Cadaverine, a metabolite of the microbiome, reduces breast cancer aggressiveness through trace amino acid receptors

1Microbes that live on the surface or the cavities of the human body affect a large set of pathophysiological processes ranging from metabolic diseases to psychiatric disorders– or neoplastic dieases,–. The number of directly tumorigenic bacteria is extremely low (~10 species), however, dysbiosis is associated with cancers of the urinary tract, cervix, skin, airways, the colon, lymphomas,, prostate and breast cancer–. Dysbiosis is often reflected as a loss of diversity of the microbiota (e.g.). In colon carcinogenesis, immunogenic microbes probably promote the malignancy. However, the majority of the aforementioned cancers are located distantly from larger depots of microbes, hence, suggesting indirect induction or promotion mechanisms. Indeed, bacterial metabolites emerge as “endocrine” agents that are produced by the microbiome, are absorbed into the circulation, and exert their biological effects distantly.

2Deconjugated estrogens,, secondary bile acids–, lipopolysaccharide or propionate (a short chain fatty acid (SCFA)) were proposed to be involved in regulating transformation or cancer cell proliferation. Nonetheless, the molecular mechanisms, through which bacterial metabolites expert their effects are largely unknown. Deoxycholic acid (DCA) was shown to reprogram the hepatocyte secretome, thereby, promoting hepatocellular carcinoma,. Another secondary bile acid, lithocholic acid was shown to inhibit proliferation of breast cancer cells through inhibiting Warburg metabolism and endothelial-to-mesenchymal transition, as well as by enhancing antitumor immunity. LCA exerted its antitumor effects through the TGR5 receptor. Importantly, the latter study showed that in early stages of breast cancer bacterial LCA biosynthesis was decreased suggesting a loss of an antiproliferative bacterial metabolite.

3Cadaverine (CAD) is produced by the decarboxylation of lysine that is performed by lysine decarboxylase (LDC) enzymes. Human cells code and express LDC, but numerous bacterial species of the human microbiome also expresses LDC either in a constant (LdcC in the LDC operon) or in an inducible (CadA in the Cad operon) fashion,. Bacteria use diamines, like cadaverine or putrescine, generated by the decarboxylation of lysine or arginine, to buffer the pH of their environment. The effects of cadaverine on cancer cells and its role in carcinogenesis is not characterized in detail. Therefore, we wanted to assess whether cadaverine can influence the behavior of breast cancer cells.

4As first step, we tested the effects of cadaverine supplementation (500 nmol/kg) to mice homotopically grafted with 4T1 breast cancer cells. Cadaverine supplementation did not alter the number of primary tumors that grew from the grafted cells (Fig. 1A), but there was a trend towards tumors with lower mass (Fig. 1B). In line with that, the number of metastases decreased (Fig. 1C) and, as with the primary tumors, there was a trend for smaller metastases in the cadaverine-treated mice (Fig. 1D). Importantly, cadaverine treatment decreased the invasivity of the primary tumors (Fig. 1E). Histological examination of the primary tumors revealed that cadaverine treatment decreased the rate of mitosis (Fig. 1F,G), the heterogeneity of nuclear morphology (Fig. 1H).

5We investigated whether cadaverine administration could influence the proliferation of cultured breast cancer cell lines. We used five different established breast cancer cell lines of which four were of human (MD-MBA-231, SKBR3, ZR-75-1 and MCF7), while one was of murine origin (4T1). The cadaverine concentration that we used corresponded to the reference concentration of cadaverine in human serum (100–800 nM),. Cadaverine slowed proliferation of 4T1, MDA-MB-231 and SKBR-3 cells as measured in SRB assay (Fig. 2A) or in colony forming assays (Fig. 2B), although the changes were not statistically significant. Importantly, the same concentrations of cadaverine did not hinder the proliferation of non-transformed primary human skin fibroblasts (Fig. 2A). We assessed whether slower proliferation could be due to the toxicity of cadaverine to cells. The proportion of the PI positive cells did not increase upon cadaverine treatment (Fig. 2D), nor did the apoptotic fraction in 4T1 cells (Fig. 2C).

