Disclosing a metabolic signature of cisplatin resistance in MDA-MB-231 triple-negative breast cancer cells by NMR metabolomics

The human TNBC cell line MDA-MB-231 (ATCC HTB-26; absence of estrogen and progesterone receptors, HER2 overexpression) was purchased from ATCC (Manassas, VA, USA). BC cells were cultured in DMEM-HG cell growth medium supplemented with 10% (v/v) FBS and maintained under a humidified atmosphere of 5% CO₂ at 37 °C. The cDDP-resistant cell line was established as previously described [30]. Briefly, MDA-MB-231 cells were continuously treated with increasing concentrations of cDDP (up to a maximum of 2 µM) during 6 months. When a consistent cell growth rate was achieved, in the presence of cDDP, the resulting cell line, designated as MDA-MB-231/R, was stored at − 80 °C. All subsequent experiments were performed within 10 passages, for both cDDP-sensitive and -resistant cell lines, maintaining the MDA-MB-231/R cell line in growth medium in the absence of cDDP. Under these conditions, the population doubling times were 25.5 ± 0.9 h and 30.6 ± 1.1 h for MDA-MB-231 (sensitive, designated as S) and MDA-MB-231/R (resistant, designated as R) cells, respectively. The two cell lines will be designated, when possible, as R and S for MDA-MB-231/R and MDA-MB-231, respectively. The cell cultures were routinely screened for mycoplasma contamination, all assays having yielded negative results.

Cells were seeded in 96-well microplates at the cell density 1.5 × 10⁴ cells/cm² (final volume 200  µL/well) and left to attach for 24 h. Label-free kinetic live monitoring of cell growth was performed using LionheartFX automated microscope (BioTek, Winooski, VT, USA) with direct image acquisition of cells in microplates at 48 h. The 4X images acquired were processed using the Gen 5 Image Analysis software (BioTek, Winooski, VT, USA) that allows for identification and counting of individual cells per image. IC₅₀ (half maximal inhibitory concentration) values were calculated for both S and R cell lines incubated for 48 h with increasing concentrations of cDDP (5, 10 and 20 μM) [30].

Changes in phosphorylated ERK1/2 (p-ERK) were detected using the AlphaScreen SureFire p-ERK1/2 Kit following the methods described in detail elsewhere [31, 32]. Briefly, 10 μL of total protein lysate were transferred into 384-well ProxiPlates and a mixture of acceptor beads and donor beads were added, following the manufacturer’s instructions. In addition, changes in phosphorylated NF-κB (p-NF-κB) were measured using the AlphaScreen SureFire p-NF-κB Kit, following the same protocol described above.

In relation to cell assays, the data was expressed as the mean ± standard error of the mean (SEM), with each of three independent experiments, in triplicate (n = 3). Statistical analysis of cell assay results was performed using (i) non-linear regression analysis of the corresponding dose–response curves to calculate the IC₅₀ values; (ii) the two-tailed Student’s t-test to compare R vs. S cells; (iii) one-way ANOVA to compare each cell line with the respective controls, followed by the Dunnett’s t-test. The GraphPad Prism 7 Software (San Diego, CA, USA) was used. A p-value lower than 0.05 was considered statistically significant.

MDA-MB-231 parental (cDDP-sensitive) and resistant cells were seeded at a density of 3 × 10⁴ cells/cm² onto 150 mm Petri dishes (ø 135.5 mm), cultured in a humidified atmosphere of 5% CO₂ at 37 °C and allowed to adhere for 24 h. The cells were then incubated and collected at 0, 24 and 48 h, with basis on the population doubling times mentioned above. At each time-point, cells were harvested using a 0.25% (v/v) trypsin–EDTA solution, washed twice with PBS and centrifuged (300 g, 5 min, 20 °C). The cell pellets were stored at − 80 °C until analysis. Three independent experiments with triplicates were performed for each cell line and time-point.

The cellular polar extracts were obtained using a biphasic extraction method of methanol/chloroform/water previously reported [33]. Briefly, cell pellets were suspended in 650 µL of 80% (v/v) methanol-miliQ water solution, transferred to microcentrifuge tubes with 150 mg of glass beads (ø = 0.5 mm) previously weighted, and vortexed for 5 min to aid cells disruption. Subsequently, 260 µL of 100% chloroform and 260 µL of 100% chloroform plus 220 µL MiliQ water were added to samples, which were vortexed for 5 min between solvents addition. Samples were stored at −20 °C for 10 min and centrifuged (2,000 g, 15 min, room temperature). The aqueous phase was collected into a new tube, vacuum-dried and stored at − 80 °C until further analysis. All samples and reagents were kept in ice during the extraction procedure. Before NMR analysis, the dry aqueous extracts were suspended in 650 µL of 100 mM sodium phosphate buffer (pH 7.4, in D₂O containing 0.25% 3-(trimethylsilyl)-propionic-2,2,3,3-d4 acid (TSP) for chemical shift referencing) and transferred into 5 mm NMR tubes.

