of the HSP70 and HSP90 3.24 0.018 protein (HIP)  Stabilizes ATP-bound Hsc70 (high affinity conformation), favoring Hsc70 chaperone activity Alternative name:  Ubiquitination and degradation of several oncogenic Suppression of proteins tumorigenicity 13  Assembly of glucocorticoid receptor (ST13)  Downregulated in colorectal carcinoma (H)  High expression: positive prognostic factor in RCC and estrogen receptor-positive breast cancer (H) Inter-alpha-trypsin  Protease inhibitor 3.24 0.018 inhibitor heavy  Synthesis and degradation of hyaluronic acid, chain 3 (ITIH3) stabilization of extra-cellular matrix  Platelet degranulation  Unfavorable prognostic marker in RCC (H)  Anti-metastatic properties in lung carcinoma cell line (H) Dolichyl-diphospho-  Ribosome binding 3.22 0.018 oligosaccharide  Protein glycosylation protein  Response to various drugs glycosyltransferase  MDR1-mediated chemotherapy resistance subunit 2 (RPN2)  Associated with EMT and metastasis  High expression: negative prognostic factor in OSA, Alternative name: RCC, urothelial and liver cancer, breast, lung, gastric, ribophorin II colorectal, and esophageal carcinoma (H) Hemoglobin subunit  Oxygen and bicarbonate transport 2.992 0.032 alpha (HBA)  Cellular oxidant detoxification, response to hydrogen peroxide  Positive regulation of apoptosis  Receptor-mediated endocytosis  Potential serum biomarker in ovarian cancer (H)

4.3.2 Proteins detected exclusively at day 0 or day 7

As noted in the previous section, we only retained proteins identified in more than 50% biological replicates at D0 and D7. A pitfall of this filtering methods is the potential exclusion of proteins detected in dogs at one time point but in none of the dogs at the other time point, some of which can still be biologically relevant even if not statistically significant. Therefore, we investigated which proteins were detected in at least one of the dog at D0 but none of the dogs at D7, or in at least one of the dogs at D7, but in none of the dogs at D0. The results are presented in Figure 1.

Figure 1. Venn diagram depicting the amount of proteins detected at both day 0 and day 7 (n=885), detected in at least one dog pre-SFN but none of the dogs post-SFN (7), and detected in none of the dogs pre-SFN but a least one dog post-SFN (21).

Using the identification methods previously described, we were able to definitively characterized 6 of 7 proteins detected only pre-SFN and 18 of 21 proteins detected only post-SFN. All of these proteins were detected in only one or two dogs at one time point, and none of the dogs at the other time point. More details about these proteins are provided in Table 5 and 6.

Table 5. Description and main biological functions of the proteins detected in at least one dog at day 0 but none of the dogs at day 7. Ca: calcium; NAD(P)HX: nicotinamide adenine dinucleotide phosphate hydrates; NFκβ: nuclear factor-kappa beta; rRNA: ribosomal ribonucleic acid; TNF-α: tumor necrosis factor alpha.

Protein name Relevant biological function(s)72-77 Detected at D0 in: NOP14 nucleolar  rRNA processing One dog (#3) protein  Small ribosomal subunit assembly Eukaryotic initiation  RNA helicase One dog (#3) factor 4A-I  Initiation of translation H3 histone family  Nucleosome structure, replication-independent One dog (#3) member 3B histone Copine 1  Ca-dependent membrane-binding protein  Positive regulation of TNF-α signaling and NFκβ One dog (#2) signaling NAD(P)H-hydrate  Epimerization of the S- and R-forms of epimerase NAD(P)HX One dog (#4)  Interaction with apolipoprotein A-I Chromosome 7  Undetermined C18orf25 homolog One dog (#7)

The first two proteins (NOP14 nucleolar protein and eukaryotic initiation factor 4A-I ) presented in Table 5, detected in dog #3 at D0 but no longer detected at D7 in any dogs, have a major role in protein translation. No other unifying biological function between these proteins could be identified.

Table 6: Description and main biological functions of the proteins detected in none of the dogs at day 0, but at least one dog at day 7. Ca: calcium; ER: endoplasmic reticulum; GM-CSF: granulocyte-macrophage colony stimulating factor; IFN-α: interferon alpha ; Ig: immunoglobulin; IL: interleukin; NFκβ: nuclear factor-kappa beta; MHC: major histocompatibility complex; mRNA: messenger ribonucleic acid; PLC: phospholipase C; TGF-β: transforming growth factor beta; TNF-α: tumor necrosis factor alpha.

Protein name Relevant biological function(s)72-77 Detected at D7 in: DExD/H-box helicase  Positive regulation of GM-CSF, INF-α and β, Two dogs (#3, 6) 58 IL6, IL8, and TNF-α  RNA helicase with DEAD box protein motifs and caspase recruitment domain  Ubiquitin protein ligase binding Serine/threonine kinase  Negative regulation of TGF-β signaling Two dogs (#5, 6) receptor associated  Negative regulation of epithelial cell migration protein and proliferation  Negative regulation of transcription Myosin heavy chain-1  Actin fi lament binding, motor activity Two dogs (#4, 7) Exosome component 7  Ig class switch and somatic hypermutation One dog (#6)  Component of the RNA exosome complex with exoribonuclease activity  Degrade unstable mRNA  Degrade histone mRNA MutS homolog 6  Ig isotype switch and somatic hypermutation One dog (#6)  G2/M checkpoint  Interstrand cross-link repair, mismatch repair, pyrimidine dimer repair  Replication fork arrest  Intrinsic apoptotic signaling pathway Chemerin chemokine-  Macrophage chemotaxis One dog (#6) like receptor 1  Complement receptor mediated signaling pathway  Negative regulation of IL12 production and NFκβ activity  PLC signaling pathway Complement 6  Part of the complement membrane attack One dog (#6) complex  Positive regulation of angiogenesis B cell receptor  MHC class 1 protein binding One dog (#2) associated protein 31  Caspase 8-mediated apoptosis  Positive regulation of ubiquitin-dependent protein degradation  Anterograde transport of proteins from the ER to the Golgi

Table 6. (Continued)

Protein name Relevant biological function(s)72-77 Detected at D7 in: Penta-EF-hand domain  Forms heterodimer with programmed cell death 6 One dog (#6) containing 1 protein  Protein ubiquitination Host cell factor C1  Control of the cell cycle One dog (#6)  Histone acetyl transferase activity  Activation of transcription factor binding Ubiquitin like modifier  Member of the E1 ubiquitin-activating enzyme One dog (#6) activating enzyme 3 family Signal sequence receptor  Protein translocation across the ER membrane One dog (#2) subunit 3 Metaxin 1  Mitochondrial protein transport One dog (#6) Cytochrome b  Transfer of electron in the mitochondrial One dog (#6) respiratory chain, generation of ATP Tubulin alpha subunit of  Microtubules organization and mitosis One dog (#6) microtubules Centrosomal protein 72  Belongs to major microtubule - organizing center One dog (#3)  Centriole replication  Spindle organization Chromodomain helicase  ATP binding One dog (#6) DNA binding protein 9  Hydrolase activity Taste receptor type 1  G-protein coupled receptor One dog (#2) member 2  Se nsory perception of sweet taste

Despite the great diversity of these proteins, some recurrent biological functions can be identified. For example seven of these proteins are involved in modulation of innate immune response and/or immunoglobulin class switch/somatic hypermutation, which is particularly interesting within the framework of multicentric lymphoma, a neoplasm arising from immune cells. Moreover, five proteins are involved in control of the cell cycle, DNA repair or apoptosis, which are described as hallmarks of cancers.113 Other common functions include protein ubiquitination (four proteins), and protein transport/trafficking (three proteins).

4.4 Mass spectrometry results: individual dogs

We compared each dog’s protein expression at D7 vs. D0. A p-value was not calculated as for each protein there was only one biological replicate at each time point (one sample at D0 and one sample at D7 for each dog). Based on the FC associated with the proteins significantly differentially expressed across all dogs at D7 vs. D0, we elected to consider a protein downregulated at D7 if the FC D7/D0 was < 0.5 (i.e. at 30

least twice as low at D7 vs. D0), and upregulated if the FC D7/D0 was > 2 (at least twice as high at D7 vs. D0). Furthermore, only proteins detected at both time points were included. After applying these criteria, of the 915 initial proteins, the number retained in the analysis ranged from 199 proteins (dog #2) to 648 proteins (dog #6). As shown in Table 7, the number of proteins up or downregulated was highly variable between dogs. For 4 dogs (#2, #5, #6 and #7), more proteins were upregulated at D7, while for 3 dogs (#1, #3 and #4), more proteins were downregulated at D7.

Table 7. Number of proteins upregulated and downregulated for each dog after seven days of oral sulforaphane supplementation. Proteins with an FC D7/D0 < 0.5 were considered to be downregulated at D7. Proteins with an FC D7/D0 > 2 were considered to be upregulated at D7. Only proteins detected at both day 0 and day 7 were included.

