Long-Chain and Very Long-Chain Ceramides Mediate Doxorubicin-Induced Toxicity and Fibrosis

Doxorubicin (Dox) is a chemotherapeutic agent with cardiotoxicity associated with profibrotic effects. Dox increases ceramide levels with pro-inflammatory effects, cell death, and fibrosis. The purpose of our study was to identify the underlying ceramide signaling pathways. We aimed to characterize the downstream effects on cell survival, metabolism, and fibrosis. Human fibroblasts (hFSF) were treated with 0.7 µM of Dox or transgenically overexpressed ceramide synthase 2 (FLAG-CerS2). Furthermore, cells were pre-treated with MitoTempo (MT) (2 h, 20 µM) or Fumonisin B1 (FuB) (4 h, 100 µM). Protein expression was measured by Western blot or immunofluorescence (IF). Ceramide levels were determined with mass spectroscopy (MS). Visualizations were conducted using laser scanning microscopy (LSM) or electron microscopy. Mitochondrial activity was measured using seahorse analysis. Dox and CerS2 overexpression increased CerS2 protein expression. Coherently, ceramides were elevated with the highest peak for C24:0. Ceramide- induced mitochondrial ROS production was reduced with MT or FuB preincubation. Mitochondrial homeostasis was reduced and accompanied by reduced ATP production. Our data show that the increase in pro-inflammatory ceramides is an essential contributor to Dox side-effects. The accumulation of ceramides resulted in a lipotoxic shift and subsequently mitochondrial structural and functional damage, which was partially reversible following inhibition of ceramide synthesis.


Introduction
Dox is a chemotherapeutic agent that belongs to the family of anthracyclines. It is used to treat various tumors. Due to its severe side-effects, especially cardiotoxicity, it is only used in limited situations [1]. Long-term use of Dox leads to dilated cardiomyopathy and fibrosis [2]. The underlying mechanisms of Dox toxicity are still not fully understood; in particular, its effects on cardiac fibrosis have not yet been thoroughly analyzed [3].
Ceramides are sphingolipids with a sphingosine backbone and attached fatty acids of various chain lengths [4] and they are involved in multiple essential cell signaling pathways [5]. Long chain ceramides have an attached fatty acid with a chain length of more than 14 C-Atoms (C14) and very-long chain ceramides with more than 22 C-Atoms (C22) and their accumulation are negatively associated with cell viability [6]. One possible way of ceramide generation is via ceramide synthases (CerS) and especially CerS2 as part of the salvage and sphingomyelin pathway [7][8][9]. The expression of CerS2 differs based on the origin of the individual cells and is mainly responsible for the formation of very long-chain ceramides [10] The accumulation of ceramides is associated with various diseases in different tissues, but its regulation and respective signaling pathways are not fully understood [6,[11][12][13].
We hypothesized that the cytotoxic side-effects of Dox are linked to long-chain and very long-chain ceramide accumulation in fibroblasts of the failing myocardium. Thus, we focused on hFSF under treatment with Dox and studied the role of CerS2 in the underlying mechanisms concerning ceramide accumulation, mitochondrial viability, and apoptosis.  Figure S1A) was performed to determine the most suited Dox concentration. Eth-D2 staining identified 0.7 µM Dox and was used for the subsequent experiments. Additionally, time-dependent lethality measurement was conducted with 0.7 µM Dox to confirm that a certain survival rate and recovery and proliferation were still observable (Supplementary Figure S1B). MT was used to assess the positive effects of the reduction in mitochondrial ROS on mitochondrial homeostasis. Concentration was adapted from Zhang et al. [14]. FuB was used to determine the beneficial effects of ceramide reduction in mitochondrial ROS production and homeostasis. Concentration was selected from Bouhet et al. [15]. Both concentrations were tested for lethality with Eth-D2, but no toxicity was detected (data not shown).

