A kidney-specific genetic control module in mice governs endocrine regulation of the cytochrome P450 gene Cyp27b1 essential for vitamin D3 activation

The vitamin D endocrine system regulates mineral homeostasis through its activities in the intestine, kidney, and bone. Terminal activation of vitamin D3 to its hormonal form, 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3), occurs in the kidney via the cytochrome P450 enzyme CYP27B1. Despite its importance in vitamin D metabolism, the molecular mechanisms underlying the regulation of the gene for this enzyme, Cyp27b1, are unknown. Here, we identified a kidney-specific control module governed by a renal cell-specific chromatin structure located distal to Cyp27b1 that mediates unique basal and parathyroid hormone (PTH)-, fibroblast growth factor 23 (FGF23)-, and 1,25(OH)2D3-mediated regulation of Cyp27b1 expression. Selective genomic deletion of key components within this module in mice resulted in loss of either PTH induction or FGF23 and 1,25(OH)2D3 suppression of Cyp27b1 gene expression; the former loss caused a debilitating skeletal phenotype, whereas the latter conferred a quasi-normal bone mineral phenotype through compensatory homeostatic mechanisms involving Cyp24a1. We found that Cyp27b1 is also expressed at low levels in non-renal cells, in which transcription was modulated exclusively by inflammatory factors via a process that was unaffected by deletion of the kidney-specific module. These results reveal that differential regulation of Cyp27b1 expression represents a mechanism whereby 1,25(OH)2D3 can fulfill separate functional roles, first in the kidney to control mineral homeostasis and second in extra-renal cells to regulate target genes linked to specific biological responses. Furthermore, we conclude that these mouse models open new avenues for the study of vitamin D metabolism and its involvement in therapeutic strategies for human health and disease.

The vitamin D endocrine system serves to regulate mineral homeostasis through its actions in the intestine, kidney, and bone (1). Vitamin D 3 itself is sequentially activated via two specific chemical modifications that occur first in the liver by CYP2R1 to 25(OH)D 3 2 and then in the kidney by CYP27B1 to 1␣,25-dihydroxyvitamin D 3 (1,25(OH) 2 D 3 ), the active hormonal form of the vitamin (2). Blood levels of 1,25(OH) 2 D 3 are also determined by rates of catabolism that represent the primary function of CYP24A1 in the kidney (3). CYP27B1 is also expressed at low levels in many non-renal target cells (NRTCs), particularly those of skin and the immune system, where local production of 1,25(OH) 2 D 3 has been suggested to preferentially influence the many pleiotropic, non-calcemic functions of 1,25(OH) 2 D 3 (4,5). Neither the mechanisms nor the overall biological impact of these non-renal cellular sources of 1,25(OH) 2 D 3 are clear, particularly in the context of circulating 1,25(OH) 2 D 3 , which is believed to be derived exclusively from the kidney. Despite these uncertainties, the regulated expression of Cyp27b1 and Cyp24a1 is well-recognized as central to the overall biological activity of the vitamin D endocrine system.
Cyp27b1 expression is known to be controlled in the kidney by many different factors. PTH represents the primary inducer (6,7), whereas both 1,25(OH) 2 D 3 (8) and FGF23 (9,10) represent the major negative regulators of Cyp27b1 expression. Indeed, the ability of PTH, FGF23, and 1,25(OH) 2 D 3 to modulate the expression of Cyp27b1 in the kidney comprises a key element of regulatory signaling that links adaptive vitamin D metabolism to the maintenance of mineral homeostasis. Interestingly, renal expression of Cyp24a1 is reciprocally regulated by these same hormones; PTH suppresses and both 1,25(OH) 2 D 3 and FGF23 induce Cyp24a1 (11)(12)(13)(14). This regulatory paradigm in the kidney highlights the critical importance of the coordinated control of both Cyp27b1 and Cyp24a1 expression by these endocrine hormones not only in the maintenance of circulating 1,25(OH) 2 D 3 , but in the orchestration of normal mineral homeostasis as well.
Despite the central importance of these two genes in vitamin D metabolism, little is known of the molecular mechanisms that underpin Cyp27b1 regulation in the kidney by PTH, FGF23, or 1,25(OH) 2 D 3 at the genomic level or of the adaptive relationship that occurs between this modulation and that of Cyp24a1 expression. It has been shown that PTH action at Cyp27b1 involves the PKA signaling pathway and likely the CREB transcription factor (15) and that 1,25(OH) 2 D 3 action involves the nuclear vitamin D receptor (VDR) (16). This contrasts with the actions of the more recently discovered phosphaturic hormone FGF23, where limited insight has been gained with regard to the identities of both the FGF receptor isoform(s) and the transcription factor(s) that are involved in Cyp27b1 expression (17,18). Importantly, research over the past decade using unbiased genomic techniques now points to the likelihood that genes such as Cyp27b1 may be regulated via genomic control regions located distal to their transcriptional start sites, a principle that we have found relevant to the vitamin D system (19,20). The absence of fundamental mechanistic insight relative to Cyp27b1 regulation has impeded progress in fully understanding the vitamin D metabolic system, its role in orchestrating mineral homeostasis, and the pathological impact of alterations in this system that can occur in a wide variety of human maladies, including autoimmune diseases and cancer.
Herein, we focused our efforts on understanding the expression and regulation of Cyp27b1 in the mouse. Our results identify a complex, multicomponent endocrine regulatory module specific to the kidney that is governed by a chromatin structure that is absent in NRTCs. This module regulates both the basal expression of Cyp27b1 in the kidney as well as its differential modulation by PTH, FGF23, and 1,25(OH) 2 D 3 , but it does not control the expression of Cyp27b1 in NRTCs by inflammatory agents such as LPS. These results highlight the differential regulation of Cyp27b1 in vivo and provide a starting point for further delineation of the precise molecular and genomic mechanisms through which both renal and non-renal tissues control the expression of Cyp27b1, production of 1,25(OH) 2 D 3 , and therefore its local and endocrine biology.