The NMR spectra were recorded on a Bruker AVANCE III spectrometer, equipped with a 5 mm TXI probe, operating at 500.13 MHz for ¹H observation, and at 298 K. Standard 1D ¹H NMR spectra of aqueous extracts were acquired using a water presaturation pulse sequence (“noesypr1d” from Bruker library, Rheinstetten, Germany), with 7002.801 Hz spectral width, 32 k data points, 2.34 s acquisition time and 2 s relaxation delay and 512 scans. Prior to Fourier transformation, each free-induction decay was zero-filled to 64 k points and multiplied by a 0.3 Hz exponential line-broadening function. Spectra were manually pre-processed including phase correction, baseline adjustment and internal calibration of chemical shifts to TSP. For peak assignment, 2D NMR homonuclear total correlation (TOCSY) and heteronuclear single-quantum correlation (HSQC) spectra were acquired for selected samples, along with comparison with existing literature and spectral databases, such as Bruker BIOREFCODE (AMIX-viewer 3.9.14, Bruker Biospin, Rheinstetten, Germany), human metabolome database (HMDB) [34] and Chenomx NMR Suite (Chenomx Inc, Edmonton, AB, Canada).

The unidimensional proton NMR spectra were converted into matrices (AMIX 3.9.14, Bruker Biospin, Rheinstetten, Germany), excluding methanol (δ 3.36, singlet) and water (δ 4.4–5.4) spectral regions. The spectra were aligned by recursive segment-wise peak alignment (RSPA) to minimize chemical shift variations (Matlab 8.3.0, The MathWorks Inc., Natick, Massachusetts, USA), and normalized to total spectral area to account for different cells numbers. Multivariate analysis was carried out using unsupervised principal component analysis (PCA) and supervised partial least squares–discriminant analysis (PLS-DA), upon unit variance (UV) scaling (SIMCA-P 11.5; Umetrics, Umeå, Sweden). PLS-DA models with corresponding values of predictive power (Q²) higher than 0.50 were considered statistically robust. PLS-DA loadings were back-transformed, multiplying each variable by its standard deviation, and colored according to variable importance to the projection (VIP) (Matlab 8.3.0, The MathWorks Inc., Natick, MA, USA). The respective loading plots revealed the resonances relevant for class separation, which were selected for area integration (AMIX 3.9.14, Bruker BioSpin, Rheinstetten, Germany), normalization, and variation assessment by univariate analysis. Univariate analysis of metabolites combined effect size (ES) [35] and statistical significance (Shapiro–Wilk test to assess data normality, Student’s t-test or Wilcoxon test for normally distributed or non-normally distributed data, respectively) (R statistical software). For multiple testing, p-values of significantly changed metabolite levels (|ES|> ES error and p < 0.05) were corrected by false discovery rate (FDR), based on the Benjamini and Hochberg method [36]. Significant metabolite differences between S and R cell lines were all confirmed by posterior visual inspection of the spectra, and putatively interpreted based on information derived from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database [37].

In order to demonstrate cDDP-resistance in R cells, the impact of increasing concentrations of cDDP (5, 10 and 20 μM) on S or R cells, incubated for 48 h, was evaluated through cell growth measurements (Fig. 1a). The results showed that S cells were highly sensitive to cDDP treatment, while the response of R cells to cDDP was considerably attenuated (Fig. 1a), with similar doubling times for both cell lines (25.5 ± 0.9 h and 30.6 ± 1.1 h for S and R cell lines, respectively). This was consistent with the reported cDDP half maximal inhibitory concentrations IC₅₀ (48 h) of 1.0 μM and 32.4 μM for S and R cells, respectively [30]. The above results are consistent with data previously reported for other MDA-MB-231 cDDP-resistant human cell lines [38‐40].