Dog # 1 2 3 4 5 6 7

# of proteins retained 299 199 614 478 478 648 211

# (%) of proteins 423 266 (89) 65 (33) 545 (89) 15 (3) 10 (2) 1 (0.5) downregulated at D7 (88.5) # (%) of proteins 210 33 (11) 134 (67) 69 (11) 55 (11.5) 463 (97) 638 (98) upregulated at D7 (99.5)

We also attempted to identify the protein most downregulated and the protein most upregulated at D7 for each dog. Out of these 14 proteins, we were able to definitively characterized 11 using the methods previously described. The results are presented in Table 8.

Table 8. Characteristics of the proteins most downregulated and most upregulated for each dog after seven days of oral SFN supplementation. FC: fold change.

Dog # Most downregulated at D7 Most upregulated at D7

Ras-related C3 botulinum toxin substrate 1 Hsc70-interacting protein (HIP) 1 (RAC1) FC D7/D0 = 0.011 FC D7/D0 = 21.9 Inter-alpha-trypsin inhibitor heavy chain 3 Wiskott-Aldrich syndrome (WAS) 2 (ITIH3) FC D7/D0 = 0.032 FC D7/D0 = 21.4 Ras-related C3 botulinum toxin substrate 1 Calcium-transporting ATPase (ATP2A3) 3 (RAC1) FC D7/D0 = 0.031 FC D7/D0 = 57.5

Uncharacterized protein Uncharacterized protein 4 FC D7/D0 = 0.016 FC D7/D0 = 13.7

Thioredoxin related transmembrane protein 1 Core histone macro-H2A (H2AFY2) 5 (TMX1) FC D7/D0 = 79.3 FC D7/D0 = 0.042 Signal recognition particle subunit SRP68 tRNA-splicing ligase RtcB homolog (RTCB) 6 (SRP68) FC D7/D0 = 0.02 FC D7/D0 = 60.2 Uncharacterized protein Potassium channel tetramerization domain 7 FC D7/D0 = 0.335 containing 12 (KCTD12) FC D7/D0 = 93.4

4.5 Mass spectrometry results: B-cell vs. T-cell lymphoma

4.5.1 B-cell lymphoma vs. T-cell lymphoma at Day 0

To our knowledge, a baseline comparison of the lymph node proteomes of dogs with B-LSA vs. T- LSA has not been reported previously. We performed a comparison of the protein expression of all B-LSA (n=5) at D0 vs. all T-LSA (n=2) at D0. With a method similar to that previously used for the calculation of proteins’ FC D7/D0, we determined here the FC B0/T0, reflecting the ratio of abundance of a given protein in dogs with B-LSA relative to the abundance of this protein in dogs with T-LSA. Due to the small sample size (only 2 biological replicates for dogs with T-LSA), no p-value was calculated, and proteins with an FC B0/T0 <0.5 or >2 were considered to be downregulated or upregulated, respectively. Moreover, only proteins detected in more than half of the biological replicates for each group (i.e. proteins detected in ≥3 32

B-LSA and in both T-LSA) were selected for further analysis. After applying these criteria, a total of 253 proteins were retained.

Amongst these 253 proteins, 183 (72.3%) were more than twice as abundant in B-LSA vs. T-LSA at D0 (mean FC B0/T0: 3.1; median: 2.6; range: 2-17). Conversely, 70 proteins (27.7%) were detected at least half as low in B-LSA vs. T-LSA at D0 (mean and median FC B0/T0: 0.3; range: 0.03-0.49). Therefore, about 3/4 of the proteins were expressed at a higher level (up to 17 times higher) in B-LSA vs. T-LSA. The remaining 1/4 of the proteins were expressed at a lower level (as low as 33 time (1/0.03) lesser) in B-LSA vs. T-LSA at D0.

The five proteins expressed at the lowest level in dogs with B-LSA when compared to T-LSA at the time of initial diagnosis were protein phosphatase Mg2+/Mn2+ dependent 1G (PPM1G, FC B0/T0 = 0.026), pleckstrin (PLEK, FC B0/T0 = 0.089), taste receptor type 1 member 2 precursor (IFFO2, FC B0/T0 = 0.09), Wiskott-Aldrich syndrome (WAS, FC B0/T0 = 0.107), and ELAV-like protein (ELAVL1, FC B0/T0 = 0.107). The five proteins expressed at the highest level in dogs with B-LSA when compared to T- LSA at the time of initial diagnosis were CD74 molecule (CD74, FC B0/T0 = 17.05), kynureninase (KYNU, FC B0/T0 = 13.98), TNF alpha induced protein 8 (TNFAIP8, FC B0/T0 = 10.07), cytochrome c oxidase subunit 4I1 (COX4I1, FC B0/T0 = 9.25), and α1-acid glycoprotein (AGP, FC B0/T0 = 7.36).

4.5.2 Response to sulforaphane supplementation of dogs with B- vs. T-lymphoma

We explored the possibility that the lymph node proteome of dogs with B-cell or T-cell LSA would respond to oral supplementation with SFN differently. For this, we used the previously described most up- and downregulated proteins in individual dogs, and compared the pattern post-SFN in T-LSA dogs (# 1 and 2) vs. B-LSA dogs (#3 to 7). The results are summarized in Figure 2.

We note that none of the proteins was upregulated in both dogs with T-LSA and downregulated in all dogs with B-LSA, and conversely no protein was downregulated in both T-LSA and upregulated in all B-LSA. The proteins upregulated in both T-LSA at D7 (RAC1 and ITIH3) were also upregulated in two to three B-LSA dogs at D7. The proteins downregulated in both T-LSA at D7 (ATP2A3, TMX1, RTCB and KCTD12) were also downregulated in one to three B-LSA dogs at D7. Therefore, based on this limited sample, the immunophenotype does not immediately appear to be associated with response of the neoplastic lymph node proteome to SFN supplementation.

Dog #1 Dog #2 Dog #3 Dog #4 Dog #5 Dog #6 Dog #7 T-LSA T-LSA B-LSA B-LSA B-LSA B-LSA B-LSA HIP * WAS * ATP2A3 * TMX1 * RTCB * RAC1 * * ITIH3 * H2AFY2 * SRP68 * KCTD12 *

Figure 2. Comparison of the proteins change pattern post-sulforaphane in dogs with T-cell lymphoma (dogs #1 and 2) vs. dogs with B-cell lymphoma (dogs #3-7). The ten proteins included here are the previously described most up and downregulated proteins in individual dogs. Green: protein downregulated post-SFN (FC D7/D0 < 0.5); Red: protein upregulated post-SFN (FC D7/D0 > 2). Grey: no major change post-SFN (FC D7/D0 ≥ 0.5 and ≤ 2) or protein not detected at D0 and/or D7. For each dog, the stars indicate the protein most up regulated and the protein most downregulated. For dog #4, the most upregulated and downregulated proteins were both uncharacterized.

4.6 Validation of mass spectrometry data with immunoblots Two proteins were selected for validation of the LC-MS/MS by immunoblots: 14-3-3 protein theta (14- 3-3-θ) and Heat shock 70 Interacting Protein (HIP). These proteins were chosen as they were amongst the proteins most significantly differentially expressed between D0 and D7 according to the LC-MS/MS results, and because the antibodies targeting these proteins had either previously been validated in dogs, or were predicted to be cross-reactive with the canine protein according to the manufacturer.95,96,100 As shown on Figure 3, immunoblots confirmed the presence of 14-3-3-θ and HIP proteins in the lymph nodes samples of all dogs, although the expression level of 14-3-3-θ detection was low in some samples (i.e. dog #1 at D0 and D7). Interestingly, although for dog #7 neither 14-3-3-θ nor HIP were detected at D0 by LC-MS/MS, both proteins were detected by immunoblot at this time point. The volume of each protein of interest band was normalized to the volume of the corresponding β-actin band, and this normalized expression was used to calculate the NER D7/D0 as defined in the material and methods section. As show on Figure 3, the changes in expression level induced by SFN supplementation for both 14-3-3-θ and HIP were overall similar when expressed as a NER D7/D0 based on the immunoblots results, or as a FC D7/D0 based on the LC-MS/MS results. The only discordant data occurred for the 14-3- 34

3-θ protein in dog #4 (upregulated at D7 based on immunoblot, downregulated based on LC-MS/MS), and dog #5 (downregulated at D7 based on immunoblot, upregulated based on LC-MS/MS).