Dox Increases CerS2 Protein Expression and Ceramide Levels
CerS2 expression was assessed with IF and showed a 1.5 ± 0.19 (p < 0.001)-fold increase in CerS2 ( Figure 1A). Increased CerS2 expression was also verified with WB (1.6 ± 0.16, p = 0.02) ( Figure 1B). Increased levels of CerS4 and CerS5 were not observed. CerS6 showed a slight increase (data not shown).
Total ceramides were calculated by the addition of all measured ceramide species of each individual measurement (C14, C16, . . . , C24:1). Afterward, mean values and standard deviation were calculated and normalized to respective control.
MitoTracker staining showed an utterly ruptured structure, mitochondrial distribution, and only a rudimentary mitochondrial network ( Figure 2D). Electron microscopy revealed beginning structural damage following 0.4 µM Dox treatment, which was characterized by an accumulation of multilamellar bodies ( Figure 2E, red arrows). Simultaneously, physiological tube-like mitochondria were also identifiable in the same cell ( Figure 2E, white arrows). An early autophagosome that encapsulated a mitochondrion targeted for digestion was detected ( Figure 2E, black arrow) and 0.4 µM Dox was selected because the structural damage in 0.7 µM Dox was too severe.
MT, as well as FuB, restored in part the physiological tube and net-like structure of the mitochondria ( Figure 2F,G). Both inhibitors were not completely able to restore the structure to normal level.

Dox impairs Mitochondrial Homeostasis and Function and Causes Apoptosis
Mitochondrial homeostasis was further assessed by the measurement of mitochondrial fusion-fission related gene expression ( Table 2). MFN1 was and remained reduced. MFN2 was increased after Dox and did not change. OPA1 was also reduced and was not reversible.
The fission gene DRP1 was and remained reduced. Mff expression was not affected by Dox but was reduced under MT or FuB. Mitochondrial fission gene FIS1 showed no expression changes after Dox but was reduced with MT or FuB.
Metabolism related gene expression was assessed to determine the effects of Dox induced mitochondrial impairment on cell metabolism. The focus was primarily on fatty acid metabolism, because of the suspected lipotoxic shift caused by increased ceramides. Fatty acid metabolism associated gene DGAT1 was increased and did not change under MT or FuB. An elevated expression level was also detected for DGAT2 and returned to normal level after pre-treatment with either MT or FuB. Expression of ATGL showed significantly increased level and remained comparatively high with MT. FuB slightly decreased ATGL expression. Dox led to a significant increase in CD36, which was unaffected by MT and was slightly decreased with FuB. Expression of the mitochondria membrane localized CPT1B was and remained reduced under any condition. PDK4 was reduced under any condition. GDH remained unchanged (Table 3).  Seahorse analysis revealed significantly reduced mitochondrial ATP production, characterized by a reduced Oxygen Consumption Rate (OCR) (0.3 ± 0.44, p < 0.001) ( Figure 3A). The reduction did neither recover with MT (0.2 ± 0.34, p < 0.001) nor FuB (0.2 ± 0.70, p < 0.001).
IF revealed translocation of cytochrome C from the mitochondria into the cytoplasm (20.3 ± 0.33, p < 0.001) ( Figure 3B). MT and FuB pre-treatment returned cytochrome C intensity to normal level ( Figure 3C

Dox Promotes Fibrosis
Fibrosis is characterized by the disassembly of extracellular matrix proteins and replacement with fibrotic ones, e.g., collagen. The resulting "scar" impairs myocardial homeostasis and function. Because of this, fibrosis-associated genes were measured with qPCR. Dox showed a significant increase in all measured mRNA expressions related to the activation of fibrosis (Table 4). MT reduced MMP14, TIMP2 and TGF-β, but still showed increased tendencies for all measured genes. Except for TIMP1 and ACTA2, all of the measured genes returned to a normal expression level after FuB. Increased ACTA2 expression was further verified with IF. IF showed more defined α-smooth muscle actin (ACTA2) filaments in Dox treated cells (white arrows) compared to the diffused structure in the solvent co ( Figure 5A). FuB pre-treatment had no effects on ACTA2 mRNA or protein expression ( Figure 5B, Table 4).

