Secretion of Acid Sphingomyelinase and Ceramide by Endothelial Cells Contributes to Radiation-Induced Intestinal Toxicity

Despite advances in treatment delivery techniques, normal tissue toxicity remains a major side-effect of radiotherapy (1). For example, patients with abdominal or pelvic malignancies treated by high doses of ionizing radiation (IR) can develop radiation gastrointestinal (GI) syndrome, an acute deleterious radiotoxicity causing lethal damage to the GI tract. This syndrome has been considered to be strictly dependent on dysfunction of intestinal stem cells or clonogens (2). When the death of those clonogens induces the collapse of crypt/villus unit, the persistence of some clonogens is sufficient to repair damaged gut epithelium. We and others challenged this single-cell target model and suggested a more complex multitarget scenario (3). In particular, a single high dose of radiation triggering microvascular endothelial apoptosis, constitutes a primary lesion leading to intestinal stem cell dysfunction, crypt damage, villi denudation, organ failure, and ultimately death by GI syndrome (3–5). This lethality can be prevented after pharmacologic inhibition of endothelial cells apoptosis by intravenous injection of basic fibroblast growth factor (3), agonist of Toll-like receptor 5 (6), sphingosine 1-phosphate (7), or angiopoietin-I (8) prior to irradiation.

Endothelial cell apoptosis depends on the early generation of sphingolipid ceramide by acid sphingomyelinase (ASM; ref. 9). Within minutes after exposure to IR, the ASM pool, initially included inside lysosomes, is translocated on the external plasma membrane hydrolyzing sphingomyelin to ceramide. Once generated, ceramide allows membrane rafts agglutination to large lipid platforms and the activation of a p38 MAPK–dependent apoptosis (10, 11). Invalidation of sphingomyelin phosphodiesterase 1 gene (Smpd1) coding for ASM in mice impedes radiation-induced endothelial cell apoptosis, enhances clonogens survival and prevents GI syndrome (3, 5). Similar effects were observed in vivo when the formation of ceramide-enriched membrane platforms was blocked after administration of an anticeramide mAb (12). Altogether, these studies definitively highlight a key role played by the endothelium in radiation-induced GI toxicity. However, the precise mechanism linking ASM/ceramide-induced endothelial cell apoptosis and intestinal epithelial radiosensitivity remains poorly characterized.

Endothelial cell death triggers the secretion of paracrine factors mediating intercellular communication, which might be a mechanism contributing to radiation-induced acute intestinal injury. In fact, using a coculture model without cell–cell contact, we showed that 15 Gy-irradiated endothelial cells induced cell-cycle arrest and the death of nonirradiated T84 intestinal epithelial cells (13). These results suggest that the secretome of endothelial cells exposed to high doses of radiation could mediate a bystander response on epithelial cells. Identification of the paracrine factors within this secretome is thus required to efficiently prevent GI syndrome.

Extracellular sources of ASM and ceramide are able to activate a deleterious response to targeted cells (11). The smpd1 gene gives rise not only to a lysosomal acid sphingomyelinase (L-ASM), but also to a secretory acid sphingomyelinase (S-ASM) through differential trafficking of a common protein precursor. S-ASM is produced mainly by endothelial cells and its secretion is up-regulated under inflammatory conditions (14, 15). In contrast, extracellular L-ASM is the result of nonspecific release from damaged cells (16). In vivo, ASM isoforms as well as ceramide are detected in most body fluids (17). Their modulation are often associated with the development of pathologies such as lung emphysema (18), hemolytic anemia during Wilson disease (19), insulin resistance, and associated renal injury (20). Interestingly, circulating ASM and ceramide are also enhanced in blood of patients with lung and liver oligometastasis after radiotherapy in correlation with a greater antitumor efficacy (21).

Here we provide novel insights into the role of sphingolipids in radiation-induced intestinal toxicity. We show, for the first time, using relevant cellular, organoid and mouse models that secreted ASM and ceramide participate to the paracrine signal triggered by endothelial cells to neighboring noncancerous intestinal epithelial cells in response to radiotherapy, leading to intestinal toxicity.

Eight- to 12-week-old ASMKO and wild-type (WT) littermate C57BL/6 mice (kindly provided by R. Kolesnick, MSKCC, New York, NY) were housed in our animal facility (SFR F. Bonamy Inserm UMS026, Nantes, France) approved by the French Association for Accreditation of Laboratory Animal Care (AFSTAL; ethical authorization number: APAFIS# 17166-2019021514472102) and maintained in accordance with the regulations and standards of Inserm and French Department of Agriculture. Whole-body irradiation (WBR) was delivered by a Faxitron CP160 irradiator (Faxitron X-ray Corp.) at 1.27 Gy/minute. D609 (Sigma-Aldrich, F) was diluted in NaCl 0.9% and WT mice were injected intraperitoneally with 40 mg/kg of D609 30 minutes before and 5 minutes after irradiation.