Morphological analysis revealed a normal spindle-shaped phenotype for both S and R cells, although R cells were characterized by an enriched portion (Fig. 1b) and a larger cell size (566.4 ± 20.7 μm² in S vs. 779.4 ± 27.4 μm² in R cells, p < 0.0001, Fig. 1c) of polyploid giant (cancer) cells (PGCCs) as compared to S cells. Furthermore, the basal phosphorylation profiles of ERK1/2 and NF-κB proteins in S and R cells were measured and compared (Fig. 1d), as the MAPK/ERK pathway plays a crucial role in the survival and development of tumor cells [41, 42], whereas the NF-κB pathway has also been linked to cancer survival [43]. For R cells, the basal phosphorylation of NF-κB showed an increasing tendency of nearly 17% (p = 0.1380), while a significant decrease was noted in the basal levels of phosphorylated ERK1/2 (near 50% decrease, p = 0.0077), compared to S cells (Fig. 1d).

The average ¹H NMR spectrum of polar extracts of untreated S cells (0 h) (Fig. 2a) reflects the presence of over 20 amino acids and derivatives, 3 choline compounds (choline, phosphocholine (PC), glycerophosphocholine (GPC)), 17 nucleotides and derivatives, 8 organic acids and 4 other compounds (dimethylamine (DMA), glycerol, m-inositol and trimethylamine N-oxide (TMAO)) (see table of assignments, Additional file 1). This is consistent with previously reported NMR data for polar extracts of this cell line [27, 44]. At 0 h, some changes are apparent through visual inspection of the average spectrum of R cells (arrows in Fig. 2b), compared to that of S cells, namely increased alanine, glutathione (reduced form, GSH) and glutamine, and decreased acetate, glutamate, adenosine-mono/di/triphosphate (AMP, ADP, ATP), pseudouridine, uridine diphosphate glucose/glucuronic acid (UDP-Glc/GlcA) and nicotinamide adenine dinucleotide (NAD⁺), all of these variations having been confirmed upon statistical analysis, as described below.

PCA and PLS-DA models were used to compare all R and S samples (full and open symbols, respectively, in Fig. 3, top) showing that the metabolic profile is significantly distinct between the two cell lines, independently of culture time and with slightly more dispersion for sensitive cells. PLS-DA LV1 loadings (Fig. 3, bottom) seem to indicate that some of the most consistent differences affect glutamate, glutamine, GSH, GPC, NAD⁺, taurine, aspartate, several nucleotides, UDP-Glc/GlcA and hypoxanthine (HX).

Such differences were confirmed by spectral integration and statistical assessment, which produced a more complete list of reliable differences between all R and S cells, independently of time point (Table 1). Overall, the results evidence that cDDP-resistant cells differ from sensitive cells in the levels of 38 identified compounds: 14 amino acids and derivatives, choline and GPC, 15 nucleotides and derivatives, 5 organic acids, DMA and m-inositol. The most significant differences (p ≤ 10^–10 and |ES|≥ 2.0) in resistant cells include (i) higher levels of GSH, glutamine, phosphocreatine (PCr), taurine and lower levels of glutamate; and (ii) overall lower levels of adenine, adenosine, NAD⁺ and UDP-Glc/GlcA (Table 1, bold and underlined metabolites), although all listed variations are important (e.g. depletion in choline, GPC, and in a wide range of nucleosides/nucleotides), all organic acids detected (including lactate) and m-inositol, as they remain statistically meaningful after FDR correction.

The average differences between R and S cell lines have a perceptible dependence on culture time (Fig. 4), with a general tendency for higher amino acid levels at 0 and 24 h in R cells (except for glutamate and proline), with several amino acids showing depletion at 48 h in the same cells. It is clear that R cells are markedly depleted in choline compounds and in many nucleotides and derivatives, at all time points (Fig. 4). Notably, analysis of distinct time points unveils several new variations in addition to the overall average changes listed in Table 1, namely in (i) tricarboxylic acid cycle (TCA) intermediates citrate, malate and succinate; (ii) amino acids alanine, leucine, N-acetyl-aspartate (NAA) and tyrosine; (iii) other compounds: ATP and glycerol. Hence, it is clear that, not only average metabolite pools are significantly different between the two cell lines, but also that time point comparisons unveil additional varying metabolites.