6We assessed whether cadaverine treatment can revert mesenchymal-like cancer cells to epithelial-like cells. First, we performed a measurement in the ECIS system in which 0.1 µM cadaverine increased resistance, suggesting better adherence of cells (Fig. 3A). To verify these findings, we stained cells with Phalloidin-Texas Red to visualize the arrangement of the actin cytoskeleton. Cadaverine treatment changed the fibroblast-like morphology of the 4T1 cells to a rather cobblestone-like morphology (Fig. 3B) that is characteristic for epithelial cells. Treatment of MDA-MB-231 and SKBR-3 breast cancer cell lines with cadaverine led to similar morphological changes (Fig. S1A,D).

7To gain insight into the molecular mechanism through which MET takes place we performed an RT-qPCR screen on EMT genes. The assay revealed differential expression of 11 genes after cadaverine treatment (Table 1), most were suppressed. MMP2, MMP3 and MMP9 support movement; Tgfb3, FgFbp1, Erbb3 and Er1 support proliferation; while Krt14, Notch1, CDH1, IgFbp4 and Spp1 support cell adhesion (Fig. 3C–E).

8In line with these observations, cadaverine-treated cells were slower in migrating to open areas in scratch assays (Fig. 4A) and also performed worse in Boyden-chamber transmigration tests (Fig. 4B). These data were further supported by the observation that MMP9 expression was suppressed by cadaverine treatment in 4T1 cells (Fig. 4C), as well as in MDA-MB-231 and SKBR-3 breast cancer cell lines (Fig. S1B,E). We assessed metabolic changes evoked by cadaverine administration using the Seahorse flux analyzer. Cadaverine treatment reduced glycolytic flux (Fig. 4D) that is a characteristic of breast cancer stromal cells. Therefore, we assessed the “stem-ness” of 4T1 cells using the aldefluor assay and found a mild reduction in cancer cell stem-ness (Fig. 4E). We found similar reductions in cancer cell stem-ness in MDA-MB-231 cells upon cadaverine treatment (Fig. S1C). We did not detect changes in lipid oxidation (Fig. S2).

9The trace amino acid receptor family serve as receptors for cadaverine. Although, most studies on TAAR focused on olfaction, a study linked TAAR1 to breast cancer. Indeed, higher expression of TAAR1, TAAR2, TAAR4, TAAR5, TAAR8 (in ER- cases) and TAAR9 provided better survival in breast cancer (Fig. 5A, Table 2). First, as TAAR receptors are G protein-dependent receptors we assessed their involvement by treating 4T1 cells with NF449, a Gsα-subunit-selective G-protein antagonist, a treatment that abolished the anti-EMT effect of cadaverine (Fig. 5B). Next, we silenced TAAR1, TAAR8 and TAAR9 in MDA-MB-231 cells (Fig. 5C). The silencing of TAAR1, TAAR8 and TAAR9 prevented the cadaverine-elicited mesenchymal-to-epithelial transition and the silencing of TAAR8 and TAAR9 prevented the cadaverine-induced decrease in MMP9 expression (Fig. 5D,E).

10We assessed the available databases to collect data on TAAR1, TAAR8 and TAAR9. TAAR1 and TAAR9 expression did not show any major association with carcinogenesis or breast cancer subtypes in contrast to TAAR8. TAAR8 expression decreased in pre-cancerous lesions, such as hyperplastic enlarged lobular units (HELUs) as compared to normal terminal ductal lobular units (TDLUs). In line with that, TAAR8 expression decreased in breast cancer. Decreases in TAAR8 expression was different between the histological subtypes of breast cancer. TAAR8 expression was lower in triple negative breast cancers as compared to non-triple negative breast cancers. There was a trend for lower TAAR8 expression in DCIS when compared to healthy tissue, TAAR8 expression was lower in ductal invasive breast cancer as compared to healthy ducts and there was a trend for lower TAAR8 expression in lobular invasive breast cancer as compared to healthy lobes. Taken together, apparently, in early stage of breast cancer TAAR8 expression decreases that is more pronounced in triple negative breast cancers.