VDR and pCREB are bound upstream of the Cyp27b1 gene in a kidney-specific manner
Dose-and time-optimized i.p. injections for an acute maximal gene expression response by PTH, FGF23, and 1,25(OH) 2 D 3 into wild-type C57BL/6 (WT) mice lead to a dramatic up-regulation of Cyp27b1 transcripts in the kidney in response to PTH and a significant suppression in response to both FGF23 and 1,25(OH) 2 D 3 (Fig. 1A). Cyp24a1 expression in the kidney, in contrast, is reciprocally suppressed by PTH and induced by both FGF23 and 1,25(OH) 2 D 3 (Fig. 1B). Based upon this regulatory paradigm and the absence of insight into transcriptional activities, we utilized ChIP-seq analysis of the mouse kidney to identify the sites of action of both PTH and 1,25(OH) 2 D 3 at the Cyp27b1 gene locus (the FGF23-activated transcription factor(s) has yet to be identified). ChIP-seq analysis of pCREB and VDR in kidneys of WT mice 1 h after a single injection (a time point previously optimized for primary genomic effects (21)) of either vehicle or cognate hormone revealed the presence of pCREB and VDR bound to DNA at the Cyp27b1 locus (Fig. 1, C and D). The former was pre-bound to DNA and generally unaffected by PTH, whereas DNA binding of the latter was largely reinforced upon 1,25(OH) 2 D 3 injection at a single overlapping site within an intron of the upstream Mettl1 gene and at three sites spanning the large intron located and Cyp24a1 (B) after treatment in 8 -9-week-old C57BL/6 wildtype mice with ethanol/PBS vehicle (Veh, gray, n ϭ 6), 230 ng/g bw PTH for 1 h (PTH, blue, n ϭ 6), 10 ng/g bw 1,25(OH) 2 D 3 (1,25D 3 , black, n ϭ 6) for 6 h, or 50 ng/g bw FGF23 for 3 h (FGF23, green, n ϭ 6). Data are displayed as relative quantitation (RQ, mean Ϯ S.E.) compared with Gapdh. *, p Ͻ 0.05 paired t test, treatment versus vehicle. Overlaid ChIP-seq data tracks for pCREB (C), VDR (D), H3K4me1 (E), and H3K9ac (F) from mouse kidney displayed as either basal or vehicle (yellow, n ϭ 3) or treated (blue, n ϭ 3) with PTH, 1,25(OH) 2 D 3 , or FGF23 as indicated with concentrations used in A and with a treatment time of 1 h. G, H3K9ac data in Cyp27b1KO mice (C27KO, blue) or VDRKO (blue) versus wildtype (WT, yellow). Overlapping track data appear as green. Regions of interest are highlighted in light gray boxes. Genomic location and scale are indicated (top), and maximum height of tag sequence density for each data track is indicated on the y axis (top left each track, normalized to input and 10 7 tags). Gene transcriptional direction is indicated by an arrow and exons by boxes.

Endocrine regulation of Cyp27b1
in the Mettl21b gene ( Fig. 1, C and D, highlighted by gray vertical boxes). Both genes produce methyltransferase-like proteins that are poorly characterized (22,23). Neither pCREB nor VDR was observed immediately upstream of the Cyp27b1 gene's transcriptional start site or in the introns of Mettl1 or Mettl21b in marrow-derived mesenchymal stem cells (MSCs) (Fig. 1, C and D, gray track) following identical hormonal treatment. We also performed a ChIP-seq analysis of the transcription-confirming histone mark H3K36me3 1 h after injection ( Fig. 2A). Modulation of H3K36me3 showed that each of these hormones, including FGF23, mediated direct regulation of the Cyp27b1 gene body in the kidney. Equally important, these experimental results identified potential sites of regulatory action for at least two of the three primary mineral-regulating hormones at the Cyp27b1 gene.

VDR and pCREB are localized to active enhancers in the introns of Mettl1 and Mettl21b
To determine whether the binding sites were localized to active enhancer regions (24), ChIP-seq analyses of the histone enhancer signature marks H3K4me1 and H3K9ac were performed in kidneys following a 1-h treatment of either vehicle or each of the three hormones (Fig. 1, E and F). Although broad sites of H3K4me1 enrichment were present across the entire Cyp27b1 locus and at the Cyp27b1 promoter region in the absence of added hormone, enrichment sites unique to the introns of Mettl1 and Mettl21b were also apparent (Fig. 1E). These potential enhancer regions in the kidney aligned pre-cisely with the more punctate pCREB and VDR sites of occupancy, yet were completely absent in similar analyses of MSCs. Results obtained through ChIP-seq analysis of the histone activity mark H3K9ac revealed similar profiles with the exception that PTH up-regulated while FGF23 strongly suppressed H3K9ac enrichment; 1,25(OH) 2 D 3 was without effect (Fig. 1F). These results suggest that both pCREB and VDR localize to unique and potentially active enhancers within specific introns of Mettl1 and Mettl21b and that the H3K9ac mark appears to be directionally regulated by PTH and FGF23 as might be expected, revealing significant activity for both of these hormones. Notably, ChIP-seq analysis of H3K9ac enrichment in kidney extracts from two mouse models of secondary hyperparathyroidism confirmed the profound effect of sustained high PTH at both the putative enhancers and at Cyp27b1 itself ( Fig. 1G) (25,26). These results support the novel finding that the regulatory regions that mediate PTH, 1,25(OH) 2 D 3 , and perhaps FGF23 action are located in selected introns of the Mettl1 and Mettl21b genes.
DNase I hypersensitivity sequencing (DHS) analysis of kidneys from untreated WT mice performed by the ENCODE Consortium revealed the presence of distinct chromatin-occupied sites across a broad region containing the Cyp27b1 locus from Tsfm to Tspan31 (Fig. 2C) (27). Although common DHS sites are seen in kidney and in many other tissues, only the kidney contained a unique DHS site within the Mettl1 intron and three unique sites in the Mettl21b intron that aligned not only with exclusive kidney enhancer marks but also with specific sites of pCREB and VDR occupancy. A similar profile of RNA pol II occupancy was also apparent consistent with an additional typical feature of enhancers ( Fig. 2D) (28). These findings support the hypothesis that the regions within the Mettl1 and Mettl21b introns likely represent a tissue-specific regulatory module that mediates transcriptional control of Cyp27b1 in the kidney by the three primary mineral-regulating endocrine hormones. Interestingly, this gene-dense region in which Cyp27b1 resides is defined by two major CCCTC-binding factor (CTCF)-occupied sites ( Fig. 2B) (29,30), prompting an examination of each gene between these sites. Both Cyp27b1 and Mettl21b but not Mettl1 or any additional genes in the region were similarly regulated by PTH, FGF23, and 1,25(OH) 2 D 3 (Fig. 3, A and B). Although mechanistically unproven, the presence of the two CTCF sites together with similarities in the regulation of both Mettl21b and Cyp27b1 advance speculation that this region may represent or reside in a small topologically associated domain (31).

