Transient Activation of the Hedgehog-Gli Pathway Rescues Radiotherapy-Induced Dry Mouth via Recovering Salivary Gland Resident Macrophages

Radiotherapy for head and neck cancers (HNC) frequently exposes nondiseased salivary glands (SG) to irradiation (IR), and consequently leads to irreversible salivary hypofunction (xerostomia; refs. 1, 2). IR-induced xerostomia exacerbates dental caries and periodontal disease and causes problems of taste, sleep, and speech, which severely impair the quality of life. Available treatments for IR-induced xerostomia, such as artificial saliva and saliva secretion stimulators, can only temporarily relieve these symptoms.

Saliva secretion occurs in acinar epithelial cells and is dependent upon a rich blood supply by microvessels surrounding acini. IR causes the irreversible impairment of acinar epithelial cells and microvascular endothelial cells in SGs through multiple mechanisms (3–5). Among the three pairs of major SGs, submandibular glands (SMG) produce two thirds of the total volume of the whole saliva in the unstimulated state and approximately 46% of that after stimulation (6), and are the focus of many research studies including ours.

We found recently that IR-impaired salivary function can be rescued by the transient activation of Hedgehog signaling within SMGs. Hedgehog signaling is triggered by the binding of Hedgehog ligand proteins such as Sonic hedgehog (Shh) to Patched receptors (Ptch1/2). In the canonical Hedgehog signaling cascade, Hedgehog-Ptch binding derepresses the membrane protein Smoothened, which in turn promotes an increase in the active form of Gli transcription factors (full length Gli1/2/3) and decrease in the repressive form of Gli (truncated Gli2/3). Consequently, the increased ratio of active to repressive form of Gli factors induces the expression of Hedgehog target genes such as Gli1 and Ptch1 (Supplementary Fig. S1; ref. 7). Our previous studies revealed that the activity of canonical Hedgehog signaling in adult mouse SMGs is very low in the healthy state, increases following the obstructive damage that induces functional SG regeneration, but is unaffected by IR (8). We transiently activated the Hedgehog pathway after IR by inducible expression or adenovirus-mediated intra-SMG transfer of Shh gene, and observed that the long-term salivary function was rescued in both mouse and miniature pig models (9–12). Transient Hedgehog activation ameliorated IR-damage of SG epithelial and endothelial cells, but the direct cellular and molecular targets of Hedgehog activation in SGs remained unclear.

Here, we report that the rescue of IR-impaired SG function by Hedgehog activation is through the recovery of SG-resident macrophages. In healthy adult mouse SGs, resident macrophages are abundant, derived from an embryonic origin but not from monocytes in circulation, and are homogeneously distributed throughout the SG (13, 14). Similar macrophage-like cells are also abundant in healthy human SGs (15). We found that mouse SMG macrophages are responsive to canonical Hedgehog signaling, and their quantities and production of proregenerative paracrine factors are decreased by IR but restored or increased by transient activation of Hedgehog-Gli pathway following IR. Consistently, the depletion of SMG macrophages compromised rescue effects of Hedgehog activation on IR-impaired salivary function. These data not only reveal the mechanism of Hedgehog-mediated rescue, but also illuminate a novel direction for developing effective treatment of IR-induced xerostomia.

C57BL/6J, Gli1-CreER^T2, and Ai9(RCL-tdT) mice were purchased from the Jackson Laboratory. All mouse procedures were approved by Texas A&M University (TAMU) Institutional Animal Care and Use Committee. Sample number per group is based on our previous papers on rescue effects of intra-SMG transfer of Shh gene (10, 11). Because Shh gene transfer rescued salivary function after IR with comparable efficiency in male and female mice (10), and >70% patients with HNC are men (16), we used male mice in this study. Eight- to 10-week-old mice weighted between 25 and 30 g were randomly grouped for treatments. The pilocarpine-stimulated whole saliva flow rates were measured as reported (9) by researchers blinded to treatments.