In order to examine each metabolite trajectory overtime, pairwise PLS-DA models were obtained for each cell line, revealing high predictive power values (Q²) of ca. 0.75 and 0.80–0.84 for S and R cells, respectively (Fig. 5a, b) and thus confirming that culture time significantly changes the metabolic profile of each cell line. In particular, this knowledge is relevant for drug exposure studies, where time course markers may be defined. Indeed, at 48 h, the metabolic profile of each cell line is markedly distinct from that at 0 h, with Q² approaching the maximum value of 1.0 (0.92–0.94) (Fig. 5c). The metabolite changes taking place as a function of time (and for the 0/48 h comparison) are quantified in Additional file 2 and Additional file 3 for S and R cell lines, respectively (most changes surviving FDR correction), and illustrated for each cell line in a heatmap form (Additional file 4). In general, both S and R cell lines exhibit a general increase in amino acids from 0 to 24 h (except for glycine which decreases in S cells), followed by a tendency for amino acid stabilization from 24 to 48 h (less number of variations, particularly in R cells). Considering the respective errors (Additional file 2 and Additional file 3), the most evident net differences (48 h vs 0 h) in R cells, compared to S cells, comprise: (i) lower/no increases in alanine, isoleucine, leucine and (N-acetyl-aspartate) NAA; (ii) more marked increases in creatine (Cr), methionine and taurine. Resistant cells show raised choline levels (no changes in S cells) and no change in GPC (decreased in S cells). The nucleotides profile of R cells is remarkably stable compared to S cells (Additional file 4), only 6 metabolites changing (against 16 in S cells), either in the same direction but lower magnitude then for S cells (ADP, hypoxanthine (HX), UDP and uridine) or in opposite direction i.e. decreased in R cells (NAD⁺ and uridine diphosphate N-acetyl-glucosamine (UDP-GlcNAc)) (Additional file 4). The S cells exhibit contrastingly strong overtime variations for adenine and uracil and derivatives (Additional file 4). As to organic acids, the R cells show similar trajectories to S cells (decreased acetate and formate, and increased fumarate, lactate and malate), although without changes in citrate, panthothenate (PA) and succinate as observed in S cells (Additional file 4). Glycerol hardly varies in R cells (strongly decreased in S cells), whereas m-inositol increases more clearly in R cells) (Additional file 4).

The trajectory graphs (Additional file 5, Additional file 6, Additional file 7) of all varying metabolites simultaneously highlight differences in levels and trajectories. Within these, the strongest distinguishing metabolites (defined as differing at least in two time points with p ≤ 0.0001) are shown in Fig. 6. Resistant cells are clearly differentiated at all time points by richer pools of glutamine, GSH, methionine, PCr (leading to higher PCr/Cr ratios at longer times, Additional file 8a), taurine and dimethylamine (DMA); and depleted pools of glutamate, GPC, adenine, adenosine, AMP, HX, NAD⁺, UDP-Glc/GlcA, UDP-GlcNAc, uridine (barely detectable in resistant cells), acetate and fumarate (Fig. 6). Consequently, R cells exhibit consistently lower glutamate/glutamine, GPC/PC, GPC/Cho and NAD⁺/NADH ratios than S cells (Additional file 8). Although ATP and ADP start off depleted in R cells (0 h), both metabolites evolve to comparable levels at 48 h in both cell lines and ADP/ATP significantly distinguishing cell lines only at 24 h (lower ADP/ATP ratio in R cells). The full set of metabolite changes in R cells, compared to S cells, and their dependence with culture time are represented in Fig. 7.

The cytotoxicity results currently obtained clearly illustrate the high cDDP-resistance of R cells as compared to S cells. This evidence is accompanied by the presence of a higher number of PGCCs in the resistant cells, confirming previous reports of a relationship between the number of PGCCs with tumor chemo-resistance and aggressiveness [45, 46]. PGCCs are cells with multiple nuclei or a single giant nucleus containing multiple complete sets of chromosomes and it is well documented that these cells are present in several solid tumors, usually in lower numbers in sensitive cells before treatment, compared to resistant cells [45]. Moreover, a positive correlation has been formerly described between the number of PGCCs and glioma stage and grade [47].