Endocrine regulation of Cyp27b1
analogous to that reported by St-Arnaud and co-workers (26) (Fig. 4B). Visual inspection of each strain revealed that M21-IKO mice were similar to their WT counterparts, whereas M1-IKO mice were smaller and phenotypically similar to C27KO mice (Fig. 4C). Both of the latter two strains were extremely frail with mortality rates considerably higher than WT mice. Blood calcium and phosphate levels were similar in both M21-IKO and WT mice, whereas PTH levels were slightly elevated, and FGF23 levels were approximately half (Fig. 4D). In contrast, M1-IKO and C27KO mice displayed debilitating hypocalcemia and hypophosphatemia, highly elevated PTH levels, and a near absence of FGF23. Systemic parameters in the three mouse strains correlated directly with BMD measurements (Fig. 5A). CT analysis, parathyroid gland hypertrophy, growth plate abnormalities, and mineralization (Fig. 5, B-E) indicated that M1-IKO and C27KO mice were similar, although the profile seen in the C27KO mouse was more extreme. Interestingly, differences in the skeletal phenotype of the M21-IKO mice from WT mice were limited to the trabec-ular compartment (Fig. 5B), potentially due to a bone remodeling issue as mineralization appeared intact (Fig. 5E).

Endocrine control of Cyp27b1 expression is altered in M1-IKO and M21-IKO kidneys
The different phenotypes of the M1-IKO and M21-IKO mice suggested that the regions in Mettl1 and Mettl21b genes were responsible for distinct regulatory functions at the Cyp27b1 gene. To examine this hypothesis, we assessed the steady-state expression of Cyp27b1 and also the ability of each of the three hormones to regulate Cyp27b1 expression in C27KO, M1-IKO, and M21-IKO mice (Fig. 1A). This latter test was executed to determine whether the gene remained sensitive to each individual hormone (Fig. 1A) but not to assess any potential physiological impact. Basal levels of Cyp27b1 were strikingly elevated in C27KO mice compared with controls (Fig. 6, A and B) because, although the gene transcribes a mutant, truncated Cyp27b1 RNA and non-functional enzyme (26), RNA production, and regulation remained fully intact. Interestingly, although high

Endocrine regulation of Cyp27b1
residual levels of PTH in C27KO mice limited response of Cyp27b1 to exogenous PTH administration, these high levels were unable to prevent FGF23 or 1,25(OH) 2 D 3 from modestly suppressing Cyp27b1 (Fig. 6, A and B). In contrast, although WT mice retained normal sensitivity to each of the three hormones, basal Cyp27b1 expression in M1-IKO mice was strikingly reduced (95%) (Fig. 6A). Because both C27KO and M1-IKO mouse strains exhibit high PTH levels, the only explanation for the dramatic up-regulation of Cyp27b1 expression in C27KO mice coincident with the dramatic down-regulation of the gene in M1-IKO mice is that in the latter strain sensitivity to PTH has been lost. Loss of circulating PTH via thyroparathyroidectomy reduces basal production of 1,25(OH) 2 D 3 , an effect that has been previously documented (7). Cyp27b1 expression in both C27KO and M1-IKO mice exhibits a similar pattern of sensitivity to FGF23 and 1,25(OH) 2 D 3 suppression and loss of sensitivity to PTH induction. Therefore, it is clear that the loss of PTH sensitivity in the M1-IKO mouse is due to deletion of the Mettl1 intronic region. M21-IKO mice exhibit a similar but more limited decrease in basal Cyp27b1 expression (75%), retain complete sensitivity to PTH, yet are insensitive to both FGF23 and 1,25(OH) 2 D 3 treatment (Fig. 6, A and B). Thus, the region that is deleted in the M21-IKO mice appears to contain an independent regulatory feature that influences basal Cyp27b1 expression, perhaps due to the actions of an additional unknown positive factor analogous to that of PTH, as well as regions that mediate suppression by FGF23 and 1,25(OH) 2 D 3 . The co-regulation of Mettl21b (Fig. 3) is also lost as a result of these individual deletions, demonstrating that the activities of these enhancers are potentially shared (data not shown). Importantly, CYP27B1 protein levels in M1-IKO and M21-IKO kidneys were also strongly suppressed (Fig. 6E). As expected, high levels of non-functional, mutated CYP27B1 protein were seen in C27KO kidneys, migrating more rapidly than the WT protein due to loss of exons 7-9 ( Fig. 6E).

Basal suppression of kidney Cyp27b1 causes secondary suppression of Cyp24a1 and Vdr
Loss of circulating 1,25(OH) 2 D 3 and up-regulation of PTH in C27KO mice results in suppression of Cyp24a1 and Vdr expression (33). Cyp24a1 expression was dramatically down-regulated in both C27KO and M1-IKO mice (Fig. 6C), likely due to