Figure 3. Immunoblots of lymph node samples for 7 dogs (d) with naïve multicentric lymphoma pre (D0) and post (D7) sulforaphane supplementation. The proteins 14-3-3-θ (A) and HIP (B) were detected in all samples (top rows). The immunoblots for β-Actin, used to normalize the expression level of each protein of interest, are also shown (second rows). The normalized expression ratio D7/D0 for each dog was calculated as the ratio of the protein of interest normalized expression at D7 vs. D0 (third rows). For comparison, the corresponding fold change at D7 vs. D0 previously determined by LC-MS/MS is also annotated (bottom rows). FC D7/D0: fold change at D7 vs. D0 (mass spectrometry); NA: not applicable; NER D7/D0: normalized expression ration at D7 vs. D0 (immunoblot).

4.7 Gene Set Enrichment analysis results

We generated a ranked list of genes from the LC-MS/MS abundance ratio data in three steps as described in the Materials and Methods section. As some canine proteins were not encoded by a gene with a human ortholog, only 514 genes were included in the reference signature. We obtained lists of gene sets from the GSEA molecular signature database (see Methods), focusing on the Gene Ontology (GO, C5 collection) and the Hallmarks of Cancer (H collection).

4.7.1 Gene Ontology

Out of a total of 5917 GO genes sets, 5052 were filtered out after restricting to our data set. Therefore, a total of 865 gene sets (14.6% of the gene sets from the GO database) were used in the analysis. The mean and median number of genes per set were 29.2 and 21, respectively (range: 10-155).

Overall, 720/865 gene sets (83.2%) were initially reported as upregulated at D7 vs. D0. We elected to include only gene sets with a nominal p-value < 0.01 and a false discovery rate (FDR) q < 0.1. The FDR is the probability that a gene set with a given enrichment score represents a false positive finding. Most commonly a FDR of 25% is used, but we elected to use here a more stringent FDR of 10%. This indicates that the result for a gene set to be significantly up- or downregulated at D7 is likely to be valid 9 out of 10 times. When these criteria were applied, only 40 gene sets (4.6%) remained. Gene sets functional annotation can also be redundant and gene sets can contain many overlapping genes (for example, “regulation of humoral immune response” and “humoral immune response”). Therefore, several additional gene sets were also excluded based on such redundancy.

After applying all these criteria, we identified 11 non-redundant gene sets (1.3%) significantly upregulated at D7 at q < 0.1. These gene sets included two main clusters: genes associated with immune response, and genes associated with protein maturation and transport, as detailed in Figure 4. Several of the gene sets upregulated at D7 contained genes encoding proteins also significantly upregulated at D7 according to the LC-MS/MS results. For example, C-type lectin domain family 3 member B (CLEC3B) was the first hit of the most upregulated gene set (“regulation of protein maturation”), and the corresponding protein was also the protein most upregulated (UniProtKB accession number E2RPB8) at D7 across all dogs according to the LC-MS/MS results (FC D7/D0 = 5.3, p = 2.9 x 10-5). 36

GO_REGULATION_OF_PROTEIN_MATURATION **

GO_REGULATION_OF_CYTOKINE_PRODUCTION **

GO_HUMORAL_IMMUNE_RESPONSE *

GO_PROTEIN_ACTIVATION_CASCADE *

GO_REGULATION_OF_IMMUNE_EFFECTOR_PROCESS *

GO_REGULATION_OF_PROTEIN_LOCALIZATION *

GO_PROTEIN_POLYMERIZATION *

GO_POSITIVE_REGULATION_OF_PROTEIN_IMPORT *

GO_POSITIVE_REGULATION_OF_INTRACELLULAR_PROTE IN_TRANSPORT *

GO_REGULATION_OF_INFLAMMATORY_RESPONSE *

GO_REGULATION_OF_PROTEIN_TARGETING *

1.6 1.65 1.7 1.75 1.8 1.85 Normalized Enrichement Score

Figure 4. Bar graph displaying the normalized enrichment score of 11 non-redundant gene sets significantly upregulated (nominal p-value <0.01) at a FDR <0.1, across all dogs at post-sulforaphane. The gene sets, whose names appear on the left of each bar, are composed of genes annotated by the Gene Ontology (GO) term. Asterisks denote statistical significance as follows: * p < 0.01; ** p < 0.001.

Furthermore, 145/865 (16.8%) gene sets were initially reported as downregulated at D7 vs. D0. After filtering with the previously described criteria, only one (0.1%) non-redundant gene set was identified as significantly downregulated at D7 at q < 0.1. This gene set GO annotation was “tRNA processing” and the normalized enrichment score (NES) was -2.02 (p <0.001). Interestingly, the penultimate hit of this set was the gene tRNA-splicing ligase RtcB homolog (RTCB). We previously discussed the corresponding protein (UniProtKB accession E2RCD6), as it was identified by LC-MS/MS as the protein most significantly downregulated at D7 across all dogs (FC D7/D0 = 0.13, p = 5.2 x 10-11).

The enrichment plot for this downregulated gene set, as well as for the most upregulated gene set (“regulation of protein maturation”) are presented on Figure 5. 37

Figure 5. Enrichment plots of the most upregulated gene set (left, “regulation of protein maturation”) and the most downregulated gene set (right, “tRNA processing”) across all dogs post-sulforaphane. The colored scale in the middle of the plot is a visual depiction of the ranked list of genes: the redder and more to the left, the most upregulated are the genes in the reference signature, the bluer and more to the right, the most downregulated are the genes in the reference signature. The vertical bars (“hits”) immediately above the color scale indicate where the genes of the gene set appear in the reference signature. The top portion of the plot displays the running enrichment score for the gene set as the analysis walks down the ranked list of genes of the reference signature. The enrichment score for a gene set is the score furthest from zero (peak of the plot for an upregulated gene set, bottom of the plot for a downregulated gene set). The enrichment score reflects the degree to which the genes contained in a gene sets are overrepresented on one side or the other of the reference signature.

4.7.2 Hallmarks of Cancer

We repeated the same analysis using the gene sets contained in the “hallmarks of cancer” collection of the GSEA molecular signature database. After applying the same previously described filtering criteria, only two of 50 sets in this collection (4%) were significantly upregulated at D7: “coagulation” (NES = 1.7, p = 0.001) and “IL2 STAT5 signaling” (NES=1.6, p = 0.006). No gene sets were significantly downregulated at D7.

Within the “coagulation” gene set, the five first hits (i.e. five most upregulated genes) were plasminogen (PLG), fibrinogen gamma chain (FGG), alpha-2-macroglobulin (A2M), apolipoprotein A1 (APOA1) and 38

Rac family small GTPase 1 (RAC1). The latter protein was previously discussed and was the most upregulated protein at D7 for dogs #1 and #3.

The first hit (i.e. most upregulated gene) of the “IL2 STAT5 signaling” gene set was switching B-cell Complex Subunit SWAP70 (SWAP70), a gene a protein (UniProtKB accession F1P9B2) involved in transducing signals from tyrosine kinase receptors to RAC, leading amongst other sequelae to lymphocyte activation, B-cell isotype switching, mast cell chemotaxis and phagocytosis. The SWAP 70 protein was also upregulated at D7 across all dogs based on the LC-MS/MS results (FC D7/D0 = 2.8), although this did not reach statistical significance (p = 0.07).

5. Discussion

In recent years, SFN has gained significant interest in human oncology due to multiple studies demonstrating both its cancer preventive and cancer suppressive activities.16,18,20,21,26,29,114,115 In particular, SFN has been the object of research in human lymphoid malignancies with promising results.30-32 The study presented here is the first to investigate the impact of SFN-supplementation in cancer-bearing dogs. The results herein show that oral supplementation with twice daily SFN was relatively well tolerated for at least one week, and induced pronounced changes in expression of several hundreds of proteins. Although the variations in proteomic profiles post-SFN varied between dogs, and several hundreds of proteins were identified, some recurrent biological functions could be detected. Specifically, proteins impacted by SFN- supplementation were involved in regulation of innate and adaptive immunity, transcription machinery, cellular response to oxidative stress and toxins, apoptosis, and protein transport, maturation and ubiquitination. Some proteins identified are associated with oncogenic pathways such as MAPK/MEK/ERK or PI3K/AKT, and/or with the prognosis of various human malignancies. Several gene sets from the GO and Hallmarks of Cancer databases that were significantly upregulated post-SFN are primarily involved in immune response or protein maturation and transport. Furthermore, the data presented indicate that the lymph node proteome of dogs with B-cell LSA and dogs with T-cell LSA differs substantially at the time of initial diagnosis.

5.1 Proteins of interest previously investigated in human malignancies

More than 900 proteins were identified across all lymph node samples in this study, however only a small proportion (24/915= 2.6 %) were significantly upregulated or downregulated post-SFN supplementation. With the exception of alanyl-tRNA synthetase (AARS), the 23 other proteins have previously been studied in relation to the pathogenesis and/or prognosis of various human malignancies, mostly representing solid tumors. As detailed in Tables 3 and 4, these include renal, liver, colorectal, pancreatic, pulmonary, head and neck, breast, and ovarian cancer.73,74,77 To our knowledge, none of these proteins have previously been studied in veterinary oncology.