Targeted irradiation of intestine was achieved with the XRAD225Cx irradiator, allowing small beams and image guided irradiation. 17 Gy were delivered at the center of intestine area (16.5 Gy were delivered to 80% of total intestine area) with two anteroposterior beams (225 kV, 13 mA, and 0.3 mm Cu filter) at 2.37 Gy/minute. The 2 cm cylindrical collimator was used with a distance source-target of 35.4 cm, providing a beam diameter of 2.3 cm and allowing to spare tissues around the intestine (Fig. 1H). Intestine accurate targeting of each mouse was obtained thanks to a 3D cone beam CT image.

Retro-orbital venous blood samples were collected at designated time points after irradiation, left to stand for 30 minutes, then centrifuged twice (6,000 × g, 15 minutes, 4°C). Serum samples were collected and stored at −80°C. Proximal jejunum obtained from euthanized animals was excised and rinsed with NaCl 0.9%. An antimesenteric border of 3-cm segment was incised, exposing the mucosal surface that was scraped before snap-freezing on dry ice and storage at −80°C. Then, mucosa was homogenized in ice-cold PBS and centrifuged (10,000 × g, 15 minutes, 4°C). Protein concentration was determined using the BCA Protein Assay Kit (Interchim). One-centimeter segment of proximal jejunum was fixed in 4% formaldehyde and embedded in paraffin blocks. Intestinal structure was determined on transverse sections of full jejunal circumferences after hematoxylin & eosin staining. Crypt–villus axis height was measured as described previously (22).

Primary human lung microvascular endothelial cells (HMVEC-L; Lonza) and human intestinal epithelial cell line T84 (European Collection of Authenticated Cell Cultures) were grown as described previously (13) and cell line authentication has been validated by Eurofins Genomics. Briefly, HMVEC-L cells were expanded to confluence in EBM-2 endothelial basal medium with 5% FBS, EGM-2MV supplements (Lonza). T84 cells were grown in DMEM: F12 (1:1, Gibco, Invitrogen) with 10% FBS, 1% l-glutamine, and 1% penicillin–streptomycin. Cells were irradiated at a rate of 1.6 Gy/minute by a Faxitron CP160 irradiator. Endothelial–epithelial coculture was established with confluent HMVEC-L in 12-well plates and proliferating T84 in Transwell inserts. Irradiation was delivered in two protocols: (i) endothelial and epithelial compartments were irradiated together; (ii) Transwell inserts containing unirradiated T84 cells were transferred into the 12-well plates where HMVEC-L had been just irradiated. All cocultures were further incubated for 24 hours before study.

Transfection of HMVEC-L was performed 36 hours before experiments using Dharmafect 4 (Dharmacon) according to manufacturer's instructions. Invalidation of ASM was achieved using Dharmacon ON-TARGET plus siRNA SMART pools (10 nmol/L; GE Healthcare) directed against human smpd1 gene containing the sequences GCACUGGGAUCAUGACUAC; GCACACCUGUCAAUAGCUU; CAGGUUACAUCGCAUAGUG; GGCACAACCUGGUAUAUCG. Scramble RNAi were performed with a pool of nontargeted sequences (UGGUUUACAUGUCGACUAA; UGGUUUACAUGUUGUGU- GA; UGGUUUACAUGUUUUCUGA; UGGUUUACAUGUUUUC- CUA).

T84 cells were incubated for 24 hours with increasing concentrations of human placental ASM (Sigma-Aldrich), C₂-ceramide (Biomol), natural C₁₆-ceramide (Avanti Lipids), and phosphocholine (Sigma-Aldrich). C₂- and C₁₆-ceramide were dissolved respectively in 20 mmol/L dimethylsulfoxide and dodecane:ethanol 2:98,v:v (respectively, DMSO and DE; Sigma-Aldrich). Final diluent concentrations in the medium were up to 0.5% and 0.05%, respectively. T84 cells were treated for 24 hours with conditioned medium collected after 24 and 72 hours from nonirradiated (control medium or CM) or 15 Gy–irradiated HMVEC-L (acute stress-associated phenotype or ASAP). CM or ASAP were pretreated with 5 μmol/L desipramine (Sigma-Aldrich) or 1 μg/mL of mouse anticeramide IgM antibody clone MID 15B4 (Enzo) or IgM isotype control (Beckman-Coulter) 1 hour before incubation with T84 cells. Cell number was determined by counting with Malassez slide. Cell death was measured in cell lysate with Cell Death Detection ELISA^plus kit (Roche Diagnostics) following manufacturer's instructions. Apoptosis was detected after incubation of T84 cells for 72 hours with CM or ASAP HMVEC-L medium including l μl/mL of CellEvent Caspase 3/7 (Thermo Fisher Scientific) and observation under brightfield and fluorescent microscopy (Nikon Eclipse Ti).