Dox Promotes Fibrosis
Fibrosis is characterized by the disassembly of extracellular matrix proteins and replacement with fibrotic ones, e.g., collagen. The resulting "scar" impairs myocardial homeostasis and function. Because of this, fibrosis-associated genes were measured with qPCR. Dox showed a significant increase in all measured mRNA expressions related to the activation of fibrosis (Table 4). MT reduced MMP14, TIMP2 and TGF-β, but still showed increased tendencies for all measured genes. Except for TIMP1 and ACTA2, all of the measured genes returned to a normal expression level after FuB. Increased ACTA2 expression was further verified with IF. IF showed more defined αsmooth muscle actin (ACTA2) filaments in Dox treated cells (white arrows) compared to the diffused structure in the solvent co ( Figure 5A). FuB pre-treatment had no effects on ACTA2 mRNA or protein expression ( Figure 5B, Table 4). Collagen was measured in the cell supernatant of Dox treated fibroblasts. Dox was replaced with a growth medium and the cells we incubated for additional 72 h. An increase in collagen was detected after this 72 h in Dox (1.48 ± 0.12, p < 0.001) ( Figure 5C). FuB pre-treatment returned collagen to control level after 72 h incubation (0.95 ± 0.24, p = 0.71) ( Figure 5D). We did not measure collagen in the cell supernatant and ACTA2 protein expression of MT preincubation because of the minor effects on the fibrosis related gene expression (Table 4).

CerS2 Knockdown Improves Dox-Mediated Mitochondrial Damage and Fibrosis
To further confirm the role of CerS2 in Dox-mediated cell toxicity, we knocked down CerS2 and subsequently treated the transfected cells with 0.7 µM Dox. Transfection success was assessed with WB (0.45 ± 0.33, p = 0.03) ( Figure S2A). FACS was performed, to confirm the functional effects ( Figure S2B). FACS measurement showed increased ceramide level in siCerS2 + 0.7 µM Dox compared to siRNA control + solvent co. The positive control (siRNA control + 0.7 µM Dox) showed further elevated ceramide levels compared to the other two conditions.

CerS2 Knockdown Improves Dox-Mediated Mitochondrial Damage and Fibrosis
To further confirm the role of CerS2 in Dox-mediated cell toxicity, we knocked down CerS2 and subsequently treated the transfected cells with 0.7 µM Dox. Transfection success was assessed with WB (0.45 ± 0.33, p = 0.03) ( Figure S2A). FACS was performed, to confirm the functional effects ( Figure S2B). FACS measurement showed increased ceramide level in siCerS2 + 0.7 µM Dox compared to siRNA control + solvent co. The positive control (siRNA control + 0.7 µM Dox) showed further elevated ceramide levels compared to the other two conditions.

CerS2 Overexpression Leads to an Increased Ceramide Production
We confirmed overexpression with IF and determined a 1.6 ± 0.1 (p < 0.001) increase ( Figure 7A). Additionally, transfection showed a 10.9 ± 0.13-fold (p < 0.001) increase in the marker protein FLAG and a 4.2 ± 0.44 fold (p = 0.04) increase in CerS2 ( Figure 7B). Changes of CerS5 and CerS6 were not observed (data not shown).

CerS2 Overexpression Leads to an Increased Ceramide Production
We confirmed overexpression with IF and determined a 1.6 ± 0.1 (p < 0.001) increase ( Figure 7A). Additionally, transfection showed a 10.9 ± 0.13-fold (p < 0.001) increase in the marker protein FLAG and a 4.2 ± 0.44 fold (p = 0.04) increase in CerS2 ( Figure 7B). Changes of CerS5 and CerS6 were not observed (data not shown). MS revealed increased total ceramide levels as well as very long chain ceramides. Moreover, a significant increase in C16, C18, C22, C24 and C24:1 was detected (Table 5).  MS revealed increased total ceramide levels as well as very long chain ceramides. Moreover, a significant increase in C16, C18, C22, C24 and C24:1 was detected (Table 5).
FuB reduced ceramide level in FLAG-CerS2. C14 was significantly reduced with FuB while other ceramides returned to normal level (Table 6).
Qualitative MitoTracker staining displayed severely impaired mitochondrial integrity and structure ( Figure 8D). The structural impairment was also confirmed with electron microscopy. FLAG-CerS2 showed multilamellar bodies ( Figure 8E, red arrows), indicating late lysosomes and autophagy. Pre-treatment with MT or FuB restored in part the physiological net-like structure of the mitochondria (Figure 8D,F). The beneficial effects of FuB seem to be more prominent than the positive effects of MT.