Furthermore, a basal phosphorylation of NF-κB is observed in S and R cells, although in the R cells a tendency to an increase in p-NF-κB is observed consistent with a slight higher activation of the NF-κB pathway and, hence, cell survival, as shown by reports of the relationship between NF-κB pathway activation, through the IKKα pathway, and cell survival under conditions of cDDP exposure [48]. In addition, a significant decrease in the basal levels of phosphorylated ERK1/2, compared to S cells, seems to characterize the resistance of TNBC cells to cDDP. Again, this is consistent with previous reports associating cancer cells (e.g. cervical carcinoma) (in)sensitivity to cDDP with the downregulation of ERK pathway activation [49, 50]. Considering that the ERK1/2 pathway may crosstalk with other signaling pathways, and that reports are found on both ERK 1/2 and the NF-κB being involved in cDDP-acquired resistance in BC cells [51, 52], we propose that the constitutive activation of NF-κB and of ERK1/2 pathways is modified by exposure of S cells to cDDP, with a slightly increase of NF-κB as well as a downregulation of ERK1/2 pathways that might be associated with TNBC survival and its more aggressive characteristics. A similar hypothesis has been advanced in cervical carcinoma, for which increased cDDP-resistance has been observed in association with reduction of activation of the MEK to ERK2 pathway (in the presence of a MEK1-selective inhibitor, 2’-amino-3’-methoxyflavone) and involving, at least in part, an increase of NF-κB activation [50].

Two of the main metabolic features that R cells maintain overtime is lower anaerobic glycolytic activity (lower lactate levels, although increasing with time in each cell line), and lower glutaminolytic activity (lower glutamate and high glutamine levels). In S cells, both processes are more active, at all times, an expected hallmark of cancer metabolism [53]. However, interestingly, both pathways seem subdued in resistant cells. This seems to contradict the increases in glycolytic and glutaminolytic activities reported in relation to Pt(II)-resistance in colon and ovarian cancer cells [12, 15‐17]. We therefore suggest that reduced glycolytic and glutaminolytic activities may be a specific characteristic of TNBC cDDP-sensitive cells. Both observations are consistent with the noted significant decrease in ERK activity. The role of ERK (and JNK) pathways in the metabolic reprogramming of highly proliferating cells (such as cancer cells) has previously been discussed in detail [54]. These authors advance that increased ERK activity accompanies a high proliferative activity in cancer cells leading to a negative regulation of the enzyme pyruvate kinase isoform M2 (PKM2, responsible for PEP to pyruvate conversion), through phosphorylation. This low PKM2 activity induces an accumulation of upstream glycolytic intermediates, which are precursors of several biomolecules (e.g. amino acids, nucleotides and fatty acids). Interestingly and paradoxically, a low PKM2 activity is also related with an increased pyruvate conversion into lactate, through the action of lactate dehydrogenase A, at the expense of NADH which is oxidized to NAD⁺. This behavior seems to be descriptive of the cDDP-sensitive cells presently studied, whereas the R cells exhibit all opposite features: decreased ERK activity, decreased lactate formation and NAD⁺ levels (compared to S cells). A similar reasoning applies to the relationship between the ERK signaling pathway and glutaminolysis [54], as a decreased ERK activity in R cells should promote decreased expression of the c-Myc transcription factor, known to lead to lower glutaminolysis activity. Articulation with NF-κB expression (slightly higher in R cells), known to also affect glutaminolysis (as well as glycolysis and OXPHOS [55]), may help to keep glutamine pools high. Glutamine may not only serve glutaminolysis for anaplerosis (which is presently seen to be slowed down in R cells), but is also related to cancer cell stemness [56], a feature which has been generally related to therapy resistance, tumor dormancy and metastatic behavior [57]. As glutamine deprivation has been observed to relate to decreased stemness properties [56], we hypothesize that the richer glutamine pool found here in TNBC cDDP-resistant cells suggests a higher stemness capacity and, hence, a higher cell adaptability (and survival) once under cDDP exposure.

As mentioned before, low glutaminolysis and, thus, low levels of glutamate in R cells should contribute to lower TCA activity, as conversion of glutamate into α-ketoglutarate should decrease. However, it is interesting to note that a decrease in other anaplerotic amino acids, particularly at later culture times, (proline from 24 h; alanine, leucine, phenylalanine, tyrosine and particularly lysine, at 48 h) may suggest their preferential use as precursors into TCA intermediates and pyruvate precursor, respectively (Fig. 7), compared to methionine and branched-chain amino acids (BCAAs) isoleucine and valine, which exhibit higher levels in R cells. BCAA metabolism has been related to cancer resistance in general [58], but it is interesting to note the different leucine behavior in R cells: decreased, as opposed to increased levels of isoleucine and valine. Indeed, leucine metabolism (in particular through the activity of branched-chain amino acid aminotransferase 1, BCAT1) has been recently suggested to lead to activated mTOR-mediated autophagy which, in turn, increases cDDP-resistance [59]. We therefore propose that lower levels of leucine may be related to such mechanism.