S-ASM and L-ASM enzymatic activities were assessed with 2 and 10 μL of mouse serum, respectively, or 10 μg of intestinal mucosa protein or 40 μL of concentrated HMVEC-L supernatant. Reaction was started by the addition of BODIPY FLC₁₂-sphingomyelin (Invitrogen; 1 nmol for serum or intestinal mucosa, 0.05 nmol for HMVEC-L supernatant) as fluorescent substrate (40 μL in 250 mmol/L sucrose containing 3% Triton X-100) and 5 mmol/L EDTA for L-ASM and 0.1 mmol/L ZnCl₂ buffer for total ASM. S-ASM measurement results from the subtraction of L-ASM to total ASM measurement. Reaction was processed at 37°C and stopped by chloroform/methanol (1:1, vol/vol). Samples were centrifuged at 1,000 × g for 5 minutes, after incubation for 30 minutes. Organic phase was harvested, evaporated under N₂, and resuspended in chloroform. Product separation was completed using silica gel Partisil K6 absorption thin layer chromatography (TLC) plates (Whatman) and chloroform/methanol (95:5, vol/vol). Resulting BODIPY-ceramide and remaining BODIPY-sphingomyelin were imaged with Chemidoc MP Imaging System (Bio-Rad) and quantified using ImageLab software.

Ceramide levels in HMVEC-L supernatants were semiquantified by DAG kinase assay. Briefly, lipids were extracted in chloroform/methanol/hydrochloric acid (1N; 100:100:1, vol/vol/vol) and hydrolyzed for 1 hour in 0.1 mol/L methanolic KOH at 37°C. 3.5 μg of DAG kinase and γ-[³²P]ATP (1 μL, 3,000 Ci/mmol) were added per sample for 30 minutes at room temperature, before stopping using 1N chloroform/methanol/hydrochloric acid (100:100:1, vol/vol/vol) and 15 mmol/L EDTA. Ceramide-1-phosphate was resolved using silica gel Partisil LK6D absorption TLC plates (Sigma-Aldrich) and chloroform/methanol/acid acetic (65:15:5, vol/vol/vol). ³²P incorporation was determined using a Typhoon 9410 phosphoimager and ImageQuant software. Ceramide amount per sample was extrapolated from a standard curve.

In mouse serum and intestinal mucosa, quantitative analysis was performed by GS-MS by using 50 μL of serum and 500 μg of proteins from the intestinal mucosa, respectively. Lipids were extracted in hexane/isopropanol (3:2, vol/vol). Ceramide was isolated using SPE cartridges (23). After methanolysis of ceramide and 0.5 nmol of internal standard C17:0-Cer (d17:1) in HCl/MeOH 1.25 mol/L, free sphingosine was derivatized for 45 minutes at 90°C in a mixture of acetic anhydride and pyridine (5:1). The sphingoid bases generated were analyzed on a 6890N GC system (Agilent) equipped with a HP-5MS column and coupled to a quadrupole 5973 mass spectrometer, operating in electron impact ionization mode. All samples were run in duplicate and reference samples were included in every run.

Phosphocholine levels in HMVEC-L supernatants and ceramide rate in mouse serum after intestinal irradiation, were quantified by LC/MS-MS performed on a Xevo TQD mass spectrometer with an ESI interface and ACQUITY H-Class UPLC System (Waters Corporation). A pool of reference standard solutions including nine ceramide species (Avanti Polar Lipids) was prepared with serial dilutions of methanol to obtain seven standard solutions, ranging from 1 to 500 nmol/L for Cer (d18:1/16:0); Cer (d18:1/18:0); Cer (d18:1/20:0) and ranging from 0 to 5 μmol/L for Cer (d18:1/22:0); Cer (d18:1/24:0); Cer (d18:1/24:1). Standards solutions and supernatants (10 μL) or plasma samples (10 μL) where then extracted with 500 μL of methanol/chloroform mixture (2/1, v/v) containing exogenous internal standard [IS; Cer (d18:1/17:0) 500 nmol/L]. Samples were mixed and centrifuged for 10 minutes at 20,000 × g (10 °C), and the supernatant dried under a nitrogen stream and solubilized in 100 μL of methanol. Sample (10 μL) were injected onto reverse-phase BEH-C18 column and the compounds were separated. Sphingolipids were detected by MS with the ESI interface operating in positive ion mode. Data acquisition and analyses were performed with MassLynx and TargetLynx version 4.1 software (Waters Corporation).