FLAG-CerS2 Leads to Reduced Mitochondrial Structure and Function and Induces Apoptosis
Mitochondria fusion gene MFN1 was significantly reduced. MT returned MFN1 expression to control level, while FuB significantly reduced MFN1 level compared to FLAG-TC and FLAG-CerS2. MFN2 expression was reduced and returned after MT and FuB. Preincubation with both inhibitors reduced OPA1 expression. FuB reduced DRP1 expression. Mff was reduced in FLAG-CerS2 and remained reduced with FuB. MT returned Mff expression to normal level. FIS1 showed no expression changes but was reduced with MT or FuB. FIS1 was reduced compared to FLAG-CerS2 under FuB (Table 7). Seahorse analysis showed a significant reduction in mitochondrial ATP production (0.82 ± 0.10, p = 0.003). MT (1.07 ± 0.16, p = 0.63) and FuB (1.0 ± 0.22, p = 0.98) returned ATP production to normal level ( Figure 9A-C).
DGAT1 expression was not altered. FuB reduced DGAT1 compared to FLAG-CerS2. DGAT2 expression was and remained reduced with MT and FuB. CD36 expression was increased. FuB pre-treatment returned CD36 to normal level and was significantly reduced compared to FLAG-CerS2. CPT1B expression was reduced. PDK4 and GDH were unaltered (Table 8).  DGAT1 expression was not altered. FuB reduced DGAT1 compared to FLAG-CerS2. DGAT2 expression was and remained reduced with MT and FuB. CD36 expression was increased. FuB pre-treatment returned CD36 to normal level and was significantly reduced compared to FLAG-CerS2. CPT1B expression was reduced. PDK4 and GDH were unaltered (Table 8). Cytochrome C was detectable in the cytosol of FLAG-CerS2 (17.2 ± 0.24, p < 0.001) ( Figure 10A). MT and FuB preincubation reduced Cytochrome C to a normal level ( Figure 10B, C). Increased cleaved caspase 9/caspase 9-ratio was detected (1.34 ± 0.10, p = 0.01) and returned to baseline level after MT (1.3 ± 0.16, p = 0.09) and FuB  FLAG-CerS2 led to increased MMP8 expression and showed increased tendencies for MMP9 and TGF-ß. Under FuB MMP8 and MMP9 expression were no longer detectable (Table 9).