The IR parameters were determined through Helical Tomotherapy Planning, and IR dose was verified with a solid-water phantom as detailed in Supplementary Figs. S2–S5. For IR, mice were anesthetized with ketamine/xylazine mixture at 0.1 mg ketamine and 0.01 mg xylazine per gram of body weight. The SGs of anesthetized mice were exposed to 15 Gy of 6 MeV electron beam produced by a medical linear accelerator (Clinac 2100C, Varian Medical Systems) as reported (11).

Adenoviral vectors (Ad) encoding GFP, rat Shh, or Gli1 (Applied Biological Materials) were expanded, purified, and titered as reported (10). These vectors were delivered into SMGs of anesthetized mice by retroductal instillation as reported at a dose of 1 × 10⁹ particles per gland on day 3 after IR (10).

RNAs were extracted from SMGs with RNeasy Mini Kit (Qiagen) and reversely transcribed with High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The quantitative PCR (qPCR) was performed with SYBR Green Master Mix (Bio-Rad) on a 7900HT Fast Real-Time PCR System (Applied Biosystems). The primers were synthesized by Invitrogen with sequences retrieved from Primerbank (http://pga.mgh.harvard.edu/primerbank). The qPCR data were analyzed with Rps9 as the reference gene using qBasePlus software (Biogazelle).

SMGs were homogenized with Tissue Protein Extraction Reagent (Thermo Fisher, 40 μL per mg tissue) followed by centrifugation at 10,000 × g for 5 minutes to collect supernatant. For Western blot, proteins were fractionated by 12% SDS-PAGE, transferred to nitrocellulose membrane, and incubated with primary antibodies to Aquaporin 5 (Aqp5; Abcam ab78486, 1:500) or Gapdh (Sigma-Aldrich, AB2302, 1:1,000). Bound primary antibodies were detected with horseradish peroxidase–conjugated secondary antibodies, observed by enhanced chemiluminescence (Pierce), photographed VersaDoc imaging system, and quantified with QuantityOne software (Bio-Rad). The protein levels of Aif1, Cx3cr1, C1q, and Csf1 were examined with corresponding ELISA kits (MyBioSource MBS922753 and MBS3805156, Lifespan F8963, and Invitrogen EMCSF1) following the manufacturers' instructions.

RNA sequencing (RNA-seq) library preparation and sequencing were outsourced to NeoGenomics Labs. Briefly, the tissue samples underwent RNA isolation using Qiagen RNeasy Mini Kit, and RNA quality was confirmed with Agilent Bioanalyzer RNA Nano Kit with RIN >7. The Illumina TruSeq Stranded mRNA Sample Prep Kit instructions were followed to generate barcoded RNA-seq libraries. These libraries were pooled in equal molar ratio and processed using a standard Illumina HiSeq 2500 pipeline. The bioinformatics analyses were performed by researchers blinded to treatments with methods described in the Supplementary Methods.

Cells were isolated from mouse SMG as reported (14). Briefly, SMGs were isolated, minced, and digested for 1 hour with RPMI 1640 medium containing 1 mg/mL of collagenase IV, 5 mmol/L CaCl₂, 50 mg of DNase I, and 8% FBS with continuous shaking at room temperature. For examining macrophage subpopulations, cells were stained with FITC-labeled Ly6C antibody, phycoerythrin (PE)-labeled Adgre1 (F4/80) antibody, or corresponding isotype controls (BD Pharmingen, 553104, 565410, 553942, and 553930, 1:100). For sorting, cells were stained with FITC-labeled antibodies for Pecam1/CD31 or Epcam (Biolegend 102405 and 118207, 1:100), PE-labeled F4/80 antibody mentioned above, or corresponding isotype controls. Dead cells were excluded with LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Invitrogen).