Furthermore, the levels of TCA intermediates detected by NMR were strongly dependent on culture time, which suggests a possible modulation of TCA activity overtime. The low NAD⁺ levels (at all time points, and not replenished by lactate production) could either arise from its use in an activated TCA cycle, and/or reflect the general significantly lower availability of nucleotides in R cells, as will be discussed below. Previous reports [54] support the hypothesis that ERK activation promotes aerobic glycolysis and ATP synthesis, subsequently used for phosphorylation. Hence, in R cells, where ERK is less active, ATP synthesis is expected not to be significantly stimulated. Indeed, ATP levels are lower than in S cells at 0 h, although tending towards equivalent levels at later times (note the lower ADP/ATP ratios at 24 h due to ATP increase). We suggest that a later enhanced ATP synthesis in R cells may be a reflection not of ERK activity but, rather, of an adaptive later TCA activation, although the precise dynamics of this pathway overtime requires further investigation, at this stage. These relatively higher ATP levels may also explain the relatively elevated PCr levels (supported by increased Cr and sarcosine, Fig. 7) noted in R cells. Cr to PCr interconversion is an important energy buffer mechanism, which produces high energy PCr particularly in cells with high requirements of energy such as cancer cells [60]. Furthermore, the PCr/Cr ratio has been related to metastasis and proliferation and we advance that a higher PCr/Cr ratios may be related to higher cDDP-resistance.

In addition, increased levels of taurine (the oxidized form of hypotaurine) and GSH are clear discriminators of R cells, indicating an interesting interplay of compounds related to antioxidant protection mechanisms, including methionine (related to taurine through cysteine, and to GSH through the transsulfuration pathway [61] (Fig. 7). Although reactive oxygen species (ROS) were not quantified in this work, the NF-κB signaling pathway is believed to be closely related to oxidative stress [62]. Hence, we propose that the maintenance of high, nearly-constant, GSH levels in R cells, compared to S cells, along with high increasing taurine levels (whereas taurine remains lower and constant in S cells), may indicate that the antioxidative mechanisms in R cells rely preferably on the hypotaurine/taurine pair, rather than on GSH/GSSG. Furthermore, taurine increase has also been reported in ovarian cDDP-resistant cells [12, 13], possibly due to the overexpression of the taurine transporter (TauT) that leads to intracellular taurine accumulation, which in turn is suggested to result in the inhibition of cDDP uptake [13, 63].

The decrease in ERK activity noted in this work may also be related to the marked overall low levels of nucleotides and several of their derivatives, mainly involving adenine, uracil and inosine (including UDP-Glc/GlcA and uridine diphosphate N-acetyl-glucosamine UDP-GlcNAc), which make up a nitrogen-base-depleted metabolic profile in R cells, which is more stable overtime than in S cells (Fig. 7). Indeed, a decreased ERK activity and its correlation with increased PKM2 activity may explain the decreased biosynthesis of nucleotides, as described above [54]. However, lower levels of nucleotides may also be related with their use as building blocks as dNTPs to support nucleotide excision repair (NER). This has been reported as a major resistance mechanism against cDDP in several types of cancer [64, 65] and is believed to justify cell death when repairs are not possible [66]. Additionally, poly(ADP-ribose) polymerase (PARP) enzymes, especially PARP1, may determine DNA damage response and maintenance of genome stability through their involvement in NER [as well as in other mechanisms such as base excision repair (BER) and homologous recombination (HR)] [67]. Therefore, we hypothesize that PARP enzymes may have an active role in a more efficient DNA repair in R cells. Upon exposure, these enzymes may interfere with the formation of cDDP-adducts with DNA’s purine bases (adenine and guanine). We hypothesize that the setup of the R cell line, through exposure to low concentrations of cDDP, may have activated these enzymes. As PARP proteins require NAD⁺ to act [67, 68], it is possible that the low NAD⁺ levels in R cells may also reflect this mechanism.

Finally, R cells are also depleted in choline and GPC, the former increasing overtime, while GPC remains stable. This reflects disturbances in membrane metabolism but the exact variation pattern contrasts with results characterizing cDDP-resistance in ovarian cancer cells, which were characterized by increased levels of GPC (although confirming decreased levels of choline as described here) [13]. This relationship between choline compounds and the exact nuances of membrane remodeling mechanisms characterizing cDDP-resistance in TNBC remains unclear, at this stage. We furthermore suggest that the changes observed may reflect distinct lipid metabolic features (eventually detectable by lipid metabolomics) characteristic of this type of resistance in TNBC.