11To get an insight whether intestinal cadaverine biosynthesis is modified in breast cancer patients, we assessed the abundance of the DNA coding for LdcC and CadA in human fecal DNA from the experimental cohort described in. We designed primers for known CadA and LdcC genes in different bacteria. When comparing healthy individuals and breast cancer patients we observed slightly decreased abundance of Escherichia coli CadA and also E. coli, Enterobacter cloacae and Hafnia alvei LdcC DNA in breast cancer patients (Fig. 6A). Decreased CadA and LdcC abundance was more pronounced in clinical stage 0 patients as compared to the pool of all patients (Fig. 6A). Subsequently, we assessed the protein levels of E. coli LdcC protein in feces by Western blotting. In the feces of stage 1 patients LdcC protein levels were markedly lower than the levels in the feces of healthy subjects (Fig. 6B), in line with the lower fecal DNA abundances.

12We assessed the GEO database to study LDC expression in human breast cancer. There was no difference in LDC mRNA expression between control and breast cancer cases– or in LDC expression of the normal breast epithelium and cancer epithelium in patients. Rather as an exception, LDC expression was lower in basal-like breast cancer as compared to control (normal) breast epithelium of non-diseased individuals.

13Finally, we assessed how expression of LDC in humans affects the outcome of breast cancer using the kmplot.com database (the acquired data is represented in Table 3). Although, differences in LDC expression did affect overall survival of the patients, in grade 1 patients higher expression of LDC was associated with significantly longer survival than lower expression of LDC (Fig. 6C). Interestingly, while LDC expression did not affect survival in ER- PR- patients, higher LDC expression correlated with better survival in ER+ PR+ patients (Fig. 6D).

14In this study we assessed how cadaverine affects the behavior of breast cancer cells. Cadaverine is an ill-characterized biogenic amine as compared to – for example – putrescine that is characterized as a pro-carcinogenic agent,. Increased levels of diamines are in positive correlation with carcinogenesis; serum level of diamines are higher in cancer patients than in healthy individuals,,. Moreover, surgical removal of tumors normalize serum diamine levels,. Based on these observations, polyamines were suggested as cancer markers, although their efficiency was not proven. While in the available literature levels of putrescine were equivocally higher in various cancer patients than in healthy individuals,,, cadaverine levels were more variable among the studies, some reporting increases in cadaverine levels,,, others with trends for decreases. Breast cancer patients have not been evaluated.

15Our data suggests that cadaverine has a tumor suppressor role in breast cancer in concentrations corresponding to the human reference range,. Cadaverine exerted its effects through inhibiting EMT, cellular movement, chemotaxis and metastasis (Fig. 7). These concentrations are much lower than the ones used in previous in vitro studies (10 µM in or 100 µM in). Moreover, similarly to another cytostatic bacterial metabolite, lithocholic acid, cadaverine, in the concentrations corresponding to its reference concentration, did not exert its cytostatic effects on primary fibroblasts suggesting that its effects are specific for tumor cells.

16Cadaverine was shown to activate TAARs, moreover, a study linked TAAR1 to breast cancer. Our results pointed out the possible involvement of TAAR1, TAAR2, TAAR3, TAAR5, TAAR8 and TAAR9 in bringing about the anticancer effects of cadaverine. High expression of these receptors in tumors associated with better survival. The lack of commercially available reagents limited us in validating these findings, we were able to test TAAR1, TAAR8 and TAAR9. All receptors were involved in regulating the mesenchymal-to-epithelial transition, however, only TAAR8 and TAAR9 modulated MMP9 expression (Fig. 7). Furthermore, the expression of TAAR8 was reduced in precancerous states and was gradually downregulated as the severity of the disease worsened; we have not found such correlations with TAAR1 or TAAR9. Available metadata suggest that TAAR8 has strong inverse correlation with breast cancer severity. Due to the limited availability of reagents, we cannot exclude that other TAARs could be equally involved in regulating cadaverine-evoked effects besides the ones we identified.