Single-cell suspensions from mouse SMGs were generated as mentioned above. Single-cell RNA-seq (scRNA-seq) libraries were generated in TAMU Genomics Core using the 10x Chromium platform and the version 3 Chromium Single Cell 3′ reagent kit following the manufacturer's instructions (10x Genomics). Samples were pooled in equimolar concentrations and sequenced on a single lane of an S4 2 × 150 PE flow cell for an Illumina NovaSeq 6000. Note that 11,048 and 18,333 cells from GFP and Shh groups were sequenced to a depth of 43,924 mean reads per cell and 1,420 median genes per cell. The bioinformatics analyses were performed by researchers blinded to treatments with methods described in the Supplementary Methods. The Uniform Approximation and Projection method was used to visualize the clusters. Putative T cells and B cells are abnormally abundant in GFP sample only likely due to blood contamination and are excluded from further analyses. Normalized data are shown in the form of feature plots.

Eight- to 10-week-old Gli1-CreER^T2;Ai9(RCL-tdT) mice were subjected to 15 Gy IR at the SG region or mock treatment. At 3 days after IR, these mice were transduced with Ad-GFP or Ad-Shh as mentioned above, and then injected with 0.1 mg 4-hydroxytamoxifen (4-OHT, Sigma) daily for 3 days. Mice were euthanized to collect SMGs 7 or 42 days after 4-OHT induction. Frozen SMG sections were first incubated in PBS or a 1:100 dilution of following primary antibodies: Aif1/Iba1 (Abcam ab48004), Krt5 (BioLegend 905901), Cd31 (BD Pharmingen 558736), Krt8 (Novus Biologicals NBP216094), or Aqp5 (Abcam ab78486) at 4°C overnight. Sections were then washed and incubated with appropriate secondary antibodies labeled with Cy7 (Vector Laboratories), and counterstained with DAPI. Sections were imaged using a fluorescent microscopy (Nikon, Eclipse 80i) and were analyzed with NIS Elements software.

Based on published protocols (17–19), clodronate liposomes (CL, 50 μg/mL, FormuMax) and control liposomes were i.v. injected (0.2 mL) into mice on days 2 and 4 after IR or mock treatment, and directly injected into SMGs (0.03 mL per gland at 3 sites) on day 4. SMGs were collected 3 days after last CL injection from some mice for flow cytometry assay. Other mice were transduced with Ad-Shh as mentioned above.

No animals or samples were excluded from the final analysis, and samples were not pooled for analyses. The numbers of experimental replicates are 3 for RNA-seq and 5 for all other assays, and technical replicates are 3 for qRT-PCR and ELISA and 2 for flow cytometry and Western blot. All statistical analysis and graphical generation of data were performed with GraphPad Prism software. All quantified data were analyzed with nonparametric Kruskal–Wallis test followed by the Dunn post test.

Besides the canonical signaling cascade mediated by Gli transcription factors, Hedgehog ligands are capable of triggering several Gli-independent noncanonical signaling cascades (20). To determine whether transient activation of Gli-dependent canonical Hedgehog signaling is able to rescue IR-impaired salivary function, we performed 15 Gy local IR to the neck region containing all major SGs in 8- to 10-week-old C57BL/6 mice. On day 3 after IR, we delivered adenoviral vectors (Ad) carrying cDNA of rat Gli1, rat Shh, or GFP into SMGs by retroductal instillation as reported (10). In mice, it is technically challenging to measure resting saliva flow rate or gland-specific stimulated saliva flow rate; therefore, we measured the stimulated whole saliva flow rates as a key SMG function readout at 28, 60, or 90 days after IR, and collected SMG samples at 7, 14, 28, or 90 days after IR (Fig. 1A). In SMGs collected 7 days after IR, the expression of Hedgehog target genes, mouse Gli1 and Ptch1 (7), was not significantly affected by IR or IR plus Ad-GFP treatment, but significantly increased in IR plus Ad-Gli1 or Ad-Shh groups compared with IR or nontreated (NT) groups (Fig. 1B and C), which confirmed the effective activation of Hedgehog-Gli pathway by Ad-Gli1. The higher levels of Hedgehog target genes in Gli1 group compared with Shh group at day 7 were likely due to the additional Hedgehog activation in SMG cells not expressing functional Hedgehog receptors. In SMGs collected 14 or 28 days after IR, the expression of these two Hedgehog target genes was not significantly affected by any treatment, indicating that Ad-Gli1 delivery activated Hedgehog signaling in irradiated SMGs transiently as Ad-Shh did (10). Local IR progressively decreased the stimulated whole saliva flow rates from day 28 to day 90 as expected, and significantly reduced the expression of acinar marker Aqp5 in SMGs by 90 days after IR at both mRNA and protein levels (Fig. 1D,–G). Both salivary function indexes were significantly rescued by the intra-SMG delivery of Ad-Shh or Ad-Gli1 with comparable efficacies (Fig. 1D,–G). Because all three pairs of major SGs were exposed to IR and only SMGs were treated, considering the approximately 46% contribution of stimulated saliva from SMGs (6), the approximately 40% stimulated whole saliva flow rates indicate an effective rescue of SMG function in Shh and Gli1 groups by day 90 (Fig. 1D). These data confirm that transient activation of canonical Hedgehog signaling is sufficient to rescue IR-impaired salivary function.

