Pyrazine ring-based Na+/H+ exchanger (NHE) inhibitors potently inhibit cancer cell growth in 3D culture, independent of NHE1

The Na+/H+ exchanger-1 (NHE1) supports tumour growth, making NHE1 inhibitors of interest in anticancer therapy, yet their molecular effects are incompletely characterized. Here, we demonstrate that widely used pyrazinoylguanidine-type NHE1 inhibitors potently inhibit growth and survival of cancer cell spheroids, in a manner unrelated to NHE1 inhibition. Cancer and non-cancer cells were grown as 3-dimensional (3D) spheroids and treated with pyrazinoylguanidine-type (amiloride, 5-(N-ethyl-N-isopropyl)-amiloride (EIPA), 5-(N,N-dimethyl)-amiloride (DMA), and 5-(N,N-hexamethylene)-amiloride (HMA)) or benzoylguanidine-type (eniporide, cariporide) NHE1 inhibitors for 2–7 days, followed by analyses of viability, compound accumulation, and stress- and death-associated signalling. EIPA, DMA and HMA dose-dependently reduced breast cancer spheroid viability while cariporide and eniporide had no effect. Although both compound types inhibited NHE1, the toxic effects were NHE1-independent, as inhibitor-induced viability loss was unaffected by NHE1 CRISPR/Cas9 knockout. EIPA and HMA accumulated extensively in spheroids, and this was associated with marked vacuolization, apparent autophagic arrest, ER stress, mitochondrial- and DNA damage and poly-ADP-ribose-polymerase (PARP) cleavage, indicative of severe stress and paraptosis-like cell death. Pyrazinoylguanidine-induced cell death was partially additive to that induced by conventional anticancer therapies and strongly additive to extracellular-signal-regulated-kinase (ERK) pathway inhibition. Thus, in addition to inhibiting NHE1, pyrazinoylguanidines exert potent, NHE1-independent cancer cell death, pointing to a novel relevance for these compounds in anticancer therapy.

With more than 2 million estimated new cases in 2018, breast cancer remains the leading cause of cancer death in women globally 1 . While the luminal and HER2-enriched breast cancer subtypes can be treated with hormone receptor-and antibody-based targeted therapies, treatment of basal-like breast cancers still relies on chemotherapy 2 . In all subtypes, acquired or de novo treatment resistance remains a paramount obstacle 2 . The tumour microenvironment plays a central role in drug resistance, by limiting exposure of the tumour cells to anti-cancer drugs, and through selection for highly aggressive cancer cells 3,4 . A hallmark of the tumour microenvironment is profound acidity, caused by the high metabolic activity and increased acid extrusion of the rapidly growing tumour cells. Upregulation of acid extrusion is a ubiquitous characteristic of aggressive tumour cells, and we and others have shown that knockdown (KD) or genetic ablation of net acid-extruding transporters reduces tumour growth in several cancer models [5][6][7][8][9][10][11] . This renders inhibition of such transporters, alone or as combination therapy, a promising therapeutic approach, as suggested already decades ago 12 . The Na + /H + exchanger isoform 1 (NHE1, SLC9A1) is a major regulator of intracellular pH (pH i ) and is widely explored as a target in cancer as well as in other diseases (see 9,13 ). (D) spheroids, respectively. One-way ANOVA test with Tukey's multiple comparisons post-test was used to determine statistically significant differences between treatment groups. * and # denotes significant differences between the treatment condition relative to DMSO and between two treatment conditions, respectively. For

EIPA, but not cariporide, potently reduces cell viability in MCF-7 and MDA-MB-231 spheroids.