Intestinal crypts were collected from 6- to 10-week-old C57BL/6 mice. Small intestines were harvested 2 cm distal to the pylorus and 2 cm proximal to the cecum and flushed with ice-cold, Ca²⁺ and Mg²⁺-free PBS. Intestinal enteroids were generated as described previously (24). Briefly, an intestinal segment was opened longitudinally and transferred to 2 mmol/L ethylenediaminetetraacetic acid (5 mol/L EDTA, Life Technologies) in PBS and rocked, buried in ice for 30 minutes. After EDTA chelation, tissues were washed and hand-shaken for 1:30 minutes. The tissue was removed and crypts were filtered through a 70-μm cell strainer (Thermo Fisher Scientific). The crypts were pelleted and resuspended in growth factor–reduced Matrigel (Corning). Approximately 150 crypts per well were plated in 35 μL of Matrigel. Each well was then overlaid with minigut media (advanced DMEM/F12, 10 mmol/L HEPES, N2 supplement, B27 supplement, 2 mmol/L GlutaMAX, 50 ng/mL EGF and 100 U/mL penicillin/100 mg/mL streptomycin (Life Technologies) supplemented with 50% L-WRN conditioned media (ATCC). Medium was changed every 4 days. As explained in Supplementary Fig. S1, forming enteroids were treated 24 hours after the initial plating with minigut media supplemented either with 50% CM, or 50% ASAP including or not 100 μmol/L D609, or 10 mU/mL ASM, or 10 μmol/L C₁₆-ceramide. The following day, enteroids were irradiated with a Faxitron CP160 irradiator, and then cultured for 5 days prior study.

Murine enteroids were imaged in bright field using an InCell 2000 platform equipped with a 4× objective (GE Healthcare). Total surface of each well was acquired generating 16 images per well. Acquisition was also performed in depth to capture all enteroids embedded in Matrigel. Stacked images were processed and stitched in batch mode using Cell Developer (GE Healthcare). The stitched tiles were then aligned and an extended depth of focus was performed to generate one single image representing all enteroids present in a well (FIJI, open-source software). Enteroid culture morphometrics analysis (number of enteroids per well and area per enteroid) was performed using a standalone version of OrganoSeg software (25).

Enteroids in Matrigel were fixed with 2% paraformaldehyde for 30 minutes at room temperature, permeabilized with 0.5% Triton X-100 in PBS for 30 minutes, then blocked in 10% goat serum buffer for 2 hours at room temperature. Enteroids were sequentially incubated with primary and secondary antibody overnight at 4°C, followed by nuclear staining (Hoechst 33342) during 20 minutes. Coverslips were then mounted using Fluoromount-G. Images were captured on a Nikon A1 confocal microscope and analyzed using FIJI. The following pairs of primary and secondary antibodies were used: rabbit polyclonal anti-Ki67 with Alexa Fluor 568 goat anti-rabbit IgG, Alexa Fluor 488 Phalloidin (respective dilution 1:200, 1:400, and 1:200; all Thermo Fisher Scientific).

Experiments were performed at least three times. Data were analyzed by either Student t or ANOVA tests with 95% confidence estimation using Prism 7.0 (GraphPad Software). P value of less than 0.05 was considered to indicate statistical significance.

GI syndrome was observed in mice 72 hours after exposure to 15 Gy as shown by the severe and extensive lesions revealed by the almost complete crypt disappearance and partially denuded intestinal mucosa (Fig. 1A). A significant 27% reduction of crypt-villus length was observed in 15 Gy–irradiated as compared with nonirradiated C57Bl/6 mice (respective axis height: 426 ± 17 μm vs. 583 ± 24 μm, P ≤ 0.05; Fig. 1B). D609 treatment on irradiated WT mice significantly reduced intestinal radiotoxicity (Fig. 1A). Crypt-villus axis height in D609 treated WT mice was significantly higher than in irradiated WT mice and 20% lower than in sham-irradiated mice (502 ± 7 μm, P ≤ 0.05 vs. 15 Gy–irradiated mice and nonirradiated WT mice; Fig. 1B). As expected, invalidation of ASM in mice prevented radio-induced intestinal lesions with crypt-villi as long as the ones in nonirradiated WT or ASMKO mice (568 ± 33 μm, P = 0.5 vs. nonirradiated WT mice or ASMKO mice).