To identify cells that directly respond to canonical Hedgehog signaling in SMGs, we used Gli1-CreER^T2;Ai9(RCL-tdT) mice that carry Cre-inducible tdTomato (tdT) gene in the Rosa26 locus. After Cre activation with 4-OHT, Gli1-expressing Hedgehog-responsive cells and their progeny will be labeled with tdT⁺ in these mice (21). At 3 days before 4-OHT treatment, some mice were firstly treated with 15 Gy IR at the head and neck region. On the same day before 4-OHT treatment, some irradiated mice received intra-SMG transfer of Ad-GFP or Ad-Shh as mentioned above. 4-OHT was administrated i.p. and daily from days 0 to 2, and SMGs were collected in 7 and 42 days for section (Fig. 2A). In both day 7 and 42 SMGs, compared with NT samples, tdT⁺ cells were slightly decreased by IR or IR+Ad-GFP, but significantly increased by IR+Ad-Shh (Fig. 2B,–E). In all day 7 SMGs, tdT⁺ cells were rare and sparsely distributed in the stroma, basal epithelia, and vessel-like structures (Fig. 2B). Compared with day 7 samples, in day 42 SMGs with the same treatment, tdT⁺ cells greatly increase, become more concentrated, and form ring-shaped structures or cord-like structures adjacent to large ducts, but are still relatively rare (Fig. 2B,–E). At all samples at both time points, immunofluorescent (IF) staining indicated tdT⁺ cells express macrophage marker Aif1/Iba1, basal/myoepithelial marker Keratin-5 (Krt5), or endothelial marker Cd31/Pecam1 (Fig. 2F,–K). These tdT⁺ cells are rarely positive for duct marker Krt8 and negative for acinar marker Aqp5 at both days 7 and 42, but form ring-shaped structures surrounding some Krt8⁺ ducts or Aqp5⁺ acini at day 42 (Supplementary Fig. S6A–S6D). These data indicate that cord-like and ring-shaped structures are microvessels and interconnected macrophages (14) or myoepithelial cells. Collectively, our lineage-tracing data suggest that Gli1⁺ Hedgehog–responsive cells in SMGs comprise small subset of macrophages, basal/myoepithelial cells, and endothelial cells, and these cells expand to a limited extent during recovery from IR damage even after transient Hedgehog activation.