Pharmacological inhibition of NHE1 using EIPA or cariporide sensitizes p95HER2-expressing MCF-7 human breast cancer cells grown in 2D culture to cisplatin (a purine crosslinker, which has an effect similar to that of DNA-alkylating agents) chemotherapy 31,32 . We therefore first asked whether NHE1 inhibitors can sensitize cancer cells to clinically relevant anticancer treatments. To maximize relevance to in vivo conditions, we grew cells as 3D spheroids, which mimic the tumour microenvironment and better model anticancer treatment response than 2D cultures 3,19,20,22 . Native MCF-7 cells -a model of luminal A breast cancer -were grown for 2 days as spheroids, followed by 7 days of treatment with the anti-oestrogen tamoxifen (2 µM), cariporide (10 µM), EIPA (10 µM), or a combination of tamoxifen and either inhibitor. The tamoxifen concentration was chosen based on a dose-response screen ( Supplementary Fig. S1), and concentrations of cariporide and EIPA were chosen to ensure inhibition at the high Na + concentration and serum content of growth medium, compared to the low Na +and serum-free conditions used to determine Ki values. Spheroid growth was monitored by brightfield imaging (Fig. 1A), and a cell viability assay was performed on day 9 (Fig. 1B). As expected, 2 µM tamoxifen treatment resulted in spheroids with visibly frayed edges from day 7 and about 45% reduced cell viability at day 9 compared to untreated controls (p = 0.0181). Cariporide had no detectable effect on spheroid appearance or cell viability. In contrast, EIPA treatment resulted in small, irregular spheroids, and reduced day 9 viability to about 10% of that of control cells (p < 0.0001). Combination of tamoxifen and EIPA exacerbated spheroid disintegration and further reduced viability to less than 5% of that of controls (p < 0.0001). Propidium iodide (PI) staining was performed to visualize the spatial arrangement of cell death in the spheroids ( Supplementary Fig. S2). PI staining was visible in spheroid cores only in control-and cariporide-treated spheroids, but throughout the spheroids after EIPA panel 1B, the p-value is 0.0116 and 0.0147 for control vs. tamoxifen + cariporide and tamoxifen vs. tamoxifen + EIPA, respectively, while for panel 1D the p-value is 0.0011 and 0.0143 for control vs. chemotherapy + cariporide and for EIPA vs. chemotherapy + EIPA, respectively. Error bars denote SEM. 3-6 n. (E,G) Representative traces demonstrating BCECF-fluorescence as a function of time during pH i recovery of MCF-7 (day 4) (E) and MDA-MB-231 (day 4) (G) spheroids after NH 4 Cl pre-pulse. Fluorescence is depicted as the 485/440 nm ratio of BCECF intensity. 2-4 n (F,H) Quantification of pH i recovery rate of MCF-7 (pvalue = 0.343 for control vs. both cariporide and EIPA) (F) and MDA-MB-231 (p-value = 0.348 for control vs. cariporide, and 0.0461 for control vs. EIPA) (H) spheroids determined as the slope of the first 120 s after maximum acidification. Error bars denote SEM. 2-4 n. Statistical significance determined using a one-way ANOVA test with Tukey's multiple comparisons test. 2-3 n. (2020) 10:5800 | https://doi.org/10.1038/s41598-020-62430-z www.nature.com/scientificreports www.nature.com/scientificreports/ treatment, consistent with the near-complete loss of cell viability. Treatment with tamoxifen resulted in evenly distributed, sparse PI staining of the nearly disintegrated spheroids.
We next subjected human triple-negative breast cancer (TNBC) MDA-MB-231 cells to a chemotherapeutic regimen (18.75 nM cisplatin, 18.75 nM doxorubicin and 0.0625 nM 5-FU, doses chosen based on a pilot dose-response screen, Supplementary Fig. S1). This combination mimics the clinical TNBC-treatment termed CAF, except that we used another alkylating agent, cisplatin, to replace cyclophosphamide, which has to be metabolized in the liver for conversion to its active DNA-alkylating form and therefore cannot be used in cell culture. Compared to untreated controls, viability was reduced to about 45% by the chemotherapeutic regimen (p = 0.0009), by about 50% by 10 µM EIPA (p = 0.0005), and further reduced, to about 25%, by the combination of chemotherapy and EIPA (p < 0.0001), whereas cariporide had no effect (Fig. 1C,D).
Importantly, both EIPA and cariporide inhibited recovery of pH i following an acid load in MCF-7 (Fig. 1E,F) and MDA-MB-231 spheroids (Fig. 1G,H) as well as in 2D culture ( Supplementary Fig. S3). The dramatically different effects of EIPA and cariporide on cell viability thus did not reflect different capacity for inhibition of NHE1 activity at the doses used. In MDA-MB-231 spheroids, cariporide appeared less potent than EIPA in inhibiting pH i recovery. Given the much tighter architecture of MCF-7-compared to MDA-MB-231 spheroids 33 , it seems unlikely that cariporide penetrated the former more effectively than the latter, and we suggest that the apparent lower potency of cariporide in MDA-MB-231 cells reflects the upregulation of other net acid extrusion pathways in these cells during spheroid growth (see Discussion).
Collectively, these results show that Luminal A and TNBC breast cancer cells grown as 3D spheroids exhibit remarkably reduced cell viability upon long-term treatment with the pyrazinoylguanidine-type NHE1 inhibitor EIPA, but not with the benzoylguanidine-type inhibitor cariporide. This effect is partially additive to that of clinically relevant chemotherapy and does not reflect differential effects of the compounds on NHE1 activity.

EIPA-induced loss of viability is cell-type specific and most pronounced in cancer cells.