Endocrine regulation of Cyp27b1
the high levels of PTH and low levels of FGF23 (Fig. 4D). Although Cyp24a1 remained sensitive to up-regulation by 1,25(OH) 2 D 3 in the kidneys of each strain, both FGF23 induction and PTH suppression appeared to be fully quenched. Surprisingly, M21-IKO mice also exhibit Cyp24a1 suppression, although this level was less dramatic. Nevertheless, Cyp24a1 retained sensitivity to PTH down-regulation and FGF23 and 1,25(OH) 2 D 3 up-regulation. As basal CYP24A1 protein levels were undetectable in WT, C27KO, M1-IKO, and M21-IKO mice, direct assessment of protein down-regulation was not possible (data not shown). The more modest secondary reduction in Cyp24a1 expression in M21-IKO mice is likely due to the limited increase in PTH but also to a reduction in FGF23, which would limit Cyp24a1 expression (considered below) (10). Vdr was also suppressed in C27KO and M1-IKO kidneys (Fig. 6D), a reduction that was confirmed directly at the protein level (Fig. 6E). The modest up-regulation of Vdr by 1,25(OH) 2 D 3 and FGF23 seen in WT mice appears to be lost in C27KO and M1-IKO mice (Fig. 6D), perhaps due to the low calcemic state of both mouse strains (34). Collectively, these studies of Cyp27b1, Cyp24a1, and Vdr expression in C27KO, M1-IKO, and M21-IKO mice suggest that their phenotypes are due to aberrant basal Cyp27b1 expression, selective loss of hormonal regulation in the kidney, and complex reciprocal homeostatic regulation of Cyp24a1. These activities appear to be intrinsic to the regulatory regions themselves, as ChIP-seq analysis of the Cyp27b1 locus in both M1-IKO and M21-IKO mice revealed that the deletion of either the Mettl1 or the Mettl21b intronic enhancer (underlined in red in Fig. 7, A and B) had no effect on either VDR or pCREB binding in the opposite regulatory region. In this analysis, VDR binding in the C27KO mouse was used as a control, as VDR levels in the kidneys of both C27KO and M1-IKO mice show similar downregulation and thus impact the level of genomic VDR binding in ChIP-seq analyses comparably. A similar analysis of epigenetic signature marks in these mice also revealed a lack of crossstructural alteration (data not shown). Importantly, as highlighted in Fig. 6, A and B, Cyp27b1 expression in M1-IKO mice is insensitive to the high circulating levels of PTH, and this is demonstrated amply through the lack of H3K9ac enrichment that is seen in both C27KO and VDRKO mice (Fig. 7C).

Endocrine regulation of Cyp27b1
in both systemic (Fig. 4D) and vitamin D metabolic (Fig. 6F) profiles appear instrumental in sustaining the near normal levels of 1,25(OH) 2 D 3 , thereby preserving mineral homeostasis in these particular mice. We emphasize that C27KO and M1-IKO mice were unable to produce 1,24,25(OH) 3 D 3 , a metabolite that is measured in the blood of WT mice and also appears to be produced at low but significant levels in M21-IKO mice. This represents the first robust measurement of a metabolite that, while comprising the first catabolic product of CYP24A1 action on 1,25(OH) 2 D 3 , is also the second most biologically active vitamin D metabolite known (35,36). Finally, the central importance of low 1,25(OH) 2 D 3 in the aberrant phenotype in the M1-IKO mouse was confirmed by the ability of a diet containing exogenously added 1,25(OH) 2 D 3 to rescue the abnormal skeletal phenotype of the M1-IKO mouse (Fig. 8A). An additional experiment in which vitamin D was removed from the diet, thereby reducing circulating 25(OH)D 3 levels, revealed that although the phenotype of the C27KO mouse was unaffected by this treatment regimen, as expected of a mouse that cannot make 1,25(OH) 2 D 3 , the phenotype of the M1-IKO mouse worsened and after 5 weeks was similar to that of the C27KO mouse (Fig. 8, B and C).

Repressive abilities of the intronic Mettl21b enhancer
Similarities between the ability of 1,25(OH) 2 D 3 and FGF23 to suppress Cyp27b1 expression in the kidney and their union in the intron of Mettl21b prompted the idea that 1,25(OH) 2 D 3 might suppress Cyp27b1, at least in part, through FGF23 by inducing transcription of the Fgf23 gene and raising its blood levels, an established function of 1,25(OH) 2 D 3 (37,38). Additional evidence includes the following. 1) ChIP-seq studies across the Cyp27b1 locus suggested that although both PTH induced and FGF23 suppressed H3K9ac enrichment within the intronic regions of Mettl1 and Mettl21b and at the Cyp27b1 promoter, 1,25(OH) 2 D 3 was without effect at this time point (Fig. 1F). 2) a reduction of VDR binding within the Mettl21b intron in the M1-IKO mouse due to Vdr down-regulation did not appear to limit the ability of either 1,25(OH) 2 D 3 or FGF23 to suppress Cyp27b1 expression (Fig. 7). To test this hypothesis, we correlated the ability of 1,25(OH) 2 D 3 to suppress Cyp27b1 expression over time in WT kidneys with its ability to induce Fgf23 in bone. The results revealed that following an acute but modest up-regulation of Cyp27b1 expression, 1,25(OH) 2 D 3 provoked a rapid suppression of Cyp27b1 that reached near baseline within 4 -6 h (Fig. 9A). This suppression was fully sustained for at least 24 h, an unusual feature of 1,25(OH) 2 D 3 activity alone, and likely much longer given previous studies (39). Importantly, Fgf23 transcripts were also induced by 1,25(OH) 2 D 3 in the skeleton within this same time frame (Fig.  9B). Based upon this finding, we then correlated these temporal responses with serum levels of PTH and FGF23 as well as calcium and phosphate. Although PTH was mildly suppressed by 1,25(OH) 2 D 3 , FGF23 was increased from a residual baseline of 200 to ϳ1700 pg/ml at 12 h (Fig. 9A). Both calcium and phosphate were also induced by 1,25(OH) 2 D 3 within this time period, an in vivo response to the hormone that has been well-established. These results support the idea that both 1,25(OH) 2 D 3 and FGF23 may act either sequentially or together to suppress Cyp27b1 expression through a sustained effect mediated by the up-regulation of FGF23. It is worth noting that the mobilization of phosphate may also play an important role because both 1,25(OH) 2 D 3 and phosphate are the most striking regulators of FGF23 that are currently known (40). A similar time course of Cyp27b1 response to 1,25(OH) 2 D 3 was also seen in M1-IKO kidneys (Fig. 9E), although this response was less robust because the basal expression of Cyp27b1 and serum levels of FGF23 are already highly suppressed (Figs. 4D and 6A). In WT mice, the temporary reduction of FGF23 and elevation in PTH allows PTH to transiently increase Cyp27b1 expression. However, Cyp27b1 expression is not acutely increased in the M1-IKO mice due to its insensitivity to PTH. Importantly, although this study suggests that the suppression of Cyp27b1 by 1,25(OH) 2 D 3 may involve the induction of FGF23 either directly by the hormone or via its mobilization of phosphate, proof of this hypothesis 29.0 ± 2.9 (12) 3396.7 ± 287.5* (8) 194.2 ± 13.2 (8) 6.7 ± 1.8* (7) 6.2 ± 2.5* (5) 8.9 ± 3.6* (5) 164.9 ± 13.9* (12) 6.8 ± 1.5*