17Our data suggest that bacterial cadaverine biosynthesis decreases in the gut in early stage breast cancer, resulting in lower production of an anti-cancer bacterial metabolite. In that sense, lower cadaverine production is in line with the lower production of another cytostatic microbial metabolite, lithocholic acid. On a broader scale, these observations suggest that the simplification of the gut microbiome in early stage breast cancer, leads to a decreased production of antiproliferative bacterial metabolites like cadaverine or lithocholic acid. This observation is further supported by the observation that there is a positive correlation between breast cancer incidences or recurrence and cumulative antibiotic use–. Antibiotic exposure increased breast cancer risk in a dose-dependent fashion, nevertheless, these are correlative studies, and the chance of uncontrolled confounding is high and certain studies have found no association,,–. Apparently, the cumulative antibiotic dose is in correlation with increases in breast cancer frequency regardless of antibiotic class, although, the strongest effects were associated with the intermitting use of tetracyclins and macrolids.

18The most drastic changes to the microbiome was associated with in situ carcinoma or stage I carcinoma, in later stages cadaverine production re-increased and reached levels similar to the level of the controls. These findings are similar to our previous observations, in breast cancer the abundance of the cadaverine-producing bacteria was the lowest in in situ carcinoma and stage I carcinoma patients. Nevertheless, the in silico data that we gathered for TAAR8 suggest that in parallel to the simplification or dysbiosys of the microbiome in tumors the expression of the cadaverine-sensing apparatus is probably downregulated.

19There are other sources of cadaverine. Milk ducts in the breast are colonized by bacteria and the duct microbiome changes upon breast cancer– that may contribute to changes in the breast pool of cadaverine. In addition to these, human cells also synthesize cadaverine. High expression of human LDC prolongs survival in early stage breast cancer patients, further supporting the potential anti-cancer properties of cadaverine. Nevertheless, the share of these sources (healthy breast tissues vs. tumor vs. host vs. gut) in cadaverine production is not explored.

20This study puts cadaverine on the list of bacterial metabolites that govern carcinogenesis. Lithocholic acid and cadaverine can suppress cancer hallmarks of breast cancer cells. In hepatocellular carcinoma, lipopolysaccharide and deoxycholic acid (DCA) have been identified as promoters, while propionate is an inhibitor. These observations support the hypothesis that bacterial metabolites influence carcinogenesis in organs remote from the gut. Nevertheless, this field requires considerably more research to understand the interactions between the microbiome and distantly located tumors.

21All methods were performed according to the relevant guidelines.

22Chemicals, among them, cadaverine were from Sigma-Aldrich (St. Louis, MI, USA) unless otherwise stated.

234T1 murine breast cancer cells, ZR-75-1 human breast cancer cells were maintained in RPMI-1640 (Sigma-Aldrich, R5886) medium containing 10% FBS, 1% penicillin/streptomycin, 2 mM L-glutamine and 1% pyruvate at 37 °C with 5% CO2.

24MDA-MB-231 and SK-BR-3 human breast cancer cells were maintained in DMEM (Sigma-Aldrich, 1000 mg/l glucose, D5546) containing 10% FBS, 1% penicillin/streptomycin, 2 mM L-glutamine and 10 mM HEPES at 37 °C with 5% CO2.

25MCF7 human breast cancer cells were maintained in MEM (Sigma-Aldrich, M8042) medium containing 10% FBS, 1% penicillin/streptomycin, 2 mM L-glutamine and 10 mM HEPES at 37 °C with 5% CO2.

26Human primary fibroblast cells were maintained in DMEM (Sigma-Aldrich, 1000 mg/l glucose, D5546) containing 20% FBS, 1% penicillin/streptomycin, 2 mM L-glutamine and 10 mM HEPES at 37 °C with 5% CO2.

27Transfections were performed as in.

28Sulphorhodamine B assay and colony forming assays were performed as in,.

29For the detection of cell death we used simple propidium iodide (PI, Biotium, Fremont, CA, 40016) uptake assays (as in), while to differentiate between apoptosis and necrosis we used an Annexin V + PI double staining assay kit (Invitrogen, Oregon, USA, V13242).