To find clues for underlying mechanisms of Hedgehog-mediated rescue of IR-impaired salivary function, we collected SMGs from all 5 treatment groups mentioned in Fig. 1 at 7 days after IR for high-throughput RNA-seq. The RNA-seq dataset has been deposited to Gene Expression Omnibus with the record number GSE133482. The differentially expressed genes in each group versus NT or IR only group respectively were identified (Supplementary File S2) and subjected to Ingenuity Pathway Analysis (IPA). The p53 signaling activity is elevated by DNA damages following IR for weeks (22) and not affected by intra-SMG Shh gene transfer (11). As indicated by the analysis of canonical pathways, p53 signaling is activated in all groups treated with IR to comparable extents (Fig. 3A), which is confirmed by qRT-PCR assays on p53 target genes Cdkn1a/p21, Mdm2, and Trp53inp1 (Supplementary Fig. S7A–S7C), indicating that IR injury levels are comparable between different treatment groups. Notably, IR inhibited multiple macrophage-related pathways, including pattern recognition, phagocytosis, GMCSF (Csf2) signaling essential for the maintenance of tissue-resident macrophages (23), and HGF signaling related to prorepair effects of macrophages (Fig. 3A; ref. 24). These IR-inhibited pathways were restored or activated by intra-SMG gene transfer of Shh or Gli1 but not by GFP (Fig. 3A). As indicated by the analysis of biological functions, IR inhibited angiogenesis as reported (3), and reduced quantity and proliferation of mononuclear leukocytes, phagocytosis of myeloid cells, and activation of phagocytes, whereas Shh or Gli1 but not GFP gene transfer after IR restored or enhanced these functions compared with IR treatment alone (Fig. 3B).

To validate these RNA-seq data, we performed flow cytometry analysis of macrophage markers in SMGs of adult C57BL/6 mice collected 3 or 7 days after IR with or without intra-SMG transfer of GFP or Shh gene. F4/80⁺ macrophages are abundant in NT SMGs, and most of them (>85%) are resident macrophages (Φ) as indicated by the low expression of Ly6c (Fig. 3C and D; ref. 25). IR significantly decreased numbers of F4/80⁺/Ly6c^low resident macrophages but not F4/80⁺/Ly6c^high infiltrating macrophages (Φ) by 3 and 7 days, whereas Shh but not GFP gene delivery restored this index (Fig. 3C,–E). The percentages of F4/80⁺/Ly6c^high infiltrating macrophages increased in Shh and GFP groups compared with NT or IR SMGs, but were still much lower than those of F4/80⁺/Ly6c^low resident macrophages (Fig. 3C,–E), suggesting that their increases are likely caused by the delivery of adenoviral vectors and less essential for salivary function than resident macrophages. Quantitative RT-PCR data confirmed that IR significantly downregulated the mRNA levels of macrophage markers (Cd11c, F4/80, Aif1/Iba1, and Fcgr1), phagocytosis receptors (Fcgr1 and Axl), and tissue-resident macrophage marker Cx3cr1 (26) by 3 and 7 days, whereas transfer of Shh or Gli1 but not GFP gene after IR restored or increased them on day 7 (Fig. 3F). Consistent changes of Aif1 and Cx3cr1 proteins were validated with ELISA (Fig. 3G and H). These data demonstrate that IR severely impaired SG-resident macrophages, whereas Hedgehog activation following IR restored them.

To determine whether SG macrophages are required for Hedgehog-mediated rescue of salivary function after IR, we depleted SMG macrophages with CL as reported (17–19) 1 day after intra-SMG transfer of Shh or 4 days after IR (Fig. 4A). In SMGs collected 3 days after CL injection alone or with IR or IR+Shh, flow cytometry indicated that CL efficiently depleted both Ly6c^low and Ly6c^high F4/80⁺ macrophages (resident and infiltrating macrophages, >45% decrease) in all three groups (Fig. 4B,–D). The stimulated saliva flow rate were measured at 90 days after IR or corresponding mock treatment (NT), and SMGs were collected at day 90 to examine the level of acinar marker Aqp5. At 90 days after treatment, mice subjected to CL injection alone maintained their saliva production and the mRNA and protein expression of Aqp5 in SMGs at levels comparable with NT group, whereas mice subjected to IR plus CL injection showed decreases of both indexes comparable with IR group (Fig. 4E–H). Notably, in mice subjected to Shh gene transfer plus CL injection after IR, these two indexes are comparable with IR group and significantly lower than the IR+Shh group (Fig. 4E,–H). These data indicate that SMG macrophages are required for the rescue of salivary function by transient Hedgehog activation following IR.