Endocrine regulation of Cyp27b1
will require delineation of the exact sites of FGF23 action in the Mettl21b intron following identification of the transcription factor(s) involved and/or elimination of the ability of 1,25(OH) 2 D 3 to up-regulate Fgf23 expression.

Regulation of Cyp27b1 expression in NRTCs is functionally unique from the kidney
Previous studies have suggested that Cyp27b1 is expressed in a number of NRTCs and may be responsible for local conversion of 25(OH)D 3 to 1,25(OH) 2 D 3 (4). However, its expression does not appear to be regulated by either PTH or FGF23 and perhaps only modestly by 1,25(OH) 2 D 3 . Cyp27b1 expression is extremely low in NRTCs, and these hormones do not regulate Cyp27b1 in skin, bone, liver, and other specific cell types, although 1,25(OH) 2 D 3 modestly suppressed Cyp27b1 in the skin and in thyroparathyroid gland tissue (TPTG) (Fig. 10A and  11A). Based upon ChIP-seq analysis of MSCs (Figs. 1 and 2), we reasoned that the absence of Cyp27b1 regulation by PTH and FGF23 and, in specific NRTC cases, by 1,25(OH) 2 D 3 could be explained by the absence of the endocrine enhancers. To test this idea, we compared the basal expression levels of Cyp27b1 in several NRTCs from C27KO, M1-IKO, and M21-IKO mice with those of individual littermate controls. Basal levels of Cyp27b1 expression in skin, calvaria, and L5 vertebrae were unaffected by either of the two genomic deletions (Fig. 10B). Interestingly, both basal and ex vivo T cell receptor-inducible levels of Cyp27b1 expression in T cells isolated from either WT or M1-IKO spleens were also unaffected ( Fig. 10C) (41). These general findings provide key support for the idea that the regulatory module that mediates the actions of PTH, FGF23, and 1,25(OH) 2 D 3 at Cyp27b1 is absent in NRTCs and thus prevents similar regulation by the three endocrine hormones. Based upon tissue selectivity, we have termed this the kidney-specific

Endocrine regulation of Cyp27b1
endocrine regulatory module (KS-ERM). Surprisingly, although the calcium and phosphate hormones are unable to control Cyp27b1 expression in NRTCs, the ability of the inflammatory modulator LPS to induce Cyp27b1 was fully apparent (Fig. 10B). LPS induction of Cyp27b1 expression has been previously described, particularly in immune cells such as macrophages, where it is suggested to be involved in the ability of vitamin D to modulate both innate and acquired immunity (42). Importantly, the ability of LPS to induce the Cyp27b1 gene in these cell types was unaffected by the enhancer deletions introduced into the M1-IKO and M21-IKO mice, although LPS activity appeared to be potentially unmasked in the kidneys of these mutant mice as a consequence of the reduced basal Cyp27b1 expression in both strains (Fig. 10C). We conclude that LPS and T cell receptor activation of Cyp27b1 in distinct NRTCs is mediated by a regulatory region(s) we have termed the NRTC-regulatory module (NRTC-RM) whose location lies outside that defined by the two separate components of the KS-ERM (Fig. 10D).

Cyp27b1 locus in TPTG tissue contains signatures of both NRTCs and renal cells
Cyp27b1 is expressed at reduced levels in isolated cells of the parathyroid gland (PTG) and in ex vivo cultures of both PTG cells and organs of several species (43,44). However, the regulatory properties of the Cyp27b1 gene itself have not been explored in this tissue in vivo. We therefore examined in a concluding set of studies the potential for Cyp27b1 regulation in the parathyroid gland, based upon the strong linkage between the function of this organ and its central involvement in the regulation of both 1,25(OH) 2 D 3 metabolism and mineral homeostasis. This was enabled by excising total TPTG tissue after hormone injection into WT, C27KO, M1-IKO, or M21-IKO mice and conducting both RNA expression and ChIP-seq analysis in this set of glands. Although the tissue is more complex, this approach allowed us to avoid time-consuming isolation methods and ex vivo culturing and to assess analyses directly from the mice. Although much has been made of vitamin D action to suppress expression of the PTH gene, loss of the VDR by selective gene knock-out in the parathyroid glands of mice has revealed only a minor impact on PTH expression, suggesting that the primary actions of vitamin D in this tissue remain to be understood (45). Of the three hormones and LPS administered to WT mice, only 1,25(OH) 2 D 3 was statistically active and modestly suppressed Cyp27b1 expression (Fig. 11A). Loss of the enhancer in the M1-IKO mouse resulted in a striking suppression of Cyp27b1 much like that seen in the kidney; this suppression was not seen in the TPTG of M21-IKO mice (Fig. 11, A and B). We therefore treated WT mice with a single dose of 1,25(OH) 2 D 3 and examined the Cyp27b1 locus using ChIP-seq for the presence of the VDR at sites observed in the kidney (Fig. 1D). Despite the presence of the VDR at accepted target genes for 1,25(OH) 2 D 3 in the TPTG, including Cyp24a1, VDR occupancy was not detected anywhere within the Cyp27b1 locus (data not shown). To explore further, we examined the epigenetic landscape surrounding the Cyp27b1 gene in the TPTG, and we assessed the basal levels of the enhancer signature marks H3K4me1 and H3K9ac and compared them with those in the kidney and in MSCs. Although several minor similarities were apparent, neither of these epigenetic marks displayed a pattern consistent with that seen in the kidney (Fig.  11C). Similar to the kidney, H3K4me1 shows a minor enrichment in the M21-IKO region (furthest left highlighted region, Fig. 11C) as well as the M1-IKO region; however, H3K9ac is more comparable to the MSC control and devoid of activity in both regions. We then examined basal epigenetic H3K4me1 and H3K9ac landscapes across the Cyp27b1 gene in TPTG tissue derived from either C27KO or M1-IKO mice and compared them with those obtained from their WT littermates (Fig. 11D). Although modest H3K4me1 enrichment was seen in a single region in the Mettl21b intron and perhaps in the Mettl1 intron in these contrasting ChIP-seq analyses, there was no evidence of residual enrichment of either the H3K4me1 or H3K9ac activity marks in either intron. These results suggest that despite minor functional similarities, the Cyp27b1 gene locus in the TPTG reflects an enhancer landscape that is more similar to NRTCs than to the kidney. We conclude that the KS-ERM is a control module unique to the kidney, providing regulatory capabilities at the Cyp27b1 gene that mediate unique sensitivity to PTH, FGF23, and 1,25(OH) 2 D 3 .