Then, ASM and ceramide were quantified in the serum of 15 Gy–irradiated WT and ASMKO mice. Irradiation of WT mice resulted in a significant serum enrichment of both S-ASM and L-ASM activities in a time-dependent manner after 15 Gy irradiation in WT mice (respective increase 2.12- and 2.79-fold 96 hours after 15 Gy vs. 0 Gy; Fig. 1C and D). Enhancement of ASM activities was concomitant with a sustained enrichment of ceramide concentration in the serum of irradiated WT mice starting as early as 30 minutes and remaining elevated for at least 96 hours (1.6-fold 96 hours after 15 Gy vs. 0 Gy; Fig. 1E). In contrast, ceramide concentration was not enhanced in ASMKO mice following radiation exposure, in agreement with their GI syndrome radioprotection (Fig. 1F). Indeed, ceramide concentration in the serum of ASMKO mice was significantly lower than those measured in WT littermates, 24 and 72 hours postradiation (both time P < 0.05; Fig. 1G). Interestingly, activities of S- and L-ASM from intestinal mucosa of WT mice were only transiently enhanced 24 hours after irradiation (respective increase 1.9- and 1.6-fold 24 hours after 15 Gy vs. 0 Gy; Supplementary Fig. S2A and S2B) and did not correlate with the time-dependent enrichment of intestinal ceramide observed after irradiation in both WT and ASMKO mice (Supplementary Fig. S2C–S2E). Finally, to precise ceramide secretion origin, total abdomen irradiation (TAI) was performed using a stereotactic device at 16.5 Gy, a dose able to induce a GI syndrome. In this condition, increase of ceramide was also observed in blood serum from locally irradiated WT mice (P < 0.5 versus nonirradiated mice; Fig. 1H) demonstrating, a least, the location of ceramide secretion.

Because endothelial cells are a high source of extracellular ASM, secretion of both ASM isoforms and ceramide was assessed in supernatants of irradiated HMVEC-L. A 30% increase in S-ASM and L-ASM activities was measured at 24 hours and maintained up to 72 hours post-15 Gy (Fig. 2A and B). Ceramide slightly increased 24 hours after irradiation. However, a 21% of ceramide enrichment was observed at 72 hours in the supernatant of irradiated endothelial cells (mean ± SD ceramide concentration at 72 hours for medium from nonirradiated vs. 15 Gy–irradiated HMVEC-L: respectively 9.66 ± 1.77 vs. 11.96 ± 2.65 μmol/L/10⁶ cells; Fig. 2C). The impact of those released factors by irradiated endothelial cells was assessed in the radiobiological response of intestinal epithelial cells. Treatment with either exogenous human placental ASM or natural C₁₆-ceramide for 24 hours efficiently repressed cell growth in a dose-dependent manner, respectively by 57.5% and 97% at the highest tested concentration (1U/mL ASM and 100 μmol/L ceramide as compared with vehicle-treated T84; Fig. 3A and B). Accordingly, treatment with either ASM or exogenous natural C₁₆-ceramide provoked T84 apoptosis in a dose-dependent manner (Fig. 3A and B). Of note, C₁₆-ceramide triggered more apoptosis than ASM (22-fold at 100 μmol/L vs. 3-fold at 1,000 mU/mL, respectively). Non-natural C₂-ceramide resumed C₁₆-ceramide effect on T84 epithelial dysfunctions. Indeed, this compound, but not the inactive C₂-dihydroceramide, significantly reduced cell number and induced apoptosis in T84, even at low concentration (Table 1).

Impact of secreted ASM and ceramide from irradiated endothelial cells on intestinal epithelium radiosensitivity was then studied in a noncontact coculture model (13). First, HMVEC-L model invalidated by siRNA was developed and validated by TLC analysis (Fig. 4A and B). Importantly, addition of the sole supernatant of irradiated HMVEC-L was sufficient to reduce nonirradiated T84 cell proliferation (Fig. 4C) and to induce their apoptosis (Fig. 4D). Indeed, the secretome of irradiated HMVEC-L triggered a 41% decrease in cell number and 1.9-fold increase in apoptosis in nonirradiated T84 cells (P < 0.05) after 24 hours of transwell coculture. ASM silencing by siRNA in irradiated HMVEC-L prevented the inhibition of T84 proliferation and limited their apoptosis (Fig. 4C and D). Only a 12% decrease in cell number and 1.6-fold increase in apoptosis were observed (P > 0.05). These effects were clearly dependent on ASM since reduced T84 proliferation and apoptosis were observed after 24 hours of coculture with irradiated HMVECs transfected with sham siRNA.