30ECIS assays were performed similarly to. ECIS measurements are used to assess cell-to-cell and cell-to-surface connections.

31Immunocytochemistry was performed similarly to,.

32The human fecal DNA library from breast cancer patients was published in. For the determination of the abundance of Lysine decarboxylase coding DNA in human fecal DNA samples 10 ng of DNA was used for qPCR reactions. Primers are listed in Supplementary Table 1. The amplicons were subsequently sequenced using the same primers as listed in Supplementary Table 1.

33Reverse transcription-coupled PCR (RT-qPCR) was performed similarly to.

34RIPA buffer (50 mM Tris, 150 mM NaCl, 0.1% SDS, 1% Triton X-100, 0.5% sodium deoxycholate, 1 mM EDTA, 1 mM Na3VO4, 1 mM NaF, 1 mM PMSF, protease inhibitor cocktail) was used to lyse cells of fecal samples. Samples were sonicated (Qsonica Q125 Sonicator, Newtown, Connecticut) 3 times for 30 seconds with 50% amplitude. After centrifugation, 25 µl 5 X SDS sample buffer (50% glycerol, 10% SDS, 310 mM Tris HCl, pH 6.8, 100 mM DTT, and 0.01% bromophenol blue) and 8 µl β-mercaptoethanol were added to every 100 µl supernatant. Samples were heated 10 minutes at 96 °C and kept on ice until loading.

35SDS PAGE and Western blotting was performed as in.

36Scratch assays were performed as in. Scratch assays were employed to assess cell movement.

37Matrigel invasion assay was carried out on 4T1 cells using Corning BioCoat Matrigel Invasion Chamber (354480). 4T1 cells were seeded in the chambers (50000 cells/well) in serum free medium, and were grown overnight. Cells were then treated with different concentration of cadaverine (0.1 µM, 0.3 µM, 0.8 µM). The lower chamber contained full 4T1 medium with 100 ng/ml SDF1-alpha (Sigma, SRP4388) as chemoattractant. After 48 hours of cadaverine treatment cells were prepared according to the manufacturer’s instructions and stained with Hematoxylin-Eosin (VWR, 340374T and 341972Q) dye. Cells were then pictured with Opera Phoenix High Content Screening System and pictures were analyzed using Harmony 4.6 Software. Invasion index was calculated from the percentage of invading cells through Matrigel membrane and control membrane.

38Changes to oxygen consumption rate (OCR, a readout of mitochondrial oxidation) and in pH (ECAR, a readout of lactate production and glycolytic flux) after cadaverine treatment were performed similarly to.

39TBARS assay was performed to assess lipid peroxidation as in.

40Aldefluor assay for assessing stem-ness was performed similarly to.

41Animal studies were conducted similarly to.

42Animal experiments were approved by the Institutional Animal Care and Use Committee at the University of Debrecen and the National Board for Animal Experimentation (1/2015/DEMÁB) and were carried out according to the NIH guidelines (Guide for the care and use of laboratory animals) and applicable national laws. Animal studies are reported in compliance with the ARRIVE guidelines,.

43We used BALB/c female mice (4 months of age, 20–25 g). Animals were bred in the “specific pathogen free” zone of the Animal Facility at the University of Debrecen, and kept in the “minimal disease” zone during the experiments. 4 mice were housed in one cage (standard block shape 365 × 207 × 140 mm, surface 530 cm2; 1284 L Eurostandard Type II. L from Techniplast). Dark/light cycle was 12 h, and temperature was 22 ± 1 °C. Mice had ad libitum access to food and water (sterilized tap water). A total of 32 female mice were used in the study, 16 randomly selected control and 16 cadaverine fed mice. The study was performed in two runs at two different occasions, each run comprising of 8 vehicle-treated and 8 cadaverine-treated mice.

44Tumor was formed in mice by the grafting of 4T1 cells. 4T1 cells were suspended (2 × 106/mL) in ice cold PBS-Matrigel (1:1, Sigma-Aldrich). 16 female BALB/c mice received 50 µL injections to the inguinal fat pads below the lower abdominal nipples on both sides (105 cells/injection site).