The thioredoxin related transmembrane protein 1 (TMX1), a protein involved in redox homeostasis and cellular response to oxidative stress, was significantly downregulated post-SFN (FC D7/D0 = 0.3). Interestingly, it has been proposed that increased ROS production is a key mechanism through which SFN induces tumor cells apoptosis. In prostate cancer cells, SFN led to apoptosis by depleting GSH levels and increasing oxidative stress.20 Mi et al. found that the activity of antioxidant proteins such as thioredoxin and glutathione S-transferase P in lung cancer cells lines could be inhibited by covalent binding of SFN to 40

the proteins’ cysteine residues.37 To our knowledge, the TMX1 protein has not been evaluated in LSA or leukemia of any species; however, thioredoxine is involved in resistance to doxorubicin, cisplatin and etoposide in human T-cell leukemia cell lines.81 In addition, the dual targeting of the thioredoxin and glutathione anti-oxidant systems was synergistic to induce death of B-LSA cell lines.116 Therefore, further studies are needed to investigate to what extent downregulation of TMX1 and alterations of cellular tolerance to oxidative stress is involved in SFN-induced tumor cell death. In addition, more investigations are needed regarding potential additive/synergetic effects, or conversely the need for avoidance, of SFN with various ROS-producing chemotherapy such as doxorubicin, one of the most commonly used agent to treat canine and human LSA.

The protein proteasome 26S non-ATPase regulatory subunit 2 (PSMD2) was also significantly downregulated post SFN (FC D7/D0 = 0.5). Previous studies showed that SFN and other isothiocyanates can directly bind to, and inhibit, 26S and 20S proteasome subunits leading to cell cycle arrest and apoptosis in multiple myeloma cell lines.37 Our results, although derived from a different type of hematopoietic malignancy, support these findings of SFN-induced proteasome inhibition. Moreover, drugs targeting the proteasome 26S subunit, such as bortezomib, have been used successfully, alone or in combination with various chemotherapeutics and HDACi, to treat myeloma and various types of human LSA. 117,118 Therefore, the downregulation of PSMD2 after a week of SFN supplementation in our study population indicate potential SFN tumor-suppressive effects.

5.2 Proteins of interest previously explored in veterinary oncology

To our knowledge, none of the proteins that were significantly up- or down-regulated at D7 across all dogs have previously been shown to play a role in canine LSA. However, several related proteins have been investigated in veterinary oncology. For example TMX1 (previously discussed for its role in redox homeostasis), was downregulated 3.5-fold post-SFN in our study. The related protein thioredoxin- containing domain C5 was upregulated in metastatic canine mammary carcinomas when compared to non- metastatic mammary carcinomas.79

Additionally, the protein 14-3-3-θ modulates numerous cellular processes including the cell cycle, apoptosis, transcription, cellular stress response and various signaling pathways via recognition of phosphoserine and/or phosphothreonine.73,74,77 In the study presented herein, 14-3-3-θ was significantly downregulated post-SFN (FC D7/D0 = 0.2). Interestingly, Mi et al. identified >30 proteins to which SFN could covalently bind and inhibit; 14-3-3-θ as well as several other proteins of the 14-3-3 family were among those identified.37 The protein 14-3-3-σ, which belongs to the same family, was not detected in 41

histologically normal kidney, but was aberrantly expressed in 38% of canine RCC, and its expression was associated with a decreased survival time.78 This protein was also detected in 97% of normal and neoplastic mammary tissues evaluated, but more intensely in neoplastic epithelial cells infiltrating the vasculature and lymph nodes95. In canine bladder carcinomas, 14-3-3-σ was also expressed mainly by highly infiltrative neoplastic cells, which suggest that this protein is related to the acquisition of an aggressive phenotype.96 Therefore, the role of TMX1 and 14-3-3-θ in the pathogenesis and outcome of canine LSA, and the relevance of their downregulation by SFN, should be investigated further.

5.3 Recurrent biological functions

Although dozens of proteins were identified in this study, some recurrent similarities among their biological functions can be discerned. Most strikingly, numerous proteins or gene sets influenced by SFN are involved in innate or adaptive immunity. The proteins CLEC3B (involved in cellular response to TGF- β), ITIH3 and ITIH4 (acute phase protein in response to IL6 and platelet degranulation) were significantly upregulated across all dogs post-SFN. In addition, seven of the 21 proteins not detected in any dogs at D0, but in at least one dog at D7 (DExD/H-box helicase 58, serine/threonine kinase receptor associated protein, exosome component 7, MutS homolog 6, chemerin chemokine-like receptor 1, Complement 6 and B-cell receptor associated protein 31) are known to play major roles in various immunological process including regulation of cytokine production (GM-CSF, INF-α and β, IL6, IL8, IL12, TNF-α, TGF-β), macrophage chemotaxis, complement receptor signaling pathway, formation of the membrane attack complex, MHC class I binding, immunoglobulin class switch and somatic hypermutation. These findings are reinforced by the GSEA analysis in which four of 11 upregulated gene sets from the GO database and 1 of 2 gene sets from the Hallmarks of Cancer database are involved in modulation of the immune system. To our knowledge, such a pronounced influence of SFN on proteins and gene sets involved in innate or adaptive immunity has not been reported in previous researches. Immunotherapy with monoclonal antibodies and immune checkpoints inhibitors has revolutionized the treatment of human LSA17,119 and encouraging results have recently been reported when combining different autologous vaccines with dose-intense chemotherapy in canine multicentric LSA.120-122 Therefore, the relevance of the SFN-induced changes in the canine immune system should be investigated further, for example in studies combining SFN with autologous vaccines or other immunotherapy modalities in dogs affected by multicentric LSA.

Other recurrent biological functions for the proteins upregulated post-SFN included cellular response to oxidative stress and cellular detoxification, negative regulation of gene transcription, cell cycle checkpoints, positive regulation of apoptosis, positive regulation of intra-cellular protein transport, and 42

protein ubiquitination. Conversely, many of the proteins downregulated post-SFN are involved in gene transcription, tRNA and mRNA processing and maturation, protein translation and post-translational modification, and negative regulation of apoptosis.

5.4 Oncogenic pathways involved SFN has previously been found to interact with several key oncogenic pathways in colon and prostate cancer, acute leukemia and multiple myeloma cell lines. These pathways include PI3K/AKT, MAPK/MEK/ERK, P53 and mitochondrial or death receptor apoptotic pathways.20,29-31,34 Several proteins identified in this study are known to interplay with these pathways. For example, LMNA and PLEK, which are positively associated with the PI3K/AKT pathway, were both significantly downregulated post-SFN in this study. The protein KCTD12, which inhibits the ERK pathway123, was upregulated >2-fold in three dogs post-SFN. The protein HBA is positively associated with apoptosis, and was significantly upregulated post- SFN, while LMNA is negatively associated with apoptosis, and was significantly downregulated post-SFN. Some of these proteins may later prove to be pivotal to understand how SFN can influence various oncogenic pathways, and play a role in tumor prevention or suppression.

5.5 Marked variability in individual dogs’ proteome response to sulforaphane supplementation

The lymph node proteome response of individual dogs to SFN-supplementation was variable. For example, for dog #2, #5, #6, and #7 more proteins were upregulated at D7, while for dog #1, #3 and #4, more proteins were downregulated at D7. Some proteins, such as HIP, were significantly up or downregulated when all dogs were considered together, but the changes in individuals dogs was highly variable. Additionally, the protein most upregulated and the protein most downregulated at D7 was different for almost all dogs, the only exception being Ras-related C3 botulinum toxin substrate 1 (RAC1) which was the protein upregulated to the highest extent for dog #1 and #3 (one T-cell and one B-cell LSA). When evaluated together, the 11 identified proteins upregulated or downregulated to the highest extent post-SFN in individual dogs, no protein was upregulated in both dogs with T-LSA and downregulated in the 5 dogs with B-LSA, or downregulated in both T-LSA and upregulated in the 5 B-LSA. Two proteins were upregulated in both T-LSA dogs (RAC1 and ITIH3), but were also mostly upregulated in dogs with B- LSA. Four proteins were downregulated in both T-LSA (ATP2A3, TMX1, RTCB and KCTD12), but were also downregulated in several B-LSA dogs at D7. Overall, the results of this study indicate that the SFN- induced alterations in the lymph node proteome of dogs with multicentric LSA was not predictable based on clinical response to SFN (SD vs. PD), lymphoma cell size or immunophenotype. Moreover, a statistically 43

significant change at a group scale is not necessarily predictive of responses for individual patients. It is possible that other factors, such as patient’s genetic background or genetic and epigenetic aberrations unique to each LSA, could play a more important role in the response of each individual to SFN.