Discussion
The phenotypes of the mice carrying KS-ERM deletions contrast with those of WT mice and suggest that reduced expression of Cyp27b1 stimulates striking adaptive homeostatic responses that involve the differential production of PTH and FGF23 as well as 1,25(OH) 2 D 3 . These responses impact not only Cyp27b1 and Cyp24a1 expression but vitamin D metabolism as well and highlight both the genomic as well as homeostatic circuits that are essential for the maintenance of normal mineral balance. A summary of the key players involved in this homeostatic response and our working hypothesis is illustrated in Fig. 12A, where the physiological actions of PTH, FGF23, and 1,25(OH) 2 D 3 are likewise shown reflecting the specific mouse models that have been used to describe the interrelationships. Sites of activity controlled by the enhancers deleted in the M1-IKO and M21-IKO strains are also depicted as well. As documented, most of these models are functionally linked to the aberrant regulation of Cyp27b1 expression or activity in the kidney, which leads to abnormal 1,25(OH) 2 D 3 production. Decreased circulating levels of 1,25(OH) 2 D 3 result in hypocalcemia and hypophosphatemia, prompting a rise in PTH and a suppression of FGF23. In the case of normal mice, homeostatic changes in PTH and FGF23 correct the underexpression of Cyp27b1 and suppress Cyp24a1 expression, raising blood 1,25(OH) 2 D 3 levels that together with PTH and FGF23 normalize extracellular calcium and phosphate metabolism.
In M1-IKO mice, however, and prompted by the down-regulation of Cyp27b1, the gene's up-regulation is disrupted due to the lack of Cyp27b1 sensitivity to PTH. Thus, calcium and phosphate lowering persists, causing a sustained rise in PTH and a further decrease in FGF23. These two hormonal changes also suppress Cyp24a1 expression that both decreases 1,25(OH) 2 D 3 turnover and compromises 24,25(OH) 2 D 3 and 1,24,25(OH) 3 D 3 production, thereby eliciting an elevation in 25(OH)D 3 substrate levels. This rise may contribute to a Endocrine regulation of Cyp27b1 homeostatic effort to raise circulating blood levels of 1,25(OH) 2 D 3 , which is largely futile in the case of the M1-IKO mice. In contrast, the reduced expression of Cyp27b1 in M21-IKO mice with retained gene sensitivity to PTH is sufficient to mediate appropriate up-regulation of 1,25(OH) 2 D 3 to maintain a normal calcium and phosphate balance. This largely prevents uncontrolled up-regulation of PTH while stimulating FGF23 synthesis. The homeostatic response also prevents the striking suppression of Cyp24a1 observed in the M1-IKO mice, thereby contributing to the successful preservation of 1,25(OH) 2 D 3 . It is possible that the full brunt of the loss of Cyp27b1 sensitivity to FGF23 and 1,25(OH) 2 D 3 suppression in the M21-IKO mice may be observed only under conditions in which 1,25(OH) 2 D 3 and FGF23 are elevated, and both hypercalcemia and hyperphosphatemia are evident (38). Nevertheless, the current results in M21-IKO mice highlight not only the nature of the exquisite homeostatic circuits that prevail in vivo to maintain mineral balance but demonstrate how the levels of key players are also modulated to achieve a normalized calcium and phosphate outcome. The inability of PTH to induce Cyp27b1 expression in M1-IKO mice almost fully compromises the capacity of Cyp27b1 and Cyp24a1 to maintain 1,25(OH) 2 D 3 levels, highlighting the essential role of PTH in the control of skeletal growth and development as well as the maintenance of mineral homeostasis.
The presence of a unique module in the kidney that mediates the regulation of Cyp27b1 by key hormones involved in mineral homeostasis together with its absence in NRTCs suggests that the primary role of Cyp27b1 in the kidney is to produce circulating blood levels of 1,25(OH) 2 D 3 that can orchestrate the expression of genes in the intestine, kidney, and bone, which are responsible for global mineral homeostasis. The absence of this  Figure 11. Parathyroid gland expression and genomic occupancy of Cyp27b1. A, gene expression of Cyp27b1 after treatment in 8 -9-week-old C57BL/6 wild-type mice with ethanol/PBS vehicle (Veh, gray, n ϭ 6), 230 ng/g bw PTH for 1 h (PTH, blue, n ϭ 6), 10 ng/g bw 1,25(OH) 2 D 3 (1,25D 3 , black, n ϭ 6) for 6 h, 50 ng/g bw FGF23 for 3 h (FGF23, green, n ϭ 6), or 10 mg/kg lipopolysaccharides (LPS, red, n ϭ 6) examined in the TPTG. *, p Ͻ 0.05 paired t test, treatment versus vehicle. B, gene expression of Cyp27b1 in C27KO, M1-IKO, and M21-IKO mice (KO, striped, n ϭ 6) basally (vehicle) compared with wild-type (WT, solid, n ϭ 6) littermates in the TPTG. *, p Ͻ 0.05 paired t test, KO versus WT. Data displayed as relative quantitation (RQ, mean Ϯ S.E.) compared with Gapdh. Overlaid data tracks for genomic occupancy by ChIP-seq of H3K4me1 and H3K9ac basally (yellow) in kidney, TPTG, and MSC (C) or C27KO and M1-IKO (blue) versus WT (yellow) (D). Overlapping track data appear as green. Regions of interest are highlighted in light gray boxes. Genomic location and scale are indicated (top), and maximum height of tag sequence density for each data track is indicated on the y axis (top left each track, normalized to input and 10 7 tags). Gene transcriptional direction is indicated by an arrow and exons by boxes.