To determine whether the dramatic effect of EIPA on 3D growth was cancer cell specific, we cultured a series of cancer-and non-cancer cell lines as spheroids and exposed them to EIPA (0-10 µM) for 7 days ( Fig. 2A-C). The cell lines investigated were: T47D (breast cancer, luminal A), SKBr-3 (breast cancer, HER2-positive), HCT116 (colon cancer), BxPC-3 (pancreatic cancer), MCF10A (non-tumorigenic mammary epithelial cells), and NIH3T3 (mouse fibroblasts). In the cancer cells, the greatest loss of viability was seen for MCF-7 cells, followed by MDA-MB-231, BxPC-3 and SKBr-3 cells. Growth of non-carcinogenic cells was less affected by EIPA: spheroids of NIH3T3 fibroblasts were fully unaffected up to 10 µM, the highest concentration used, whereas viability of MCF10A spheroids was if anything increased by EIPA at concentrations up to 5 µM, yet decreased at 10 µM EIPA ( Fig. 2A-C). Cariporide at 10 µM had no detectable effect on spheroid viability for any of the cell lines tested ( Supplementary Fig. S4).
These results show that EIPA cytotoxicity is cell type dependent, with strong loss of viability in some breast cancer subtypes, while other cell types are less or not detectably affected by EIPA.

Cytotoxic effects of other NHE1 inhibitors in MCF-7 and MDA-MB-231 spheroids.
To understand which functional groups are associated with cytotoxicity of NHE1 inhibitors, we next compared the ability of various pyrazinoylguanidine-and benzoylguanidine-type NHE1 inhibitors to induce cell death in MCF-7 and MDA-MB-231 spheroids. Table 1 shows the chemical structures and estimated dose resulting in a 50% reduction in cell viability (EC 50 ) compared to reported 50% inhibitory concentration (IC 50 ) values for NHE1 inhibition and the lipophilicity (LogD and/or LogP values). Ki represents the inhibition constants for NHE1 activity, as reported in the literature. Figure 2D-G shows corresponding dose-response curves for amiloride, DMA, HMA, and eniporide. For MCF-7 spheroids, the order of cytotoxicity was HMA > EIPA > DMA > amiloride, with an EC 50 of HMA of only 2.5 µM. MDA-MB-231 cells were generally less sensitive than MCF-7 cells, but the order of cytotoxicity was similar. Neither cariporide nor eniporide, another benzoylguanidine-type compound with a reported IC 50 for NHE1 of only 4.5 nM, had detectable effects on spheroid viability.
These results demonstrate that pyrazinoylguanidine-type NHE1 inhibitors potently reduce viability in MCF-7 and MDA-MB-231 spheroids, whereas benzoylguanidine-type compounds have no detectable effect. As both classes of compounds inhibit NHE1 activity in these spheroids to a similar extent, and reported Ki values for NHE1 are lowest for the benzoylguanidines (Fig. 1E-H The EIPA-mediated cytotoxicity of MCF-7 and MDA-MB-231 cells is NHE1-independent. To conclusively determine whether the cytotoxic effects were NHE1-independent, MCF-7 and MDA-MB-231WT cells and two NHE1 CRISPR/Cas9 KO clones for each cell type 33 were grown as spheroids and subjected to EIPA (0-10 µM) as above (Fig. 3). Notably, the effect of EIPA on viability was identical in WT and NHE1 KO spheroids, for both MCF-7 and MDA-MB-231 cells. Similarly, the response of MCF-7 NHE1 shRNA KD spheroids (~50% reduced NHE1 expression compared to vector controls 33 ) to cariporide, EIPA, and tamoxifen treatment was essentially identical to that of WT MCF-7 cells (compare Fig. 1 to Supplementary Fig. S5).
These results demonstrate that the cytotoxic effect of EIPA is independent of NHE1. www.nature.com/scientificreports www.nature.com/scientificreports/ Pyrazinoylguanidine NHE1 inhibitors accumulate intracellularly in breast cancer cell spheroids.