5.6 Immunoblots validated the mass spectrometry results

Two proteins, HIP and 14-3-3-θ, were chosen to validate some of our LC-MS/MS results via immunoblots. Based on LC-MS/MS, 14-3-3- θ was downregulated at D7 in dogs #1, #2, #3, #4 but upregulated in dogs #5 and 6. The results were confirmed in four dogs via immunoblots, while for 2 dogs (#4 and #5), the LC-MS/MS and immunoblots results were reversed. We suspect that these differences are due to manipulation errors at one or several step of the Western blots (protein quantification and dilution, electrophoresis, transfer on nitrocellulose membrane, antibody binding) since in our opinion, a human error is more likely to have occurred during these experiments than during mass spectrometry. Interestingly, neither HIP nor 14-3-3- θ were detected by LC-MS/MS for dog #7 at D0, but both were detected by Western blot at this time point. This highlights the complementarity of both techniques in proteomic analysis.

5.7 Perspectives from previous canine lymphoma proteomic studies

To our knowledge, five proteomic studies in canine LSA have been published to date.63-67 In two studies only the mass-to-charge ratio (m/z) of the proteins peaks or the peptide mass profile were reported, but not the identity of the proteins.64,67 In the other studies, the proteins differentially expressed between LSA dogs and healthy dogs included haptoglobin, C-reactive protein, macrophage capping protein, prolidase, triosephosphate isomerase, glutathione S-transferase α2 HS glycoprotein, α2-macroglobulin, apolipoprotein A1 precursor, apolipoprotein E, clusterin precursor, haptoglobin, α-antichymotrypsin, antithrombin III, inter α-trypsin inhibitor, and lipopolysaccharide binding protein. 63,65,66 63 Interestingly, amongst the dozens of proteins of interest identified in the current study, only 3 proteins have been reported in previous canine LSA proteomic studies. Inter-alpha-trypsin inhibitor was detected by Atherthon et al and by us via LC-MS/MS (ITIH3 and ITIH4 were significantly upregulated post-SFN). Proteins α2 macroglobulin and apolipoprotein A1 were also both detected by Atherthon et al., and by us in the GSEA (both were genes from the “coagulation” gene set, significantly upregulated post-SFN). 78 There is limited data on the serum or lymph node proteome of dogs with multicentric LSA, and we hope that this work will lay the ground for further research in this area.

5.8 Differences in the proteome of B-cell vs. T-cell lymphoma

Although not the initial aim of this study, this is the first report evaluating differences in the lymph node proteomes of dogs with B-LSA or T-LSA. Due to the small sample size (5 B-LSA and 2 T-LSA) a statistical analysis was not performed in order to avoid overzealous conclusions or type II error. Nevertheless, it is worth reporting that we detected 183 proteins expressed at a level at least twice as high in B-LSA vs. T-LSA, and 70 proteins expressed at a level at least twice as high in T-LSA vs. B-LSA at the time of initial diagnosis. Whether some of these proteins could be involved in the prognostic difference between most B- and T-cell LSA in dogs remains unknown at this point, and more studies in this field are required. We did not collect lymph node biopsies, thus, the definitive LSA grades remain unknown, which is admittedly a limitation of this study. The inclusion of an unbalanced proportion of high grade/low grade LSA in the B-LSA and the T-LSA groups could also explain some of the differences seen in their proteomic profiles.

5.9 Clinical response to sulforaphane supplementation

Sulforaphane supplementation was well tolerated in dogs with multicentric LSA in this study. Most dogs experienced adverse events, which were low grade in all cases. Many events were noted prior to the start of the study and SFN supplementation, and the most significant adverse events occurred in dogs that experienced LSA progression. Therefore, the authors attribute most adverse events to the constitutional, metabolic or hematologic changes secondary to LSA. None of the adverse events reported were considered likely to be related to SFN, however SFN adverse event causality is possible. The dose of SFN used in this study was based on previous human studies22 as well as our team’s assessment of a single SFN dose in healthy dogs.25 It remains unknown if this dose could be safely increased, or if a higher dose could impact the lymph node proteome to a larger extend.

No dog demonstrated a beneficial clinical response after a week of SFN supplementation. This finding was not unexpected. Although numerous studies have demonstrated a major impact of SFN on cancer cell growth and survival in vitro19,27-30,32, and SFN can slow the growth or decrease the volume of tumor xenografts30,31, a measurable gross benefit to SFN-single-agent therapy in patients with naturally occurring cancer has not been reported.

5.10 Limitations

This prospective study is naturally limited by the small number of dogs recruited. However, it was designed as a pilot study in veterinary medicine whose primary goals were to 1) determine if SFN can have an impact on the proteome of cancer-bearing dogs, as has previously been shown in rodents and human, and 2) determine the extent of these changes and identify promising targets to be investigated in follow-up studies. In this regards, we believe that our goals were met. Nevertheless, we cannot exclude the possibility that the changes noted in the expression level of various proteins occurred for reasons unrelated to SFN supplementation. Factors such as physiologic variation in protein production and catabolism in a patient, sampling of different lymph nodes at D0 and D7, differences in the LSA subtypes or grades, or metabolic changes due to progression of the LSA during the week of SFN supplementation may be considered. Similarly, although immunophenotype does not immediately appear to be associated with response of the neoplastic lymph node proteome, this lack of correlation cannot be definitively established due to low sample size. Recruiting a more homogeneous population of dogs, for example only dogs with diffuse large B-cell lymphoma diagnosed based on lymph node biopsy, would have also been ideal in order to reduce variability in our results. We discussed numerous proteins and biological functions influenced by SFN in this manuscript, however these data will need to be validated with a larger population of dogs and a more targeted investigation of individual proteins or groups of proteins. More dogs with multicentric LSA are currently recruited by our group in order to validate these preliminary results, and investigate the impact of SFN on other cellular components. In particular, studies investigating the impact of SFN on genome-wide methylation status, histone acetylation, and transcriptomic profile in dogs with multicentric LSA are currently underway.

6. Conclusion

This prospective study provides for the first time important information regarding the use of SFN in tumor-bearing dogs. We showed that SFN is safe and well tolerated when administered to dogs with multicentric LSA, and can have a profound effects on their lymph node proteome. A total of 915 proteins were detected across all dogs, 14 of which were detected at a significantly lower level (ranging from 2 to 8 times lower) and 10 proteins at a significantly higher level (ranging from 3 to 5 times higher) after 7 days of SFN. Twenty one proteins were not detected in any dog pre-SFN but detected in some dogs post-SFN, and seven proteins were detected in some dogs pre-SFN but not in any dog post-SFN. Several hundreds of proteins were detected at a level at least twice as high (up to 93 times higher) or twice as low (up to 91 times lower) when the lymph node proteome of each individual dog was compared pre- and post-SFN. The qualitative changes in proteomic profile post-SFN varied between dogs, but did not appear to correlate with LSA cell size or immunophenotype. Recurrent proteins biological functions were identified, such as regulation of innate or adaptive immunity, regulation of genes transcription, post-translational modifications, protein transport and ubiquitination, cellular response to stress, and apoptosis. Moreover, many proteins identified in this study have been associated with oncogenic pathways such as MAPK/MEK/ERK or PI3K/AKT, or with the prognosis of various human solid tumor and hematopoietic malignancies. Eleven gene sets from the GO database and 2 gene sets from the Hallmarks of Cancer database were significantly upregulated post-SFN, and primarily involved in immune response or protein maturation and transport. Finally, the lymph node proteome of dogs with naïve B-LSA differed substantially from that of dogs with naïve T-LSA, with more than 250 proteins detected at level at least twice as high (up to 17 times higher) or twice as low (up to 33 time lower) in dogs with B-LSA when compared to dogs with T-LSA. Continued investigation of SFN is veterinary medicine is warranted, for example by combining this natural compound with conventional chemotherapy protocols in canine LSA, or exploring its use in canine cancers whose human counterparts have been shown to be impacted by SFN, such as mammary, intestinal or prostatic carcinoma.