To strengthen the involvement of secreted ASM and ceramide from irradiated endothelial cells on intestinal epithelial cell radiosensitivity, HMVEC-L medium was collected after 24 and 72 hours in absence (CM) or after 15 Gy radiation (ASAP). Interestingly, a 30% decrease of T84 cell number was observed in T84 cells cultured in presence of ASAP (vs. CM collected at similar time: both P < 0.05; Fig. 4E). Furthermore, addition of desipramine (5 μmol/L), an ASM inhibitor, in ASAP restored epithelial T84 cell proliferation (vs. CM, P > 0.05; Fig. 4F). Similarly, addition of a blocking α-ceramide antibody (mAb) in ASAP completely abolished the ability of ASAP to reduce T84 proliferation while no effect was observed with its isotype (vs. CM72, P > 0.05; Fig. 4G). Finally, T84 cell apoptosis visualized by activation of the caspases 3 and 7, is induced by ASAP treatment (Supplementary Fig. S3). Pretreatment with desipramine or α-ceramide mAb blocked ASAP-induced apoptosis.

To definitely prove their role in epithelial crypt dysfunction, secreted ASM and ceramide effects have been directly tested on enteroids (a.k.a. intestinal organoids). On the basis of a previous method (24), intestinal crypts isolated from C57BL/6 mice were cultured in vitro for 7 days, to obtain 3D epithelial enteroids with budding structures (Fig. 5A, D, and G). Irradiation significantly reduced the number and size of enteroids by more than 20% and 30%, respectively (both P < 0.05; Fig 5A–C). Interestingly, whereas it did not alter survival of non-irradiated organoids, ASM addition in the media significantly increased enteroids radiosensitivity by 34%, as compared with the sham cultures (P < 0.05; Fig. 5B). The size of enteroids was also decreased by ASM disregarding irradiation (Fig. 5C). Using similar experimental modality, ceramide curiously reduced the number of organoids independently of radiation exposure without affecting their size (Fig. 5D–F). Finally, ASAP containing ASM and ceramide resumed results obtained directly with ASM or ceramide. Both enteroid number and size were significantly reduced after ASAP (respectively, P < 0.01 and P < 0.05 vs. CM), but were not affected when D609 was added in ASAP (Fig. 5G–I).

To determine whether ASM and ceramide affect the intestinal proliferating cells, the frequency of enteroid buds was analyzed (Fig. 6A). In correlation to its ability to limit organoid growth, ASM inhibited the formation of differentiated buds in nonirradiated and 5 Gy–irradiated enteroids (respectively, 37% and 17% decrease vs. sham-control; P < 0.05; Fig. 6B and C). As shown by Ki67 immunostaining, dividing cells are localized at the apex of the buds (Fig. 6D). When ASM treatment displayed a loss of Ki67⁺ proliferating cells and budding in nonirradiated or irradiated enteroids, ceramide affected none of those events (Fig. 6E–G).

Intracellular ceramide generation through relocation and activation of ASM mediated the apoptosis of endothelial cells exposed to ionizing radiation, leading to a deleterious GI syndrome (3, 10, 11). However, how endothelial cell response to ionizing radiation leads to intestinal epithelial radiosensitivity is not well understood. Recently, Bodo and colleagues proposed that endothelial apoptosis mediated by ASM/ceramide is a secondary lesion preceded by a rapid microvascular ceramide-mediated ischemia/reperfusion, leading to homology-directed misrepair and tumor cell death (26). This explanation cannot be excluded to explain the role of irradiated endothelial cells in GI syndrome. Because epithelial T84 cell proliferation arrest and cell death are also mediated by a paracrine secretion of irradiated endothelial cells (13), we aimed to determine whereas systemic ASM and ceramide acutely secreted by irradiated endothelial cells can directly trigger epithelial dysfunctions using in vitro, ex vivo, and in vivo models. In our study, we clearly demonstrate that endothelium radiosensitivity leads to a paracrine response inducing intestinal failure after radiation.