Endocrine regulation of Cyp27b1
endocrine module in NRTCs and the simultaneous presence of an independent module outside the KS-ERM that controls Cyp27b1 expression in NRTCs (NRTC-RM) suggests that the expression of Cyp27b1 in these cell types serves an alternative function to regulate the production of autocrine 1,25(OH) 2 D 3 that controls cell-specific gene activation and local pleiotropic biological activities (46). If these separate roles of Cyp27b1 in the kidney and NRTCs are correct, it implies the existence of two relatively distinct spheres of vitamin D influence, the first comprising the control of mineral metabolism and the second consisting of the regulation of numerous biological activities that are cell type-specific and independent of mineral metabolism. Most importantly, this thesis indicates that the mechanistic basis for these two biologically active domains of vitamin D action are largely determined by the differential regulation of Cyp27b1 in these two tissue classes. These two domains are unlikely to be fully independent, however, considering that extracellular calcium and phosphate levels impact all cell types and that endocrine regulation by vitamin D also influences circulating 25(OH)D 3 , the common substrate for Cyp27b1 activity in both tissue types. Conversely, certain biological actions of vitamin D in NRTCs are also likely to impact the orchestration of mineral homeostasis by the intestine, kidney, and bone as well. For example, the potential impact of vitamin D on insulin secretion, utilization, and energy metabolism is likely to impact the maintenance of mineral homeostasis (47). We argue that the current mouse models we have created, as well as additional models in preparation, will allow us to isolate the activities of each sphere of influence independently, thereby permitting a precise determination of not only the impact of locally produced 1,25(OH) 2 D 3 on the biological activities of NRTCs but of the dependence of these activities on circulating 1,25(OH) 2 D 3 produced by the kidney. This is important because a central clinical issue currently is the impact of vitamin D supplementation on 25(OH)D 3 levels that are hypothesized to represent a primary determinant of the local production of 1,25(OH) 2 D 3 (48,49). Of greater importance, however, is the fact that both spheres of vitamin D influence may also be altered in disease, as illustrated in chronic kidney disease (40), autoimmune disease (50), and cancer (51)

Endocrine regulation of Cyp27b1
The implications for the locations of regulatory enhancers for the expression of both Cyp27b1 and Mettl21b are significant relative to genome-wide association studies attempting to correlate single nucleotide polymorphisms to the prevalence of autoimmune disease, particularly multiple sclerosis (52)(53)(54)(55). In these studies, disease-associated SNPs have been identified in an extended region that contains the CYP27B1 gene as well as neighboring genes that include METTL1 and METTL21B (FAM119B) (56,57). Indeed, some genotypes appear to be preferential for the altered expression of CYP27B1, although observed co-regulation of the METTL21B gene has introduced some complexity. Importantly, the human CYP27B1 locus and adjacent genes on chromosome 12 are syntenic on the reverse strand with the mouse haplotype. In fact, a VISTA analysis of genomic conservation indicates not only that this is the case but that the only non-exonic regions that reach significance are those in which we have identified enhancers (Fig. 12B). Our discovery that the regulatory regions that control the mouse Cyp27b1 gene in the kidney are embedded within the introns of Mettl1 and Mettl21b and that Mettl21b is also co-regulated by PTH, FGF23, and 1,25(OH) 2 D 3 are therefore potentially relevant. We hypothesize that the SNPs that have been identified in this region and particularly in the METTL1 gene in humans may hint at the locations of important regulatory enhancers for both CYP27B1 and METTL21B. It is also of significance that if non-coding SNPs in these genomic regions are linked to the expression of CYP27B1, our data would suggest that multiple sclerosis is associated with endocrine production of 1,25(OH) 2 D 3 and not local production by CNS-invasive immune cells. Thus, it will be of interest to determine whether the regulation of the human CYP27B1 gene in the kidney is modulated through a distal organization similar to that in the mouse and to determine the functional nature of the SNPs that have been identified. In that regard, our studies may have broader implications for vitamin D in other human autoimmune diseases as well.
In summary, a kidney-specific regulatory module has been identified upstream of the Cyp27b1 gene that controls both the basal and hormone-regulated expression of the gene in the kidney in mice in vivo. This module is absent in the Cyp27b1 region in NRTCs, which accounts for the absence of regulation by PTH, FGF23, and 1,25(OH) 2 D 3 in these cell types. Dissection of this region in mice via CRISPR/Cas9 genome-editing methods has revealed distinct elements responsible for basal expression in the kidney as well as for PTH induction and 1,25(OH) 2 D 3 and FGF23 suppression. The phenotypes of these mouse strains highlight both the genomic and homeostatic circuits that control Cyp27b1 expression. We conclude that these studies open new avenues to study vitamin D metabolism and its involvement in both mineral metabolism and in other pleiotropic biology in vivo in health and disease.
Traditional genotyping PCR was completed with GoTaq (Promega, Madison, WI), and all real-time qPCR was completed with the StepOnePlus using TaqMan for gene expression assays (Applied Biosystems, Foster City, CA). Primers were obtained from IDT (Coralville, IA).

Gene expression
Dissected tissues were frozen immediately in liquid nitrogen and stored at Ϫ80°C. Frozen tissues were homogenized in TRIzol reagent (Life Technologies, Inc.), and RNA was isolated per the manufacturer's instructions. 1 g of isolated total RNA was DNase-treated, reverse-transcribed using the High Capacity cDNA kit (Applied Biosystems), and then diluted to 100 l with RNase/DNase free water. qPCR was performed using primers specific to a select set of differentially expressed genes by Taq-Man analyses. TaqMan gene expression probes (Applied Biosystems) were used for RT-PCR and are found in Table 1.