Bibliography

1. Vail DM, Pinkerton ME, Young KM. Canine lymphoma and lymphoid leukemias. In: Withrow SJ, Vail DM, Page RL, eds. Withrow&MacEwen's Small Animal Clinical Oncology. Saint Louis: Elsevier Saunders Co; 2013:608-638. 2. Kaiser HE. Animal neoplasia: a systemic review. In: Kaiser HE, ed. Neoplasms-comparative pathology in animals, plants and man. Baltimore: Williams & Wilkins; 1981. 3. Ito D, Frantz AM, Modiano JF. Canine lymphoma as a comparative model for human non-Hodgkin lymphoma: recent progress and applications. Veterinary immunology and immunopathology 2014;159:192-201. 4. Garrett LD, Thamm DH, Chun R, et al. Evaluation of a 6-month chemotherapy protocol with no maintenance therapy for dogs with lymphoma. J Vet Intern Med 2002;16:704-709. 5. Moore AS, Cotter SM, Rand WM, et al. Evaluation of a discontinuous treatment protocol (VELCAP- S) for canine lymphoma. J Vet Intern Med 2001;15:348-354. 6. Chun R, Garrett LD, Vail DM. Evaluation of a high-dose chemotherapy protocol with no maintenance therapy for dogs with lymphoma. J Vet Intern Med 2000;14:120-124. 7. Childress MO, Ramos-Vara JA, Ruple A. A randomized controlled trial of the effect of prednisone omission from a multidrug chemotherapy protocol on treatment outcome in dogs with peripheral nodal lymphomas. Journal of the American Veterinary Medical Association 2016;249:1067-1078. 8. Lurie DM, Gordon IK, Theon AP, et al. Sequential low-dose rate half-body irradiation and chemotherapy for the treatment of canine multicentric lymphoma. J Vet Intern Med 2009;23:1064-1070. 9. Rassnick KM, McEntee MC, Erb HN, et al. Comparison of 3 protocols for treatment after induction of remission in dogs with lymphoma. J Vet Intern Med 2007;21:1364-1373. 10. Lurie DM, Kent MS, Fry MM, et al. A toxicity study of low-dose rate half-body irradiation and chemotherapy in dogs with lymphoma. Veterinary and comparative oncology 2008;6:257-267. 11. Williams LE, Johnson JL, Hauck ML, et al. Chemotherapy followed by half-body radiation therapy for canine lymphoma. J Vet Intern Med 2004;18:703-709. 12. Imai Y, Maru Y, Tanaka J. Action mechanisms of histone deacetylase inhibitors in the treatment of hematological malignancies. Cancer science 2016;107:1543-1549. 13. Mensah AA, Kwee I, Gaudio E, et al. Novel HDAC inhibitors exhibit pre-clinical efficacy in lymphoma models and point to the importance of CDKN1A expression levels in mediating their anti- tumor response. Oncotarget 2015;6:5059-5071. 14. McCleary-Wheeler AL, Lomberk GA, Weiss FU, et al. Insights into the epigenetic mechanisms controlling pancreatic carcinogenesis. Cancer letters 2013;328:212-221. 15. Cerchietti L, Leonard JP. Targeting the epigenome and other new strategies in diffuse large B-cell lymphoma: beyond R-CHOP. Hematology American Society of Hematology Education Program 2013;2013:591-595. 16. Hsu A, Wong CP, Yu Z, et al. Promoter de-methylation of cyclin D2 by sulforaphane in prostate cancer cells. Clinical epigenetics 2011;3:3. 17. Xue K, Gu JJ, Zhang Q, et al. Vorinostat, a histone deacetylase (HDAC) inhibitor, promotes cell cycle arrest and re-sensitizes rituximab- and chemo-resistant lymphoma cells to chemotherapy agents. Journal of cancer research and clinical oncology 2016;142:379-387. 48

18. Clarke JD, Hsu A, Yu Z, et al. Differential effects of sulforaphane on histone deacetylases, cell cycle arrest and apoptosis in normal prostate cells versus hyperplastic and cancerous prostate cells. Molecular nutrition & food research 2011;55:999-1009. 19. Myzak MC, Tong P, Dashwood WM, et al. Sulforaphane retards the growth of human PC-3 xenografts and inhibits HDAC activity in human subjects. Experimental biology and medicine (Maywood, NJ) 2007;232:227-234. 20. Clarke JD, Dashwood RH, Ho E. Multi-targeted prevention of cancer by sulforaphane. Cancer letters 2008;269:291-304. 21. Tortorella SM, Royce SG, Licciardi PV, et al. Dietary Sulforaphane in Cancer Chemoprevention: The Role of Epigenetic Regulation and HDAC Inhibition. Antioxidants & redox signaling 2015;22:1382- 1424. 22. Clarke JD, Hsu A, Riedl K, et al. Bioavailability and inter-conversion of sulforaphane and erucin in human subjects consuming broccoli sprouts or broccoli supplement in a cross-over study design. Pharmacological research 2011;64:456-463. 23. Clarke JD, Riedl K, Bella D, et al. Comparison of isothiocyanate metabolite levels and histone deacetylase activity in human subjects consuming broccoli sprouts or broccoli supplement. Journal of agricultural and food chemistry 2011;59:10955-10963. 24. Clarke JD, Hsu A, Williams DE, et al. Metabolism and tissue distribution of sulforaphane in Nrf2 knockout and wild-type mice. Pharmaceutical research 2011;28:3171-3179. 25. Curran KM, Bracha S, Wong CP, et al. Sulforaphane absorption and histone deacetylase activity following single dosing of broccoli sprout supplement in normal dogs. Veterinary medicine and science 2018;4:357-363. 26. Beaver LM, Kuintzle R, Buchanan A, et al. Long noncoding RNAs and sulforaphane: a target for chemoprevention and suppression of prostate cancer. The Journal of nutritional biochemistry 2017;42:72- 83. 27. Myzak MC, Dashwood WM, Orner GA, et al. Sulforaphane inhibits histone deacetylase in vivo and suppresses tumorigenesis in Apc-minus mice. FASEB journal : official publication of the Federation of American Societies for Experimental Biology 2006;20:506-508. 28. Myzak MC, Hardin K, Wang R, et al. Sulforaphane inhibits histone deacetylase activity in BPH-1, LnCaP and PC-3 prostate epithelial cells. Carcinogenesis 2006;27:811-819. 29. Wiczk A, Hofman D, Konopa G, et al. Sulforaphane, a cruciferous vegetable-derived isothiocyanate, inhibits protein synthesis in human prostate cancer cells. Biochimica et biophysica acta 2012;1823:1295- 1305. 30. Suppipat K, Park CS, Shen Y, et al. Sulforaphane induces cell cycle arrest and apoptosis in acute lymphoblastic leukemia cells. PloS one 2012;7:e51251. 31. Jakubikova J, Cervi D, Ooi M, et al. Anti-tumor activity and signaling events triggered by the isothiocyanates, sulforaphane and phenethyl isothiocyanate, in multiple myeloma. Haematologica 2011;96:1170-1179. 32. Fimognari C, Nusse M, Cesari R, et al. Growth inhibition, cell-cycle arrest and apoptosis in human T-cell leukemia by the isothiocyanate sulforaphane. Carcinogenesis 2002;23:581-586. 33. Robertson KD, Ait-Si-Ali S, Yokochi T, et al. DNMT1 forms a complex with Rb, E2F1 and HDAC1 and represses transcription from E2F-responsive promoters. Nature genetics 2000;25:338-342. 49

34. Rizzo VL, Levine CB, Wakshlag JJ. The effects of sulforaphane on canine osteosarcoma proliferation and invasion. Veterinary and comparative oncology 2017;15:718-730. 35. Agyeman AS, Chaerkady R, Shaw PG, et al. Transcriptomic and proteomic profiling of KEAP1 disrupted and sulforaphane-treated human breast epithelial cells reveals common expression profiles. Breast cancer research and treatment 2012;132:175-187. 36. Angeloni C, Turroni S, Bianchi L, et al. Novel targets of sulforaphane in primary cardiomyocytes identified by proteomic analysis. PloS one 2013;8:e83283. 37. Mi L, Hood BL, Stewart NA, et al. Identification of potential protein targets of isothiocyanates by proteomics. Chemical research in toxicology 2011;24:1735-1743. 38. Baur AS, Shaw P, Burri N, et al. Frequent methylation silencing of p15(INK4b) (MTS2) and p16(INK4a) (MTS1) in B-cell and T-cell lymphomas. Blood 1999;94:1773-1781. 39. Li Y, Nagai H, Ohno T, et al. Aberrant DNA methylation of p57(KIP2) gene in the promoter region in lymphoid malignancies of B-cell phenotype. Blood 2002;100:2572-2577. 40. Rossi D, Capello D, Gloghini A, et al. Aberrant promoter methylation of multiple genes throughout the clinico-pathologic spectrum of B-cell neoplasia. Haematologica 2004;89:154-164. 41. Thompson RC, Vardinogiannis I, Gilmore TD. The Sensitivity of Diffuse Large B-Cell Lymphoma Cell Lines to Histone Deacetylase Inhibitor-Induced Apoptosis Is Modulated by BCL-2 Family Protein Activity. PloS one 2013;8:e62822. 42. Clozel T, Yang S, Elstrom RL, et al. Mechanism-based epigenetic chemosensitization therapy of diffuse large B-cell lymphoma. Cancer discovery 2013;3:1002-1019. 43. Zullo KM, Guo Y, Cooke L, et al. Aurora A Kinase Inhibition Selectively Synergizes with Histone Deacetylase Inhibitor through Cytokinesis Failure in T-cell Lymphoma. Clinical cancer research : an official journal of the American Association for Cancer Research 2015;21:4097-4109. 44. Yuan B, Zhang J, Wang H, et al. 6-Thioguanine reactivates epigenetically silenced genes in acute lymphoblastic leukemia cells by facilitating proteasome-mediated degradation of DNMT1. Cancer research 2011;71:1904-1911. 45. Ribrag V, Kim WS, Bouabdallah R, et al. Safety and efficacy of abexinostat, a pan-histone deacetylase inhibitor, in non-Hodgkin lymphoma and chronic lymphocytic leukemia: results of a phase II study. Haematologica 2017;102:903-909. 46. Bryan JN, Jabbes M, Berent LM, et al. Hypermethylation of the DLC1 CpG island does not alter gene expression in canine lymphoma. BMC genetics 2009;10:73-73. 47. Sato M, Mochizuki H, Goto-Koshino Y, et al. Prognostic significance of hypermethylation of death- associated protein kinase (DAPK) gene CpG island in dogs with high-grade B-cell lymphoma. Veterinary and comparative oncology 2018;16:409-415. 48. Sato M, Mochizuki H, Goto-Koshino Y, et al. Hypermethylation of the death-associated protein kinase CpG island in canine B-cell lymphoid tumors. Veterinary immunology and immunopathology 2014;161:222-231. 49. Fujiwara-Igarashi A, Goto-Koshino Y, Mochizuki H, et al. Inhibition of p16 tumor suppressor gene expression via promoter hypermethylation in canine lymphoid tumor cells. Research in veterinary science 2014;97:60-63. 50. Yamazaki J, Jelinek J, Hisamoto S, et al. Dynamic changes in DNA methylation patterns in canine lymphoma cell lines demonstrated by genome-wide quantitative DNA methylation analysis. Veterinary journal 2018;231:48-54. 50