GI syndrome induces the death of C57Bl/6 WT mice within 6–7 days after whole body exposure to dose above 15 Gy (3). In those conditions, we show, for the first time, that enzymatic activities of the 2 ASM isoforms, especially S-ASM, are enhanced in the serum of WT mice, as early as 30 minutes and up to 3 days. A similar time response was observed for serum ceramide secretion. Importantly, ceramide release in the serum was also found after local irradiation of the abdomen (TAI), showing a local secretion. In ASMKO mice that do not develop acute intestinal failure after irradiation, no change in ceramide was measured highlighting ASM dependence of ceramide in the serum after irradiation. Furthermore, we reported that in vivo injection of D609, an ASM inhibitor, improved intestinal damage, albeit to a lesser extent than smpd1 invalidation in mouse, reinforcing the involvement of the ASM/ceramide pathway in mediating radiation-induced GI toxicity. In contrast, we were able to measure significant generation of ceramide in the intestinal mucosa 24 hours after irradiation, even in ASMKO mice, suggesting that ceramide generation in the mucosa might be induced by an alternative de novo metabolic pathway involving the ceramide synthase activity, independently of ASM expression (Supplementary Fig. S2). Moreover, we recently demonstrated that acute elevation of plasma ceramide is correlated with tumor control efficacy in a well-defined phase II study including 35 patients with lung and liver oligometastases from colon cancer treated by hypofractionated stereotactic body radiotherapy (SBRT) combined with irinotecan (21). Other study confirmed secreted ceramide as a biomarker of radiotherapy efficacy when increase of secreted ceramide was observed in the plasma of 5 out of 7 patients with bulky tumors responding to spatially fractionated grid radiotherapy (SFGRT; ref. 27). In this study, we corroborated the clinical relevance of increased circulating ASM activity and ceramide, providing new information on their role in radiation-induced intestinal toxicity.

Soluble factors released by irradiated HMVEC-L primary endothelial cells are critical for the induction of intestinal epithelial cell damage (13). We hypothesized that those soluble factors included ASM and ceramide. This hypothesis is supported by the fact that endothelium is an important source of S-ASM and its activity is enhanced by 2–3-fold in response to inflammatory cytokines in endothelial supernatant as well as in the serum of mice (14, 15). In this study, we reported that high dose of radiation induces an acute secretion of both ASM isoforms and ceramide by endothelial cells. Interestingly, phosphocholine, which is the other product of the hydrolysis of sphingomyelin by sphingomyelinase, is not secreted by irradiated endothelial cells (Supplementary Fig. S4A). Furthermore, in contrast to what we observed with ASM and ceramide, adding exogenous phosphocholine does not affect T84 dysfunction, showing the specific deleterious effect of ASM/ceramide (Supplementary Fig. S4B). As observed in the mouse serum, S-ASM is the most prominent ASM form in the supernatant of irradiated endothelial cells. Both ASM and ceramide may be secreted during the apoptosis of irradiated endothelial cells. However, S-ASM release by HUVECs is dependent on endothelial cell activation, but not cell death after TNFα treatment (15). Surviving HMVEC-L cells from radiation exposure develop an activation phenotype, defined by the release of proinflammatory cytokines, such as CXCL8, in the extracellular medium and in the mouse bloodstream (Supplementary Fig. S5; ref. 28). Further studies must define whether both ASM and ceramide are released by dying or activated endothelial cells.

Modulation of circulating ceramide and ASM is proposed as therapeutic target. Intravenous or intraperitoneal injection of neutralizing ceramide antibodies abolish platelet-activating factor–mediated pulmonary edema (16) or emphysema through the blockage of VEGF receptors (18). Amitriptyline, an ASM inhibitor, significantly reduces plasma ceramide level and decreases obesity, insulin resistance, and associated renal injury (20). Modulation of ASM or ceramide secreted by irradiated endothelial cells should also impact intestinal dysfunction. Exogenous treatment with either human recombinant ASM or natural C₁₆-ceramide on T84 human intestinal cells or enteroids induces a dose-dependent decrease of cell growth and increase of apoptosis (Figs. 3–6; Supplementary Fig. S2). Exogenous ASM enhances intracellular ceramide in T84 (Supplementary Fig. S6), which might be involved in epithelial radiosensitivity. The uptake of bioactive extracellular ceramide by endothelial cells triggers new synthesis of endogenous ceramide to activate proapoptotic signaling (29). Interestingly, intravenous treatment of 7 Gy–irradiated mice with recombinant human ASM reduced their survival (median survival 7 Gy: 11 days vs. 7 Gy + ASM: 4 days. P < 0.05; Supplementary Fig. S7) demonstrating that circulating ASM radiosensitizes mice to radiotherapy. Inhibition of secreted ASM/ceramide limits epithelial dysfunctions. Secretome of irradiated endothelial cells invalidated for smpd1 gene is not able to induce a paracrine response, leading to T84 cell-cycle inhibition or death (Fig. 4C and D). Then, pharmacologic inhibition of ASM and immunologic blockage of ceramide from ASAP of irradiated endothelial cells strongly inhibits epithelial cells radiosensitization (Fig. 4F and G). Interestingly, in vivo intravenous injection of an anticeramide mAb in mice prevents intestinal failure after high-dose radiotherapy (12). The authors are proposing that the antibody is inhibiting the formation of ceramide enriched raft platform leading to endothelial apoptosis. If we do not exclude this hypothesis, we have also considered that anticeramide mAb is blocking circulating ceramide in the bloodstream limiting epithelial dysfunction and intestinal failure.