Amiloride and its derivatives are weak bases, with reported pK a values ranging between 8.5 and 8.7 for amiloride, DMA, HMA, and EIPA 15,34 . In contrast, cariporide and eniporide are weak acids, with pK a values of around 6.2 and 6.0, respectively 35,36 . We therefore speculated that the cytotoxicity of pyrazinoylguanidine-type NHE1 inhibitors might reflect their accumulation in acidic organelles during spheroid growth. To investigate this, we exploited the intrinsic fluorescence properties of these compounds 37 at 420 nm after excitation at 340-400 nm. Spheroids were grown for two days, treated with cariporide, EIPA or HMA (all at 10 µM), and accumulation examined after 5 days in continued presence of the inhibitor, compared to 4 h after inhibitor addition (when the compounds have permeated the spheroids but accumulation is expected to be minimal). Compared to control-and cariporide-treated spheroids, spheroids treated with EIPA or HMA displayed a markedly higher intrinsic fluorescence already after 4 h of treatment, with peak fluorescence intensity observed at around 390 nm (Fig. 4A,B). The absolute fluorescence intensity at this wavelength, quantified over multiple experiments, increased significantly from 4 h to 5 days of treatment for EIPA-and HMA-treated spheroids -despite their smaller volume (caused by inhibitor-induced loss of viability). In marked contrast, controland cariporide-treated spheroids exhibited no detectable change in intrinsic fluorescence, despite the similar spectral properties (Fig. 4A). To assess where in the cells the compound accumulation occurred, cells in 2D culture were treated with EIPA (10 µM) or HMA (10 µM) for 2-4 days, fixed and their intrinsic fluorescence analysed by UV illumination (Fig. 4C). Compound accumulation was detectable as the appearance of distinct, primarily punctate structures in  . The pyrazinoylguanidine NHE1 inhibitors EIPA and HMA accumulate intracellularly in MCF-7 spheroids and monolayer cultures. MCF-7 spheroids grown for 2 or 7 days were treated with cariporide (10 µM), EIPA (10 µM) or HMA (10 µM) for 4 h or 5 days, respectively. DMSO served as a vehicle for HMA. Inhibitor fluorescence intensity was measured for three individual spheroids per condition. (A) Traces of an excitation scan from 340-400 nm. Emission was recorded at 420 nm. Error bars denote SEM. 3-4 n (except for DMSO which is 1-2 n). (B) Quantification of fluorescence intensity. A one-way ANOVA with Tukey's multiple comparisons test was used for testing statistical significance between conditions. * and **** denote p = 0.0353 and <0.0001, respectively, between treatment and their respective control; #### denotes p < 0.0001 between 4 h and 5 days (390 nm) of the same condition, § § denotes p = 0.0016 between cariporide and HMA, and § § § § denotes p < 0.0001 between cariporide and EIPA. Error bars denote SEM. 3-4 n (except for DMSO which is 1-2 n). It should be noted that the fluorescence intensities have not been corrected for spheroid volume, and since HMA-and EIPA-treated spheroids are much smaller on day 7 than control-and cariporide-treated Scientific RepoRtS | (2020) 10:5800 | https://doi.org/10.1038/s41598-020-62430-z www.nature.com/scientificreports www.nature.com/scientificreports/ the perinuclear region, which were never observed in control-or cariporide-treated cells. The punctate accumulation overlapped partially, but not completely, with staining for the lysosomal marker lysosomal-associated membrane protein 1 (LAMP-1) (Fig. 4D, arrow heads). No marked changes in lysosomal organization were detectable. In some cells, EIPA as well as HMA treatment induced the appearance of large perinuclear vacuolar structures ( Fig. 4D-F, arrows). This pattern of vacuolization is characteristic of the cell death process known as paraptosis, characterized by damage to, and dilation of mitochondria and endoplasmic reticulum (ER) 39,40 . We therefore assessed the status of mitochondria, ER and Golgi by immunofluorescence analysis of their respective markers, mitochondrial import receptor subunit TOM20, protein disulfide isomerase (PDI), and giantin. EIPA treatment was associated with extensive mitochondrial fragmentation (Fig. 4E), intracellular vacuolization and a diminished Golgi area (Fig. 4G). The vacuoles appeared to be "pushing" on the nucleus (Fig. 4D-F) and surrounded by mitochondrial membrane (Fig. 4E) and ER (Fig. 4F).
These results show that EIPA and HMA undergo marked accumulation in cancer cell spheroids. Accumulation appears to be primarily lysosomal, and is associated with extensive perinuclear vacuolization, mitochondrial fragmentation, and ER reorganization.
EIPA and HMA treatment elicit DNA damage, ER stress, mitochondrial fragmentation, autophagic arrest, and PARP cleavage. To gain insight into the mechanisms underlying the cytotoxic effects of the pyrazinoylguanidines, MCF-7 spheroids were lysed 48 h after treatment with 10 µM EIPA, and immunoblotted for phosphorylated retinoblastoma protein (pRb), Poly-ADP Ribose Polymerase (PARP) and cleavage of caspase-7 (the main apoptotic effector caspase in MCF-7 cells, which lack caspase-3). This revealed that EIPA treatment elicited a 5-fold reduction in pRb level (p < 0.0001), an 8-fold increase in the c-PARP/PARP ratio (p = 0.0132), and slight caspase-7 cleavage (Fig. 5A-C). Some NHE1 inhibitors have been proposed to intercalate in DNA and cause DNA damage 41 . In agreement with this, Histone 2AX phosphorylation (γH2AX) was increased in EIPA-treated spheroids (Figs. 5D, 5 days of treatment). Finally, with increasing EIPA concentrations, PI-permeable cells were distributed throughout the spheroids rather than restricted to the hypoxic core as in controls (Fig. 5E).
To allow more detailed subcellular analysis, MCF-7 cells were next grown as monolayers for 4 days under control conditions or in presence of 10 µM EIPA or HMA, and subjected to immunofluorescence analysis and Western blotting. Similar to the results obtained in 3D spheroids, EIPA and HMA exerted potent cytotoxic effects, while there was no detectable toxicity of cariporide (Fig. 5F). Both EIPA and HMA elicited PARP cleavage and γH2AX upregulation, but not p53 upregulation (Fig. 5F). Pyrazinoylguanidine treatment also induced marked nuclear upregulation of the ER stress-induced transcription factor C/EBP Homologous Protein (CHOP (also known as GADD153)) ( Fig. 5F,G). Autophagy is dependent on recruitment of ER membrane for autophagosome formation, and CHOP can induce cell death through both stimulation of apoptosis and inhibition of autophagy 42 . Accordingly, EIPA-and HMA treatment was associated with marked accumulation of LC3B-II and p62, the combination of which is indicative of arrested autophagy (Fig. 5F).