51. Ferraresso S, Bresolin S, Aricò A, et al. Epigenetic Silencing of TFPI-2 in Canine Diffuse Large B- Cell Lymphoma. PloS one 2014;9:e92707. 52. Pelham JT, Irwin PJ, Kay PH. Genomic hypomethylation in neoplastic cells from dogs with malignant lymphoproliferative disorders. Research in veterinary science 2003;74:101-104. 53. Tomiyasu H, Goto-Koshino Y, Fujino Y, et al. Epigenetic regulation of the ABCB1 gene in drug- sensitive and drug-resistant lymphoid tumour cell lines obtained from canine patients2013. 54. Kisseberth WC, Murahari S, London CA, et al. Evaluation of the effects of histone deacetylase inhibitors on cells from canine cancer cell lines. American journal of veterinary research 2008;69:938- 945. 55. Nagamine MK, Sanches DS, Pinello KC, et al. In vitro inhibitory effect of trichostatin A on canine grade 3 mast cell tumor. Veterinary research communications 2011;35:391-399. 56. Lin TY, Fenger J, Murahari S, et al. AR-42, a novel HDAC inhibitor, exhibits biologic activity against malignant mast cell lines via down-regulation of constitutively activated Kit. Blood 2010;115:4217-4225. 57. Flesner BK, Kumar SR, Bryan JN. 6-Thioguanine and zebularine down-regulate DNMT1 and globally demethylate canine malignant lymphoid cells. BMC veterinary research 2014;10:290. 58. Fulkerson CM, Dhawan D, Jones DR, et al. Pharmacokinetics and toxicity of the novel oral demethylating agent zebularine in laboratory and tumor bearing dogs. Veterinary and comparative oncology 2017;15:226-236. 59. Hahn NM, Bonney PL, Dhawan D, et al. Subcutaneous 5-azacitidine treatment of naturally occurring canine urothelial carcinoma: a novel epigenetic approach to human urothelial carcinoma drug development. The Journal of urology 2012;187:302-309. 60. Cohen LA, Powers B, Amin S, et al. Treatment of canine haemangiosarcoma with suberoylanilide hydroxamic acid, a histone deacetylase inhibitor. Veterinary and comparative oncology 2004;2:243-248. 61. Wittenburg LA, Gustafson DL, Thamm DH. Phase I pharmacokinetic and pharmacodynamic evaluation of combined valproic acid/doxorubicin treatment in dogs with spontaneous cancer. Clinical cancer research : an official journal of the American Association for Cancer Research 2010;16:4832- 4842. 62. Rizzo V, Wakshlag JJ, Chandler J. The effects of sulforaphane on neoplastic cell proliferation and death In: ACVIM conference, Seattle, WA 2013. 63. Atherton MJ, Braceland M, Fontaine S, et al. Changes in the serum proteome of canine lymphoma identified by electrophoresis and mass spectrometry. Veterinary journal 2013;196:320-324. 64. Gaines PJ, Powell TD, Walmsley SJ, et al. Identification of serum biomarkers for canine B-cell lymphoma by use of surface-enhanced laser desorption-ionization time-of-flight mass spectrometry. American journal of veterinary research 2007;68:405-410. 65. McCaw DL, Chan AS, Stegner AL, et al. Proteomics of canine lymphoma identifies potential cancer- specific protein markers. Clinical cancer research : an official journal of the American Association for Cancer Research 2007;13:2496-2503. 66. Ratcliffe L, Mian S, Slater K, et al. Proteomic identification and profiling of canine lymphoma patients. Veterinary and comparative oncology 2009;7:92-105. 67. Wilson CR, Regnier FE, Knapp DW, et al. Glycoproteomic profiling of serum peptides in canine lymphoma and transitional cell carcinoma. Veterinary and comparative oncology 2008;6:171-181. 51

68. El-Aneed A, Cohen A, Banoub J. Mass Spectrometry, Review of the Basics: Electrospray, MALDI, and Commonly Used Mass Analyzers. Applied Spectroscopy Reviews 2009;44:210-230. 69. Veterinary cooperative oncology group - common terminology criteria for adverse events (VCOG- CTCAE) following chemotherapy or biological antineoplastic therapy in dogs and cats v1.1. Veterinary and comparative oncology 2016;14:417-446. 70. Atwell LL, Zhang Z, Mori M, et al. Sulforaphane Bioavailability and Chemopreventive Activity in Women Scheduled for Breast Biopsy. Cancer prevention research 2015;8:1184-1191. 71. Vail DM, Michels GM, Khanna C, et al. Response evaluation criteria for peripheral nodal lymphoma in dogs (v1.0)--a Veterinary Cooperative Oncology Group (VCOG) consensus document. Veterinary and comparative oncology 2010;8:28-37. 72. Bateman A, Wu C, Bridge A. UniProt Consortium available from www.uniprot.org. In: 2018. 73. Uhlen M, Lundberg E, Lindskog Bergström C, et al. The Human Protein Atlas available from www.proteinatlas.org. In: 2018. 74. Ponten F, Jirstrom K, Uhlen M. The Human Protein Atlas--a tool for pathology. The Journal of pathology 2008;216:387-393. 75. Thul PJ, Akesson L, Wiking M, et al. A subcellular map of the human proteome. Science (New York, NY) 2017;356. 76. Uhlen M, Fagerberg L, Hallstrom BM, et al. Proteomics. Tissue-based map of the human proteome. Science (New York, NY) 2015;347:1260419. 77. Uhlen M, Oksvold P, Fagerberg L, et al. Towards a knowledge-based Human Protein Atlas. Nature biotechnology 2010;28:1248-1250. 78. Suarez-Bonnet A, Lara-Garcia A, Stoll AL, et al. 14-3-3sigma Protein Expression in Canine Renal Cell Carcinomas. Veterinary pathology 2018;55:233-240. 79. Klopfleisch R, Klose P, Weise C, et al. Proteome of metastatic canine mammary carcinomas: similarities to and differences from human breast cancer. Journal of proteome research 2010;9:6380-6391. 80. Wakasugi N, Tagaya Y, Wakasugi H, et al. Adult T-cell leukemia-derived factor/thioredoxin, produced by both human T-lymphotropic virus type I- and Epstein-Barr virus-transformed lymphocytes, acts as an autocrine growth factor and synergizes with interleukin 1 and interleukin 2. Proceedings of the National Academy of Sciences of the United States of America 1990;87:8282-8286. 81. Wang J, Kobayashi M, Sakurada K, et al. Possible Roles of an Adult T-Cell Leukemia (ATL)- Derived Factor/Thioredoxin in the Drug Resistance of ATL to Adriamycin. Blood 1997;89:2480-2487. 82. Kienzle N, Young D, Silins SL, et al. Induction of pleckstrin by the Epstein-Barr virus nuclear antigen 3 family. Virology 1996;224:167-174. 83. Song M, Wang Y, Zhang Z, et al. PSMC2 is up-regulated in osteosarcoma and regulates osteosarcoma cell proliferation, apoptosis and migration. Oncotarget 2017;8:933-953. 84. Suh KS, Tanaka T, Sarojini S, et al. The role of the ubiquitin proteasome system in lymphoma. Critical reviews in oncology/hematology 2013;87:306-322. 85. Langlands FE, Dodwell D, Hanby AM, et al. PSMD9 expression predicts radiotherapy response in breast cancer. Molecular cancer 2014;13:73. 86. Belt EJ, Fijneman RJ, van den Berg EG, et al. Loss of lamin A/C expression in stage II and III colon cancer is associated with disease recurrence. European journal of cancer (Oxford, England : 1990) 2011;47:1837-1845. 52