To determine how Hedgehog-Gli signaling regulates SG function through macrophages, we performed scRNA-seq of adult mouse SMGs harvested 7 days after intra-SMG delivery of Ad-Shh or Ad-GFP. The identities of cell clusters were determined by the expression of marker genes as reported (Fig. 5A; Supplementary Table S1; refs. 27, 28). The cells responsive to Hedgehog-Gli signaling were identified by expression of Ptch1, the Hedgehog receptor and target gene. In the GFP group, Ptch1⁺ cells are mostly enriched in putative resident macrophages (rMp, Cx3cr1⁺Ccr2⁻, Supplementary Fig. S8A; ref. 26) and are moderately enriched in innate lymphoid cells (ILC; ref. 29), ductal cells, and endothelia cells (Fig. 5A and B). In the Shh group, Ptch1⁺ cells strikingly increased in endothelial cells, acinar/ductal progenitors (ADP, Aqp5⁺/Krt8⁺/Epcam^high), and ductal progenitors (DP, Krt8⁺/Epcam^high and Ascl3⁺ or Sox9⁺; ref. 28), and moderately increased in ILCs, basal/myoepithelial cells, and mesenchymal stem cells (MSC; Fig. 5A and B). Meanwhile, Hedgehog activation strikingly increased percentages of ADP, DP, and endothelial cells in all SMG cells, and moderately increased those of basal/myoepithelial cells, MSCs, and rMp, but did not increase that of ILCs (Fig. 5C). However, Hedgehog activation did not significantly increase the levels of proliferation markers Pcna and Mki67 in all Ptch1⁺ cells or Ptch1⁺ endothelial cells (P > 0.05), and only a small percentage of Mki67⁺ cells are Ptch1⁺ (Supplementary Fig. S8B). These data suggest that Hedgehog activation likely expanded these cells through indirect mechanisms.

Macrophages or monocytes were reported to promote proliferation of cultured intestine epithelial cells and endothelial cells through HGF-Met signaling (24, 30). In mouse SMGs, HGF-Met signaling is inhibited by IR but upregulated by Hedgehog activation after IR (Fig. 3A). Consistently, scRNA-seq indicated that Hgf⁺ cells are enriched in putative resident macrophages with or without Hedgehog activation, and increased greatly by Hedgehog activation (Fig. 5D and E). Without Hedgehog activation, Met/Hgfr⁺ cells are rarely present in putative endothelial cells, ductal cells, ADP, and DP; after Hedgehog activation, Met/Hgfr⁺ cells greatly increased in ADP, DP, and endothelial cells (Fig. 5D and E). Another potential prorepair factor produced by macrophages is C1q that promote angiogenesis through C1q receptor Cd93/C1qRP on endothelial cells (31, 32). With or without Hedgehog activation, cells express all C1q genes (C1qa, C1qb, and C1qc) at high levels (>50th percentile) are mainly putative SMG-resident macrophages, whereas Cd93 expression is enriched in putative endothelial cells without Hedgehog activation and increased strikingly in these cells by Hedgehog activation (Fig. 5D and E). The maintenance of tissue-resident macrophages is regulated by Csf1/IL34-Csf1r and Csf2-Csf2r signaling (23). In SMGs with or without Hedgehog activation, Csf1r^high (>50th percentile) cells are mainly putative resident macrophages (Fig. 5D), and majority of resident macrophages (>70%) expressing only Csf1r but not Csf2r (Supplementary Fig. S8C). In SMGs without Hedgehog activation, Csf1⁺ cells are much more abundant than IL34⁺ cells and Csf2⁺ cells, and Hedgehog activation greatly increased the percentages of Csf1⁺ and IL34⁺ cells in all SMG cells, but did not significantly affect that of Csf2⁺ cells (Fig. 5D and E; Supplementary Fig. S8D). After Hedgehog activation, both Csf1⁺ cells and IL34⁺ cells are enriched in the putative ADP, DP, and basal/myoepithelial cells, whereas Csf1⁺ cells are also enriched in the putative endothelial cells (Fig. 5D and E). The increases of above genes by Hedgehog activation are found in both Ptch1⁺ and Ptch1⁻ cells (Fig. 5F), suggesting that Hedgehog activation induces these genes both directly and indirectly.