The Ser/Thr kinases Extracellular Signal-Regulated Kinase (ERK) and Akt are major survival kinases, which are regulated in many stress conditions, including ER stress and paraptosis. These kinases can contribute to both survival and death signalling, and are targets of clinical anticancer therapies 43,44 . MCF-7 cells were therefore subjected to treatment with EIPA with and without ERK and Akt inhibitors (Fig. 5H). ERK, but not Akt, activity was increased upon EIPA treatment, and both kinases were effectively inhibited by their inhibitors, U0126 and Akti, respectively. Strikingly, ERK inhibition strongly potentiated EIPA-induced CHOP upregulation and PARP cleavage (Fig. 5H). This suggests that ERK activation under these conditions may be a compensatory survival mechanism, inhibition of which synergizes with EIPA in eliciting MCF-7 cell death.
Together, these results indicate that the cytotoxic effects of pyrazinoylguanidine-type NHE1 inhibitors are multifactorial and include pronounced DNA damage, ER stress, mitochondrial fragmentation, autophagic arrest, and paraptosis.

Discussion
The key novel findings of this work are that 5-substituted pyrazinoylguanidine-type NHE1 inhibitors accumulate in cancer cells grown under 3D conditions and exert potent, multifaceted cytotoxic effects unrelated to NHE1 inhibition. In marked contrast, the benzoylguanidine-based compounds, cariporide and eniporide, exert no detectable toxicity. To the best of our knowledge, this is the first study in which the effect of these inhibitors has been comprehensively studied across compounds and compared to CRISPR/Cas9-mediated ablation of NHE1. It is also the first in which this has been done in a 3D setting, which is superior to 2D models in predicting in vivo response to drug treatment 3,19,22 . www.nature.com/scientificreports www.nature.com/scientificreports/ Pyrazinoylguanidine cytotoxicity is NHE1-independent and cell-type specific. Genetic reduction or ablation of NHE1 reduces proliferation, invasiveness and in vivo growth of a wide range of cancer cells 5,6,[8][9][10]33 . Small molecule NHE1 inhibitors have therefore been explored as an anticancer approach, alone or in combination treatment schemes 6,31,32,45 . However, evidence from 2D systems has shown that at high concentrations (20-40 µM, and up to 500 µM for amiloride, compare to IC 50 values for NHE1 inhibition, Table 1), some pyrazinoylguanidine-type NHE1 inhibitors exert NHE1-independent cytotoxicity 24,25,28,46 . Furthermore, pyrazinoylguanidine-type inhibitors inhibit several other NHE isoforms than NHE1 and some additionally target the epithelial Na + channel (ENaC) and Na + /Ca 2+ exchangers [13][14][15] . In cancer cells, the long-term effects of pyrazinoylguanidine-type NHE1 inhibitors, which have generally been assumed to be downstream of NHE1 inhibition, include decreased proliferation, reduced metastatic potential and -viability, and sensitization to chemotherapy 32,45,47,48 . There are, however, reports of NHE1-unrelated effects of these compounds, including dysregulation of ER Ca 2+ homeostasis 24 , reactive oxygen species production 25 , urokinase plasminogen activator inhibition 26 , and cell death 27,28 . The present study is the first to use NHE1 CRISPR/Cas9 KO in conjunction with extensive pharmacological analyses, to unequivocally establish the NHE1-independence of these effects. Interestingly our work mirrors recent reports taking advantage of CRISPR/Cas9 mutagenesis to demonstrate that the effect of many anticancer drugs in clinical trials is unrelated to their presumed target 29,30 .
Our first observation was that EIPA treatment lead to a dramatic loss of breast cancer spheroid viability, both as monotherapy and in combination with tamoxifen or chemotherapy. In contrast, the benzoylguanidine-type NHE1 inhibitor cariporide had no effect, in agreement with our previous observations in MCF-7 spheroids 33 . Both compounds were effective in inhibiting NHE1, albeit cariporide seemed less effective than EIPA in MDA-MB-231 cells grown in 3D. In contrast to MCF-7 cells, MDA-MB-231 cells exhibit strong V-ATPase relocalization to the plasma membrane upon adaptation to extracellular acidification (L.E.L., S.F.P, unpublished) -a condition encountered in the 3D environment. We therefore speculate that the incomplete inhibition of net pH i recovery by cariporide in MDA-MB-231 spheroids may reflect a cariporide-insensitive contribution to pH i regulation that underlies the incomplete inhibition in these spheroids. Critically, this does not affect the study conclusions, because the markedly different toxicity of EIPA vs cariporide was also seen in both MCF-7 spheroids and in 2D, where the two compounds had equal inhibitory effects on NHE1.