87. Gong G, Chen P, Li L, et al. Loss of lamin A but not lamin C expression in epithelial ovarian cancer cells is associated with metastasis and poor prognosis. Pathology, research and practice 2015;211:175- 182. 88. Kong L, Schafer G, Bu H, et al. Lamin A/C protein is overexpressed in tissue-invading prostate cancer and promotes prostate cancer cell growth, migration and invasion through the PI3K/AKT/PTEN pathway. Carcinogenesis 2012;33:751-759. 89. Capo-chichi CD, Aguida B, Chabi NW, et al. Lamin A/C deficiency is an independent risk factor for cervical cancer. Cellular oncology (Dordrecht) 2016;39:59-68. 90. Shahzidi S, Brech A, Sioud M, et al. Lamin A/C cleavage by caspase-6 activation is crucial for apoptotic induction by photodynamic therapy with hexaminolevulinate in human B-cell lymphoma cells. Cancer letters 2013;339:25-32. 91. Wazir U, Ahmed MH, Bridger JM, et al. The clinicopathological significance of lamin A/C, lamin B1 and lamin B receptor mRNA expression in human breast cancer. Cellular & molecular biology letters 2013;18:595-611. 92. Hodgkinson VC, D EL, Agarwal V, et al. Proteomic identification of predictive biomarkers of resistance to neoadjuvant chemotherapy in luminal breast cancer: a possible role for 14-3-3 theta/tau and tBID? Journal of proteomics 2012;75:1276-1283. 93. Burmester BR, Gross MA, Walter WG, et al. Pathogenicity of a viral strain (RPL12) causing avian visceral lymphomatosis and related neoplasms. II. Hostvirus interrelations affecting response. Journal of the National Cancer Institute 1959;22:103-127. 94. Love R, Sharpless GR. Studies on a transplantable chicken tumor, RPL-12 lymphoma. II. Mechanism of regression following infection with an oncolytic virus. Cancer research 1954;14:640-647. 95. Suarez-Bonnet A, Herraez P, Martin de las Mulas J, et al. Expression of 14-3-3 sigma protein in normal and neoplastic canine mammary gland. Veterinary journal 2011;190:345-351. 96. Suarez-Bonnet A, Herraez P, Aguirre M, et al. Expression of cell cycle regulators, 14-3-3sigma and p53 proteins, and vimentin in canine transitional cell carcinoma of the urinary bladder. Urologic oncology 2015;33:332.e331-337. 97. Flahaut M, Jauquier N, Chevalier N, et al. Aldehyde dehydrogenase activity plays a Key role in the aggressive phenotype of neuroblastoma. BMC cancer 2016;16:781. 98. Rose M, Kloten V, Noetzel E, et al. ITIH5 mediates epigenetic reprogramming of breast cancer cells. Molecular cancer 2017;16:44. 99. Kurozumi S, Yamaguchi Y, Hayashi S, et al. Prognostic value of the ubiquitin ligase carboxyl terminus of the Hsc70-interacting protein in postmenopausal breast cancer. Cancer medicine 2016;5:1873-1882. 100. Huanan W, Yang x, Jin Y, et al. Expression and significance of CHIP in canine mammary gland tumors2015. 101. Werbowetski-Ogilvie TE, Agar NY, Waldkircher de Oliveira RM, et al. Isolation of a natural inhibitor of human malignant glial cell invasion: inter alpha-trypsin inhibitor heavy chain 2. Cancer research 2006;66:1464-1472. 102. Woong-Shick A, Sung-Pil P, Su-Mi B, et al. Identification of hemoglobin-alpha and -beta subunits as potential serum biomarkers for the diagnosis and prognosis of ovarian cancer. Cancer science 2005;96:197-201. 53

103. Liu J, Liu Z, Liu Q, et al. CLEC3B is downregulated and inhibits proliferation in clear cell renal cell carcinoma. Oncology reports 2018;40:2023-2035. 104. Keymoosi H, Gheytanchi E, Asgari M, et al. ALDH1 in combination with CD44 as putative cancer stem cell markers are correlated with poor prognosis in urothelial carcinoma of the urinary bladder. Asian Pacific journal of cancer prevention : APJCP 2014;15:2013-2020. 105. Paris S, Sesboue R, Delpech B, et al. Inhibition of tumor growth and metastatic spreading by overexpression of inter-alpha-trypsin inhibitor family chains. International journal of cancer 2002;97:615- 620. 106. Kurashige J, Watanabe M, Iwatsuki M, et al. RPN2 expression predicts response to docetaxel in oesophageal squamous cell carcinoma. British journal of cancer 2012;107:1233-1238. 107. Tominaga N, Hagiwara K, Kosaka N, et al. RPN2-mediated glycosylation of tetraspanin CD63 regulates breast cancer cell malignancy. Molecular cancer 2014;13:134. 108. Fujiwara T, Takahashi RU, Kosaka N, et al. RPN2 Gene Confers Osteosarcoma Cell Malignant Phenotypes and Determines Clinical Prognosis. Molecular therapy Nucleic acids 2014;3:e189. 109. Fujita Y, Yagishita S, Takeshita F, et al. Prognostic and therapeutic impact of RPN2-mediated tumor malignancy in non-small-cell lung cancer. Oncotarget 2015;6:3335-3345. 110. Zhang J, Yan B, Spath SS, et al. Integrated transcriptional profiling and genomic analyses reveal RPN2 and HMGB1 as promising biomarkers in colorectal cancer. Cell & bioscience 2015;5:53. 111. Fujimoto D, Goi T, Hirono Y. Expression of ribophorine II is a promising prognostic factor in human gastric adenocarcinoma. International journal of oncology 2017;50:448-456. 112. Ballinger CA, Connell P, Wu Y, et al. Identification of CHIP, a Novel Tetratricopeptide Repeat- Containing Protein That Interacts with Heat Shock Proteins and Negatively Regulates Chaperone Functions. Molecular and Cellular Biology 1999;19:4535-4545. 113. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 2011;144:646-674. 114. Atwell LL, Zhang Z, Mori M, et al. Sulforaphane Bioavailability and Chemopreventive Activity in Women Scheduled for Breast Biopsy. Cancer prevention research 2015;8:1184-1191. 115. Herman-Antosiewicz A, Johnson DE, Singh SV. Sulforaphane causes autophagy to inhibit release of cytochrome C and apoptosis in human prostate cancer cells. Cancer research 2006;66:5828-5835. 116. Kiebala M, Skalska J, Casulo C, et al. Dual targeting of the thioredoxin and glutathione antioxidant systems in malignant B cells: a novel synergistic therapeutic approach. Experimental hematology 2015;43:89-99. 117. Leonard JP, Kolibaba KS, Reeves JA, et al. Randomized Phase II Study of R-CHOP With or Without Bortezomib in Previously Untreated Patients With Non-Germinal Center B-Cell-Like Diffuse Large B-Cell Lymphoma. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 2017;35:3538-3546. 118. Holkova B, Yazbeck V, Kmieciak M, et al. A phase 1 study of bortezomib and romidepsin in patients with chronic lymphocytic leukemia/small lymphocytic lymphoma, indolent B-cell lymphoma, peripheral T-cell lymphoma, or cutaneous T-cell lymphoma. Leukemia & lymphoma 2017;58:1349-1357. 119. Ansell SM, Caligaris-Cappio F, Maloney DG. Immunotherapy in lymphoma. Hematological oncology 2017;35 Suppl 1:88-91. 120. Marconato L, Stefanello D, Sabattini S, et al. Enhanced therapeutic effect of APAVAC immunotherapy in combination with dose-intense chemotherapy in dogs with advanced indolent B-cell lymphoma. Vaccine 2015;33:5080-5086. 54

121. Marconato L, Frayssinet P, Rouquet N, et al. Randomized, placebo-controlled, double-blinded chemoimmunotherapy clinical trial in a pet dog model of diffuse large B-cell lymphoma. Clinical cancer research : an official journal of the American Association for Cancer Research 2014;20:668-677. 122. Sorenmo KU, Krick E, Coughlin CM, et al. CD40-activated B cell cancer vaccine improves second clinical remission and survival in privately owned dogs with non-Hodgkin's lymphoma. PloS one 2011;6:e24167. 123. Li L, Duan T, Wang X, et al. KCTD12 Regulates Colorectal Cancer Cell Stemness through the ERK Pathway. Scientific reports 2016;6:20460.