Discussion
Despite severe side-effects that include cardiotoxicity, Dox remains one of the most established and effective chemotherapeutic agents [1]. Nevertheless, the specific mechanisms of Dox-mediated cardiotoxicity and myocardial fibrosis remain to be elucidated [16,17]. Here, we have shown that Dox treatment of fibroblasts results in an increase in ceramides, also observable in CerS2 overexpression. This ceramide accumulation led to inflammation, as characterized by increased cytokines and mitochondrial ROS levels. This resulted in mitochondrial damage promoting apoptosis, all consistent with a lipotoxic phenotype. The detrimental effect on mitochondria is improved after reducing mitochondrial ROS levels and reducing ceramide levels. Additionally, the knockdown of CerS2 through gene silencing showed comparable results to FuB incubation with respect to mitochondrial homeostasis. Therefore, the current data link the cytotoxic properties of Dox to the accumulation of pro-inflammatory and lipotoxic ceramide species. Furthermore, mitigation of ceramides is beneficial for fibrosis reduction. Our findings highlight the role of long-chain and very long-chain ceramides, specifically produced by CerS2, to Dox associated adverse effects in fibroblasts.
The increase in ceramide species resulted in a lipotoxic phenotype. This is supported by the upregulation of the respective fatty acid associated genes DGAT1 [18], DGAT2 [19], ATGL [20] and CD36 [21] and downregulation of CPT1B [22]. The lipotoxic shift, in turn, instigated increased mitochondrial ROS production followed by mitochondrial damage and the promotion of apoptosis. The observed increase in mitochondrial ROS production and cytokine mRNA levels is characteristic of inflammation and cellular damage [23]. Interestingly, even though ceramide levels were elevated in FLAG-CerS2, mitochondrial ROS levels and cytokine mRNA expression were higher in Dox-treated cells. These findings indicate a contributing role of ceramide species in Dox-mediated cytotoxicity and additional mechanisms involved in the pro-inflammatory activation. This is also supported by the reducing effects of FuB. FuB is an unspecific CerS inhibitor and returned all previously increased ceramide species equally in Dox and FLAG-CerS2. Additionally, in FLAG-CerS2 FuB significantly reduced C14 ceramides. CerS2 has no affinity for C14 ceramides [24] which was observable by the unchanged C14 level in FLAG-CerS2. That hints, that the detected reduction in C14 by FuB could be linked to CerS other than CerS2. The reduction in ceramides is characterized by reduced IL-6, IL-1β, mitochondrial ROS production, and recovered mitochondrial homeostasis.
MT had no effects on cytokine mRNA levels in Dox-treated cells, while it increased TNFα expression and reduced IL-6, IL-1β level in FLAG-CerS2, even though mitochondrial ROS levels were reduced in both. The increase in TNFα could be explained the varying transfection efficiency. This is further supported by the increased tendency of TNFα in only FLAG-CerS2 cells. The reduction of IL-1β and IL-6 by FuB shows that the reduction in mitochondrial ROS can be beneficial for the inflammatory status if there is a singular cause, such as specifically increased long chain and very long chain ceramides. The siCerS2 data support these findings due to the same expression patterns such as Dox-treated hFSF. Here, also, all cytokines were increased and showed no improvement by the reduction in CerS2, followed by Dox incubation.
This further indicates that the effects of Dox on inflammation are multiple and too profound to be reduced by the inhibition of mitochondrial ROS or ceramides alone.
There was no increase in the TNFα expression in Dox treated cells. Three out of 11 experiments showed very high values for solvent co, highly increasing standard deviation and preventing the detection of the expected increase.
Interestingly, neither the preincubation with MT or FuB reduced mRNA level of the ROS-Scavengers SOD2 or GPx1, indicating that mitochondrial ROS level or ceramides are not involved in the expression regulation. This is further indicated by the respective FLAG-CerS2 data, which showed also no effects on SOD2 and GPx1. Only Catalase was positively affected by FuB preincubation implying a correlation between ceramides and this specific ROS scavenger. Dox led to an increase in the ROS inducer NOX2, which was not affected by MT. Our data concur with McLaughlin et al. who also identified increased NOX2 level after DOX treatment [25]. The reduction in mitochondrial ROS level with MT had no beneficial effects on NOX2 expression, indicating that mitochondrial ROS production does not affect this specific cytosolic ROS inducer. FuB reduced NOX2 expression when compared to Dox-treated cells, but still showed an increased tendency. This shows the beneficial effects of ceramide reduction not only on mitochondrial, but also cytosolic ROS production. This is further supported by the siCerS2 data. Here also, SOD2 Gpx1 were increased while Cat returned to a normal level comparable to the Dox data. NOX2 was also increased but showed comparable level to FuB + Dox. Mitochondrial fusion and the resulting broad netlike structures are essential for mitochondrial function and cellular viability [26]. Increased ceramide levels cause structural impairment of mitochondria net-like structure. Electron microscopy showed a pathological accumulation of multilamellar bodies in Dox-treated and transfected cells, indicating mitochondrial digestion and autophagy in both cases. These results concur with Law et al., who also identified increased mitophagy in CerS2 overexpressing human cardiomyocytes.