Epithelial cell differentiation and functions are different depending on their intestinal location in the intestinal crypt–villus axis. Clonogens and their immediate progeny at the base of the crypts are capable of dividing to regenerate an entire functional intestinal epithelium (30). Up to 12 Gy, pool of surviving clonogens is still sufficient to support complete intestinal recovery. After 15 Gy, clonogens apoptosis is more widespread, resulting in an inability to restore the integrity of small intestine (31, 32). In our study, ASM and ceramide secretion was not only investigated in coculture cell models, but also in crypt-derived enteroids model that resemble normal intestinal physiology. In contrast to other studies using enteroids to screen radioprotectors, we used the developing enteroids to decipher the role of ASM and ceramide in crypt dysfunction (33, 34). In agreement with Wang and colleagues (33), dose of 5 Gy statistically reduces the number and the size of enteroids (Fig. 5A–C) when 15 Gy was inducing massive cell death in isolated crypts preventing enteroid formation. In line with our hypothesis, both exogenous human recombinant ASM and natural C₁₆-ceramide impact crypt survival and early-stage enteroid formation. However, whereas C₁₆-ceramide reduces the number of regenerating intestinal crypts without affecting their structure, ASM treatment is inducing a significant decrease of enteroid growth without affecting their number (Fig. 5). Pharmacologic inhibition of ASM activity, by pretreating ASAP from irradiated endothelial cells with D609, significantly improved both enteroid survival and formation.

The deleterious effect of ASM on intestinal crypt growth is supported by the significant reduction of Ki67⁺ cell number located on the apex of enteroid buds and inhibition of the crypt budding process (Fig. 6A–C; ref. 24), suggesting that ASM may directly affect the proliferating cell population. On the other hand, the decrease of enteroids number after C₁₆-ceramide, in combination or not with radiation, suggests its potential role in initiating crypt cell death (Fig. 5G). The dual responses of ASM and ceramide on enteroid generation were surprising because ASM hydrolyzes sphingomyelin to generate ceramide. This discrepancy may be explained by the targeted cell type, the kinetic or stress intensity. Thus, the differential effect of ASM and C₁₆-ceramide may be, respectively, associated with either a very rapid or late wave of sphingomyelin breakdown, further contributing to intestinal injury. Then, ceramide could contribute to acute apoptotic death observed in the first 24 hours following irradiation, while ASM could be involved in chronic cell growth arrest in later time points. The relevance of this hypothesis awaits further investigation.

This study provides a new understanding of the intercommunication between endothelium and epithelium compartments to induce radiation-induced GI syndrome. Acute secretion of ASM and ceramide by irradiated endothelial cells enhances epithelial apoptosis and cell-cycle arrest leading to crypt collapse, but also limits their regeneration. Thus, inhibition of secreted ASM and ceramide potentially represent an attractive strategy to limit the deleterious side effects of radiation-induced GI syndrome and improve the quality of life for patients and overall radiotherapy tolerance and efficacy.

No potential conflicts of interest were disclosed.

Conception and design: D. Leonetti, H. Estéphan, M.-H. Gaugler, M.M. Mahé, F. Paris

Development of methodology: D. Leonetti, H. Estéphan, N. Ripoche, N. Dubois, S. Gouard, I. Corre, M.-H. Gaugler, F. Paris

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Leonetti, H. Estéphan, N. Dubois, A. Aguesse, S. Gouard, L. Brossard, M.-H. Gaugler

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Leonetti, H. Estéphan, C. Pecqueur, M.-H. Gaugler, F. Paris

Writing, review, and/or revision of the manuscript: D. Leonetti, H. Estéphan, I. Corre, C. Pecqueur, M. Neunlist, E. Hadchity, M.-H. Gaugler, M.M. Mahé, F. Paris

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Leonetti, H. Estéphan, L. Brossard, S. Chiavassa

Study supervision: F. Paris

We would like to thank Dr. Richard Hellman for reviewing English scientific writing and Ms. Natacha Galopin and Ms. Laetitia Durand for their technical help. We would like to thank the gastrointestinal organoid platform (GalOP) from the INSERM UMR 1235 -TENS for its support in intestinal organoid generation. This work was partly supported by INCA, Electricité de France (EDF), la Ligue Nationale contre le Cancer, la Région Pays de la Loire, la fondation ARC, and the CNRS-L.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.