To validate the cell type–specific expression of these factors, we sorted F4/80⁺ macrophages, Cd31⁺ endothelial cells, and Epcam^high putative epithelial progenitor cells by flow cytometry from SMGs at 7 days after Ad-GFP or Ad-Shh treatment, and examined the expression of these factors with qRT-PCR assays. Compared with cells negative or low for the corresponding marker with or without Hedgehog activation, F4/80⁺ macrophages express significantly higher levels of C1qa, C1qb, C1qc, Csf1r, and Hgf (Supplementary Fig. S9A), Epcam^high cells express significantly higher levels of IL34 and Met/Hgfr (Supplementary Fig. S9B), and Cd31⁺ endothelial cells express significantly higher levels of Csf1 and Cd93/C1qRP in both groups and higher level of Met/Hgfr in Shh group alone (Supplementary Fig. S9C). Notably, Ad-Shh treatment significantly increased the relative levels of C1q factors, Csf1r and Hgf in macrophages compared with Ad-GFP treatment, suggesting the activation of macrophages after Hedgehog activation. These data together suggest that Hedgehog activation triggers positive feedback loops between epithelial progenitors, endothelial cells, and resident macrophages through Csf1/IL34, Hgf, and C1q signaling pathways.

To determine effects of IR and transient Hedgehog activation following IR on above paracrine interactions, we examined the mRNA expression of these genes and the tissue-resident macrophage marker Cx3cr1 by qRT-PCR in SMGs at days 3, 7, 14, and 28 after IR with or without intra-SMG Ad-Shh treatment on day 3. Consistent to data in Fig. 3, the mRNA expression of macrophage-enriched Csf1r, Cx3cr1, C1qa, and Hgf was significantly decreased by IR compared with the NT group, and was significantly increased by Shh but not GFP gene transfer compared with the IR group at days 7 and 14 (Fig. 6A,–D). At day 28, the mRNA expression of Cx3cr1, C1qa, and Hgf was still significantly downregulated by IR and recovered by Shh but not GFP gene transfer. The mRNA expression of Csf1r ligands (Csf1 and IL34), proangiogenic C1q receptor CD93, and Hgf receptor Met was significantly decreased by IR at day 14 or 28 compared with the NT group, and was significantly increased by Shh but not GFP gene transfer compared with the IR group at days 7, 14, and 28 except the slight increase of Met at day 7 (Fig. 6E,–H). The consistent changes of C1q and Csf1 proteins were confirmed by ELISA at days 3 and 7 after IR (Fig. 6I and J). These data indicate that consequent to acute IR damages to SMG cells, IR leads to persistent defects of paracrine interactions between resident macrophages, epithelial progenitors, and endothelial cells; conversely, transient Hedgehog activation after IR enhances these paracrine interactions rapidly and for a prolonged period (Fig. 6K). Because transient Hedgehog activation rescues salivary function after IR through preserving epithelial stem/progenitor cells and promoting angiogenesis (9–12), whereas SG macrophages are required for such a rescue (Fig. 4), these data suggest that the macrophage-derived proangiogenic and prorepair factors, C1q and Hgf, likely play essential roles in the rescue of salivary function by Hedgehog activation.

Current research on molecular therapies of the IR-induced salivary hypofunction mainly focuses on recovering the impaired acinar epithelial cells or microvascular endothelial cells (33). Resident macrophages with an embryonic origin are essential for functional repair of multiple organs (34, 35) and abundant in SGs (13–15), but their roles in IR damages and function restoration of SGs have never been explored. We report here that IR greatly impairs SG-resident macrophages and decreases expression of macrophage-derived proregenerative factors Hgf and C1q and their receptors. We also found that Hedgehog activation after IR recovers SG-resident macrophages, either directly or through inducing expression of Csf1 and IL34 in epithelial progenitors and endothelial cell, which in turn restore the expression of macrophage-derived Hgf and C1q. Conversely, the depletion of SG macrophages abolished rescue effects of transient Hedgehog activation on salivary function following IR. These data indicate that SG-resident macrophages are essential for the functional SG regeneration following IR, which is related to their paracrine interactions with epithelial progenitors and endothelial cells.