MFN1 and MFN2 were significantly reduced in FLAG-CerS2, supporting the microscope data and the findings of Chen et al. [27]. Pre-treatment with MT and FuB improved reduced MFN2 expression but had no effects on MFN1. This indicates that the decrease of MFN1 is not associated with ceramide levels but with CerS2 protein levels due to the CerS non-specificity of FuB. FuB pre-treatment further reduced MFN1 expression in FLAG-CerS2. This paradox result could be explained by the varying transfection efficiency but further emphasizes the non-beneficial effects of FuB on MFN1. The respective Dox data also support this association. Here too, a significant reduction in MFN1 was observed, which was unaffected by FuB-caused ceramide reduction. The knockdown of CerS2 and subsequent Dox treatment led to the increase in previously observed MFN1 reduction. The knockdown data show that the expression of MFN1 is not regulated by ceramide levels but directly by the CerS2 protein level. Further research is necessary to elucidate the detailed interconnection between CerS2 and MFN1.
The MFN2 expression was unaffected by any of the used pre-treatments or the knockdown of CerS2, followed by Dox incubation. These results are contradictory to previous findings [28]. We hypothesize that the observed upregulation of MFN2 could be a compensatory mechanism by the human fibroblasts. This potential counter mechanism was also observed in HL-1 cells [25] but still needs further investigations. FuB returned previously reduced MFN2 expression to baseline levels in FLAG-CerS2 and suggests that the MFN2 increasing effects of Dox superimpose the beneficial ceramide reducing effects of FuB.
Mff [29] and FIS1 [30] were unchanged in Dox-treated fibroblasts, where an increase might have been expected. This could hint that the detected mitochondrial fission happens within the used incubation time of 24 h and that Mff and FIS1 might be upregulated within this period and returned to a normal level afterward. The measurement of Mff and FIS1 at an earlier time point could clarify the observed effects Preincubation with MT or FuB reduced the mRNA level of both genes, which was also detectable in FuB pre-treated FLAG-CerS2. The knockdown data concur and support our findings. Mff and FIS1 were significantly reduced in knockdown cells followed by Dox treatment, showing the advantageous effects of ceramide reduction in Dox-mediated cell toxicity.
The reduction in mitochondrial ROS, either primary by MT or as a secondary effect of FuB, improves mitochondrial structure more by the downregulation of fission-related genes Mff and FIS1 instead of the upregulation of the fusion-related genes (MFN1, MFN2, and OPA1). The fragmented mitochondrial structure also decreased mitochondrial function. Seahorse analysis revealed a significant reduction of OCR and, therefore, of mitochondrial ATP production in Dox and FLAG-CerS2. Interestingly, even though both inhibitors showed reduced inflammation and improved mitochondrial structure, no mitochondrial function improvement in Dox-treated cells was detected. This indicates that the detrimental effects of 0.7 µM Dox are too strong to be improved by ceramide reduction alone. Coherent to the impairment of mitochondrial homeostasis, is the induction of apoptosis. Dox and FLAG-CerS2 induce apoptosis by translocation of Cytochrome c and Casp9 activation. This activation is reduced by the preincubation of MT and FuB. Furthermore, siCerS2 also improves cell vitality by reduced Casp9 activation when compared to only Dox treated cells.
Fibrosis is a regenerative response to a pathogenic stimulus that can become maladaptive and result in a loss-of-function of the damaged tissue [31]. Our data have shown an increase in MMP8 [32], MMP9 [33], and MMP14 [34] expression. Furthermore, the mRNA expression of the fibrosis marker TGF-ß [35] and ACTA2 [36] were also increased. Additionally, the upregulation of TIMP1, TIMP2 in Dox treated cells also indicates fibrosis progression [37]. Again, the reduction in ceramides also reduced fibrosis related gene expression except for ACTA2. The reduction of CerS2 with siRNA produced similar results to FuB pre-treatment. FLAG-CerS2 mainly affected MMP8 and MMP9 expression. Both genes already appeared at a later cT cycle in the qPCR in the control, and therefore they were no longer detectable after preincubation with FuB. The other measured genes only showed an increased tendency, which could be explained by varying transfection efficiency. CerS2 overexpression did not affect TIMP1 and TIMP2 expression and remained unchanged under FuB, while FuB reduced TIMP2 expression under Dox. This could also indicate that the mRNA expression is not directly linked to ceramides but to other CerS than CerS2. Still, the reduction in ceramides reduce fibrosis progression in Dox treated fibroblasts.
Our results implicate a strongly contributed connection between increased ceramides and the detrimental effects of Dox. Even so, some of our results were contrary to our expectations. An example is the mRNA expression of MFN1 in FLAG-CerS2. Here, the overexpression led to a significant decrease in the gene, which was further enhanced upon the pre-treatment with FuB, where an increase in MFN1 would have been expected. A single or combination of reasons could explain such paradox results. Variations of transfection efficiency or too low concentrations or incubation times of either FuB or MT could be an example for the observed contrary or not strong enough effects. Further modifications and improvements could explain the observed effects.

CerS2 Overexpression and CerS2 Silencing
The original unedited p3xFLAG-CMV7 vector (FLAG-TC) and the p3xFLAG-CerS2 overexpression plasmid were designed and acquired from GENEWIZ (South Plainfield,

Statistical Analysis
Statistical analysis was performed with SigmaPlot 14.0. Data are depicted as mean values ± standard deviation as our standard way of data presentation and SD is a better representation of the data distribution and normalized to the respective control. Significance was calculated with the student's t-test (p ≤ 0.05). The graphs were designed with GraphPad Prism 8.