Unequivocally establishing the NHE1-independence, EIPA-induced cytotoxicity was indistinguishable between NHE1 KO and WT spheroids from several breast cancer types. EIPA-induced toxicity was also pronounced in several other cell types. While non-cancer cells were generally less sensitive than cancer cells, they were not fully unaffected and some cancer cells were also relatively insensitive to pyrazinoylguanidine-induced cytotoxicity. This warrants caution against previous proposals from 2D studies that these compounds are cancer cell-selective 46 and shows that sensitivity must be determined on a cell type-to-cell type basis. In contrast, cariporide had no effect on spheroid viability for any of the cell lines tested, in agreement with previous reports 28,49 .

Pyrazinoylguanidine cytotoxicity in 3D is associated with intracellular compound accumulation.
The use of 3D models is essential to anticancer drug studies, because they mimic the tumour microenvironment 3,19,20 , including the dense extracellular matrix (ECM), tortuous extracellular space, and acidic pH, as low as 6.0 in the tumour core 9 . In our study, we employed 3D spheroids prepared both with and without exogenous addition of matrix components, reflecting the varying capacity of different cancer cells for cell-cell adhesion. Given the propensity of cancer cells for endogenous matrix production 50 , and the tight spheroid morphology, we expect all spheroids used to exhibit a tortuous extracellular space. However, also spheroids prepared without exogenous ECM addition are suitable for in vitro modelling of tumour drug response 19,22 . Under these conditions, weak base drugs accumulate in the extracellular space and in acidic organelles (which are also more acidic at acidic extracellular pH 51 ) than in the less acidic cytosol and nucleus (the phenomenon of "ion trapping"). For chemotherapeutic drugs destined to function in the nucleus, this can lead to physiological treatment resistance, i.e. reduced drug concentration at the site of action 12,23 . The benzoylguanidine-type inhibitors cariporide and eniporide are weak acids with pK a values of 6.2 and 6.0, respectively 35,36 . Based on their chemical properties, they are thus expected to preferentially partition into the cytosol, whereas the pyrazinoylguanidines, which are weak bases (pK a values 8.5-8.7 15,34 ), should partition into acidic compartments. In congruence with this, EIPA and HMA, which were profoundly cytotoxic, accumulated most markedly in perinuclear lysosomes.
The pyrazinoylguanidines studied in this work bear a similar architecture comprising a guanidine group connected to a flat aromatic ring by an amide bond. However, they also have noticeable differences. In contrast to cariporide and eniporide, which are built on a phenyl aromatic ring, amiloride, DMA, EIPA and HMA are built on a heterocyclic pyrazine ring. This structure bears two nitrogen atoms at positions 1 and 4, whose electronegativity influence the electron distribution, making this ring slightly electron deficient. Furthermore, these Lewis basic N lone electron pairs render pyrazine weakly basic. The pyrazinoylguanidine compounds tested in this study all feature a 6-Cl halogen atom, meta to the 2-acyl guanidine on the pyrazine ring. While the Cl is more electronegative than C, the overall consequence of the substitution of C-H for C-Cl produces a net increase in the hydrophobicity of the molecule (see 15 ). Cariporide and eniporide bear a sulfonyl group at the same position that has very strong electron withdrawing properties giving these molecules, which bear a large hydrophobic component (see below), amphiphilic properties.
Both classes of molecules have groups in para position to the acylguanidine moiety. These are a single amino group (amiloride), and hydrophobic 5-N alkyl or cycloalkyl moieties (DMA, EIPA, HMA), isopropyl (cariporide) or a pyrrole group (eniporide). Lastly, pyrazinoylguanidines bear an ortho 3-amino group that, similar to the 5-amino group, is likely to affect the pKa value of the guanidine group by resonance. This group is not present in cariporide and is replaced by a methyl in eniporide. Taken together, the cellular toxicities of amiloride, DMA, EIPA and HMA observed here correlate well with the size and hydrophobicity of their 5-N substituents (see also Table 1). It is tempting to propose that these hydrophobic substituents affect the overall ability of these molecules to partition into membranes, penetrate cells and accumulate in intracellular compartments. Furthermore, compared to cariporide and eniporide, the structures of amiloride derivatives feature a number of potentially reactive basic groups. This makes these molecules more likely to interact and react with intracellular components. Beyond these observations, the numerous differences between these two classes of molecules, as listed above, make it impossible to unequivocally assign the observed toxicity to a single chemical group. Thus, future studies should narrow down the precise functional group(s) responsible for the toxic effects and equally important assess the potential NHE1-independent effects of bi-or tricyclic NHE1 inhibitors such as zoniporide, SM-20550, and KB-R9032.