45Animals received daily oral cadaverine treatment. Cadaverine stock was prepared in sterilized tap water at 100x concentration (15 mM) and the stock was stored at −20 °C. Cadaverine stock was diluted each day to a working concentration of 150 µM in sterile tap water before the treatment. Animals received a daily oral dose of 100 µl/30 g bodyweight from cadaverine solution (8 mice) or vehicle (sterilized tap water, 8 mice). Researchers administering cadaverine and vehicle solutions were blinded. Treatment was carried out every day during the morning hours between 9 am and 11 am. Mice were sacrificed on day 14 post grafting.

46During autopsy primary tumors were scored based on their infiltration rate into surrounding tissues based on macroscopic appearance of the tumor. “Low infiltration” class means that primer tumor remained in the mammary fat pads without any attachment to muscle tissues. In case the tumor mass attached to the muscle tissue but did not penetrate to the abdominal wall, it classified as a “medium infiltration” tumor. If the tumor grew into the muscle tissue and totally penetrated the abdominal wall, it was scored as a “high infiltration” tumor.

47Both primary and metastatic tumor masses were removed from mice and were measured on analytical balance in preweighed Eppendorf tubes.

48We assessed the abundance of the bacterial DNA coding for lysine decarboxylase in human fecal DNA samples. The human fecal samples were collected from healthy women and breast cancer patients by collaborators at the National Cancer Institute (NCI), Kaiser Permanente Colorado (KPCO), and the Institute for Genome Sciences at the University of Maryland School of Medicine, and RTI International. The study protocol and all study materials were approved by the Institutional Review Boards at KPCO, NCI, and RTI International (IRB number 11CN235). The primary study results were published in. All methods were performed in accordance with the relevant guidelines and regulations.

49We obtained informed consent from study participants.

50Another cohort was used to assess LdcC protein in the feces of healthy volunteers and breast cancer patients. The collection and biobanking of feces was authorized by the Hungarian national authority (ETT). Patients and healthy volunteers meeting the following criteria were excluded from the study according to the corresponding national guideline for fecal transplantation: (1) has previous history of breast cancer or had been operated due to neoplasia, (2) has a disease of unknown origin, (3) has chronic contagious disease, (4) had contagious diarrhea 6 months prior to enrollment, (5) taken antibiotics in the 6 months prior to enrollment, (6) had chemotherapy, biological therapy or immunosuppressive therapy 6 months prior to enrollment, (7) used intravenous drugs 12 months prior to enrollment, (8) had piercing, tattooing, acupuncture or other endangering behavior or action 12 months prior to enrollment, (9) exposition to an allergen to which the enrolled individual had been sensitized to, (10) underwent colonoscopy 12 months prior to enrollment. First morning feces was sampled; samples were frozen and deposited in the biobank two hours after defecation. Samples were stored at −70 °C until subsequent use.

51The kmplot.com database was used to study the link between gene expression levels and breast cancer survival in humans. The association of known mutations with breast cancer was retrieved from www.intogen.org/. The sequence of the CadA and LdcC ORFs were retrieved from the KEGG (www.genome.jp/kegg/)database. We assessed the NCBI GEO Profiles with the term “LDC and breast cancer” and the GENT database with the keywords “TAAR1”, “TAAR8” and “TAAR9”.

52We used two tailed Student’s t-test for the comparison of two groups unless stated otherwise. Fold data for human fecal DNA assessment were log2 transformed to achieve normal distribution. For multiple comparisons one-way analysis of variance test (ANOVA) was used followed by Tukey’s honestly significance (HSD) post-hoc test. Data is presented as average ± SEM unless stated otherwise. Texas Red-X Phalloidin-labelled fluorescent pictures were analyzed using Image J or Cell Profiler 2.0 followed by Advanced Cell Classifier 3.0. FACS results were analyzed using Flowing Software 2.0. Statistical analysis was done using Origin 8.6 software unless stated otherwise.