Besides these paracrine interactions, SG-resident macrophages likely also contribute to the restoration of salivary function after IR through proregenerative innate immune responses. These responses include the regulation of inflammation caused by IR and the removal of cell debris and dead or senescent cells, and are essential for functional regeneration of several other organs (36, 37). Notably, C1q also enhances phagocytosis by tagging damage-associated molecular patterns and apoptotic cells (38), and promotes the polarization of macrophages to the M2 phenotype and the production of anti-inflammatory cytokines (39). Therefore, the recovery of resident macrophages and C1q expression in irradiated SMGs by transient Hedgehog activation might restore salivary function through both direct prorepair effects and indirect innate immune responses. The contributions of the latter to the restoration of salivary function are worth further study.

We reported previously that transient activation of Hedgehog pathway following IR rescues salivary function through preservation of salivary epithelial stem/progenitor cells, endothelial cells, and parasympathetic innervation mainly via non-cell-autonomous mechanisms such as inhibiting cellular senescence and inducing prorepair paracrine factors (9–11), but SG cells directly responsive to Hedgehog activation are unclear. As indicated by findings reported here, Hedgehog-responsive cells are much more abundant in SG-resident macrophages than in all other types of SG cells including epithelial and endothelial cells (Fig. 5A and B), and the preservation of salivary epithelial stem/progenitor cells and endothelial cells appears largely mediated through the recovery of SG-resident macrophages and their prorepair paracrine factors (Figs. 4,–6). Meanwhile, as reported in other tissues, the recovery of resident macrophages in SGs likely also contributes to the preservation of parasympathetic innervation and the inhibition of cellular senescence by transient Hedgehog activation. The underlying mechanisms include clearing axonal debris (36) and senescent cells (37), promoting neurite outgrowth (40), enhancing DNA repair through Hgf (41), and resolving inflammation that propagates DNA damage and senescence (42).

Present approaches of preserving SG acinar epithelial cells or endothelial cells after IR rely on antiapoptosis factors, growth factors, and angiogenic factors. These factors generally have protumor risks when their levels increase in blood or tumor (33), which limits their application in HNC survivors. Our findings indicate that SG-resident macrophages and their paracrine factors are essential for the recovery of IR-impaired salivary function. These cells and factors could be manipulated safely in patients with cancer. For instances, Csf2/GMCSF that regulates the survival and proliferation of resident macrophages (23) has been approved by FDA for applications in patients with cancer (43); while C1q, a major prorepair factor secreted by SG-resident macrophages, also appears safe for application in patients with cancer (44). Therefore, it will be worth exploring whether recovering SG-resident macrophages and/or their paracrine factors is sufficient to restore salivary function following radiotherapy.

In summary, our studies demonstrate that SG-resident macrophages are essential for the Hedgehog-mediated restoration of the IR-impaired salivary function, which is related to their paracrine interactions with endothelial and epithelial cells. These findings shed light on a new research direction for the development of safer and more effective treatments of radiotherapy-induced salivary hypofunction in HNC survivors.

Q. Zhao reports grants from NIH (1R01DE022975) during the conduct of the study. F. Liu reports grants from NIH (1R01DE022975) during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

Q. Zhao: Data curation, investigation. L. Zhang: Formal analysis, visualization, methodology. B. Hai: Data curation, investigation. J. Wang: Resources, methodology. C. Baetge: Resources, methodology. M.A. Deveau: Resources, supervision, methodology. G.M. Kapler: Supervision, funding acquisition, writing-review and editing. J.Q. Feng: Resources, supervision, funding acquisition, methodology, writing-review and editing. F. Liu: Conceptualization, formal analysis, supervision, funding acquisition, visualization, writing-original draft, project administration.

This study was supported by NIH/NIDCR 1R01DE022975-01 to Q. Zhao, B. Hai, and F. Liu and TAMHSC bridge fund to F. Liu. We sincerely thank Drs. Andrew Hillhouse and Kranti Konganti in TAMU Genomics Core for scRNA-seq.

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.