This work shows that the potent NHE1-independent cytotoxic effects of EIPA and HMA in long-term 3D culture occur below or at the concentration range generally applied for NHE1 inhibition (i.e. at concentrations as low as 1 µM, compared to the 5-10 µM widely used for NHE1 inhibition). This effect involves accumulation in the endo-/lysosomal compartment and is likely to be exacerbated by the increased size and acidity of this compartment under the acidic extracellular conditions of tumours, as mimicked in spheroids 51 . Our findings also provide an explanation for the apparently paradoxical finding that the cytotoxicity of pyrazinoylguanidine-type inhibitors is greatest under acidic conditions 52 . While these compounds have not yet been used in clinical trials, it seems highly probable that a similar time-dependent accumulation would be seen upon long-term treatment of cancer patients. In conjunction with the tendency of the cytotoxic effect to be more profound in cancer cells than in normal cells, this makes the pyrazinoylguanidines potentially interesting in anticancer therapy, as discussed further below.
Pyrazinoylguanidines elicit profound ER stress, reduced autophagic flux, and paraptosis. The pattern of accumulation indicated that the pyrazinoylguanidine compounds exert their cytotoxic effects at least in part by disrupting lysosome-and ER function. Consistent with this notion, EIPA and HMA treatment caused profound upregulation of the ER stress-induced transcription factor CHOP and increased expression of both LC3B-I and -II and p62. The accumulation of both lipidated LC3B-II and of p62 is indicative of stalled autophagic flux, as expected if lysosomal function is disrupted. NHE6, −7, and 9 are present in the endo-lysosomal pathway, however, a role for inhibition of these isoforms seems unlikely to explain our findings as, similar to NHE1, these transporters are sensitive to both benzoyl-and pyrazinoylguanidines 13 . The same is true for NHE2, which is the only other plasma membrane NHE expressed at non-negligible levels across most cell types studied here (Human Protein Atlas, mRNA levels). The upregulation of CHOP is consistent with an earlier report that EIPA and HMA elicit ER Ca 2+ depletion in endothelial cells in 2D culture 24 . The reduction of autophagic flux is likely secondary to both ER stress 42 and lysosomal dysfunction, precluding autophagosome formation. Consistent with the former, our 2D experiments revealed induction of massive perinuclear vesicles surrounded by mitochondrial-and ER markers. This pattern is characteristic of paraptosis -a form of caspase-independent cell death associated with perinuclear swelling of ER and mitochondria 39,40 . Finally, in congruence with the reported EIPA-and HMA-induced ER Ca 2+ depletion 24 , paraptosis is dependent on ER-and mitochondrial Ca 2+ dysregulation in a manner proposed to involve Ca 2+ flux from ER to mitochondria at ER-mitochondrial contact sites 40 . EIPA and HMA treatment also caused γH2AX upregulation, congruent with the notion that some pyrazinoylguanidines (such as amiloride and EIPA) can intercalate into DNA 41 . However, given the lack of p53 upregulation in the highly sensitive MCF-7 cells (which are p53-wild type), and the reported resistance of MCF-7 cells to DNA-damage-induced apoptosis 53 , it seems unlikely that DNA damage contributes substantially to pyrazinoylguanidine-induced death, at least in MCF-7 cells.
Perspectives and relevance to basic research and anticancer treatment. This work clearly establishes that effects of pyrazinoylguanidine NHE inhibitors cannot be taken as evidence of a role of NHEs, warranting caution in their use and meaning that the existing literature using these compounds should be re-evaluated. It remains clear, however, from studies using other tools, that NHE1 is important for the growth and development (2020) 10:5800 | https://doi.org/10.1038/s41598-020-62430-z www.nature.com/scientificreports www.nature.com/scientificreports/ of many cancers 5,6,[8][9][10]33 . Consistent with previous reports 6,32,54 , our results point to an additive effect of EIPA in combination with tamoxifen, combination chemotherapy and ERK inhibition. In our hands, EIPA-induced cytotoxicity was particularly strongly potentiated by inhibition of ERK, a highly interesting target in e.g. KRAS mutated cancers 43,55 . This suggests a substantial potential for combination therapy which should be investigated in future studies. Specific questions to be explored include effects of sub-lethal doses of pyrazinoylguanidine-type NHE1 inhibitors, combination with other anticancer treatments, and staggered treatment combinations, taking into account the potentially different latency of action of EIPA and other treatments tested.
An important implication of these findings is that in NHE1-dependent cancers, pyrazinoylguanidine-type NHE1 inhibitors are likely to exert dual, NHE1-dependent and -independent anticancer effects, including ER stress and paraptosis. Paraptosis is attractive as an anticancer death pathway because it is independent of pro-apoptotic Bcl-2 proteins and effector caspases, which are frequently downregulated in cancer cells. In addition to exploring combination treatments, future work should establish whether the cytotoxicity is sufficiently cancer specific to be useful in a therapeutic context and identify cancer types likely to benefit from this treatment. Interesting possibilities to explore would include KRAS-or ERK-driven tumours with high NHE1 expression.
In conclusion, we show that pyrazinoylguanidine-type NHE1 inhibitors exert potent, cell type-dependent, NHE1-independent cytotoxic effects on cancer cells grown as 3D spheroids, at concentrations generally assumed to specifically inhibit NHE1. We propose that the cytotoxicity of the pyrazinoylguanidines reflects their hydrophobicity in conjunction with weak base properties, collectively resulting in their trapping and accumulation in acidic organelles, leading to DNA damage, ER stress, paraptosis, and stalling of autophagic flux. We suggest that both the NHE1-dependent and -independent effects of these compounds may render them attractive in the context of anticancer therapy.

Methods
A table describing relevant equipment and settings is provided as Supplementary Table S1.

Antibodies and reagents.
A table containing all employed antibodies as well as their cellular target, supplier and catalogue number is provided as Supplementary Table S2.
Lysis of cells in 3D culture. Spheroids were gently collected in Eppendorf tubes, followed by one wash in ice-cold PBS and lysis in LB for 5-10 min at RT with intermittent vigorous vortexing. Sonication, homogenization and centrifugation were as described for 2D culture.

Measurements of inhibitor accumulation. Monolayer cells. 24 h after seeding on glass coverslips, cells
were treated with cariporide, EIPA or HMA for 2 days before fixation in 2% paraformaldehyde (PFA, Sigma, #47608) and visualization employing Olympus IX83 microscope with a Yokogawa scanning unit, using a 60X/1.4 NA objective and aquisition using CellSens Dimension software. Intensity adjustment was performed using ImageJ software.
Spheroids. Cells were seeded for spheroid formation either 2 or 7 days prior to measurement of accumulation and treated with cariporide, EIPA or HMA for 4 h or 5 days, respectively. Fluorescence intensity of the inhibitors was measured on a Nikon Eclipse Ti microscope employing EasyRatioPro software (PTI) at wavelengths 340-400 nm with 5 nm band intervals. Emission was recorded at 420 nm and data was analysed in Excel and Graphpad Prism.
Immunofluorescence analysis. Cells cultured on glass coverslips were treated with cariporide or EIPA for 2 or 4 days. Cells were washed in cold PBS, fixed in either 2% PFA for 15 min at RT and 30 min on ice or 4% PFA for 30 min at RT and washed twice in TBST. Cells fixed in 2% PFA were permeabilized in 0.5% Triton-X-100 (Plusone, #17-1315-01) in TBST and blocked using 5% Bovine Serum Albumin (BSA, Sigma-Aldrich, #A7906). Cells fixed in 4% PFA were quenched in 50 mM NH 4 Cl in TBST and permeabilized and blocked in 5% BSA with 0.1% Saponin (Sigma-Aldrich, #47036). The cells were incubated with primary antibodies diluted in TBST plus 1% BSA overnight at 4 °C or for 1.5 h, RT, and with secondary antibodies, also in TBST plus 1% BSA, at RT for 1 h, followed by a 5 min DAPI (1:1000) incubation, wash in 1% BSA, and mounting in 2% N-propyl-gallate mounting medium (Sigma-Aldrich, #P3130). Fluorescently labeled proteins were visualized employing an Olympus IX83 microscope with a Yokogawa scanning unit, using a 60X/1.4 NA oil immersion objective and CellSens Dimension software. No image processing except merges, zooms, and color balance was performed. Image overlays and adjustments of the intensities were performed using ImageJ software.
Inhibition of Akt and MEK. Cells were seeded in 6-well plates for 24 h prior to 48 h-treatment with EIPA (10 µM), Akti (1 µM), U0126 (10 µM) or a combination thereof. Lysates were prepared as described in the above section on immunoblotting. (2020) 10:5800 | https://doi.org/10.1038/s41598-020-62430-z www.nature.com/scientificreports www.nature.com/scientificreports/ Propidium Iodide (PI)-staining of spheroids. On day 9, 100 µL medium was removed, spheroids were washed three times with heated PBS, 100 µL PBS containing 4 µM PI was added per well (2 µM final concentration) and plates were incubated at 37 °C for 10-15 min protected from light. Spheroids were washed three times in heated PBS and visualized employing an Olympus IX83 microscope with a Yokogawa scanning unit, using a 10X air objective. Subsequent image adjustments were performed using ImageJ software. Statistical analysis. Data are shown as representative images or as means with standard error of the mean (SEM) error bars as indicated. After verifying normal distribution of the data, one-way analysis of variance (ANOVA) followed by Tukey's or Dunnett's multiple comparisons post-test was used to test for significant differences when more than two groups were compared, while a paired or unpaired Student's t-test, as relevant, was employed when only two groups were analysed.

Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.