ChIP followed by sequencing (ChIP-seq)
ChIP was performed using antibodies listed under "Reagents." ChIP was performed as described previously with several modifications (21). In brief, kidney and TPTG samples were obtained from mice treated with 1,25(OH) 2 D 3 , PTH, FGF23, or the corresponding vehicle for 1 h. Two kidneys or 10 TPTGs were combined per replicate, rinsed with ice-cold 1ϫ PBS, and then subjected to immunoprecipitation using either a control IgG or the indicated antibodies. Statistical analysis and data processing for ChIP-seq assays were performed as reported previously (58). The isolated DNA (or input DNA acquired prior to precipitation) was then validated by qPCR and further prepared for ChIP-seq analysis. ChIP-seq libraries were prepared as described previously (58). Conservation analysis was performed with VISTA Point comparing the mouse (mm9) with human (hg19) genomic regions surrounding the CYP27B1 gene (59).

CRISPR-generated and transgenic mice
The guides used for CRISPR-Cas9-mediated genome editing were optimized for the least number of potential off-target sites and fewest sites within coding exons using the Zhang Laboratory CRISPR Design tool (crispr.mit.edu) 3 and cross-referenced with Liu lab CRISPR-DO (cistrome.org/crispr). 3 The guides (see Table 1) were annealed and cloned into plasmids pX330 or pX458 obtained from Zhang Laboratory via Addgene (Cambridge, MA) as described recently (60). Resulting PCR products were then transcribed in vitro utilizing the T7 MEGAshortscript kit (Life Technologies, Inc.) (61). The mixture of 50 ng/l of the produced RNA guides and 40 ng/l Cas9 protein in injection buffer (5 mM Trizma base, 5 mM Tris-HCl, 0.1 mM EDTA, pH 7.4) was injected into the pronucleus of 1 day-fertilized embryos isolated from hyper-ovulating female C57BL/6 mice, as described previously (62), and implanted into recipient females at the University of Wisconsin-Madison Biotechnology Center Transgenic Animal Facility. The resultant pups were genotyped with spanning primers (Table 1), cloned, and sequenced. The top 10 predicted potential off-target sites were examined by PCR and sequencing analysis.
The VDR-null (VDRKO) strain was produced by targeted ablation of the second zinc finger of the VDR DNA-binding domain (25). VDRKO mice were fed a standard rodent chow diet (5008; Lab Diet, St. Louis, MO). Cyp27b1 null (C27KO) mice were produced by GenOway (Strasbourg, France) with a strategy similar to that previously described (63). In brief, the mouse Cyp27b1 gene was amplified by PCR using genomic DNA from C57BL/6 ES cells as a template and cloned into the pCR4-TPOP vector (Invitrogen) to create a targeting vector. Two loxP sites were inserted in intron 6 and downstream of exon 9 and a positive selection marker, neomycin resistance gene, which is flanked by FRT sites, was introduced between exon 9 and downstream loxP site. Diphtheria toxin A was also introduced upstream of the Cyp27b1 gene as a negative selection marker to reduce the isolation of non-homologous recombined ES cell clones and to enhance the chance to isolate ES cell clones containing the distal loxP site. The targeting vector was linearized and electroporated into ES cells. The recombined ES cell clones identified by the selection, PCR confirmation, and Southern blot analysis were then injected into C57BL/6 blastocysts to create chimeric mice, and the injected blastocysts were reimplanted into OF1 pseudopregnant females and allowed to develop. To remove the selection marker flanked by FRT sites and exons 7-9 of the Cyp27b1 gene flanked by loxP sites, the recombinant mice were sequentially bred with Flp-recombinase expressing deleter mice and then Cre-recombinase expressing deleter mice. For genotyping, a primer set amplifying a DNA region between the last exon and downstream of the last exon (880 bp) and a primer set amplifying a DNA region between intron 5 and downstream of the last exon (560 bp) were used for WT and the deleted alleles, respectively (Table 1).

Animal studies
Genetically modified mice were outbred with C57BL/6 mice (Envigo, Indianapolis, IN, and The Jackson Laboratory, Bar Harbor, ME) as heterozygotes. Mice were housed in high-density ventilated caging in the Animal Research Facility of the University of Wisconsin-Madison under 12-h light/dark cycles, 72°F temperature, and 45% humidity. All mice used in this study were maintained on a standard rodent chow diet (5008, Lab Diet, St. Louis, MO), aged 8 -9 weeks, and backcrossed five or more generations with WT mice unless otherwise indicated. For the 1,25(OH) 2 D 3 rescue diet study, 3-week-old mice were fed either a 1,25(OH) 2 D 3 rescue diet (5008 chow diet ϩ 1 ng/g 1,25(OH) 2 D 3 ) or standard chow (5008) for 5 weeks, at which time the animals were sacrificed, and tissues were collected. For the vitamin D 3 -deficiency study, 3-week-old mice were fed either a vitamin D 3 -deficient diet containing 0.6% calcium and 0.4% phosphorus (TD.01102, Envigo, Madison, WI) or a control diet containing 0.6% calcium, 0.4% phosphorus, and 2200 IU vitamin D/kg (TD.97191, Envigo) for 5 weeks, at which time animals were sacrificed, and tissues were collected. All experiments and tissue collections were performed in the Animals were injected intraperitoneally with 10 ng/g body weight (bw) 1,25(OH) 2 D 3 (in propylene glycol), 230 ng/g bw PTH(1-84) (in PBS), 50 ng/g bw FGF23 (in PBS ϩ 0.1% BSA), 10 mg/kg bw LPS (in PBS), or vehicle (EtOH or PBS). Animals were sacrificed, and tissues were collected 1 h after PTH injection, 3 h after FGF23 injection, and 6 h after 1,25(OH) 2 D 3 and LPS injections. Unless otherwise indicated, all experiments were done with equal numbers of males and females (n Ն 6). There were no differences in our data between males and females; therefore, the data presented combine both males and females in all studies unless otherwise indicated.

Blood chemistry
Cardiac blood was collected at the time of sacrifice. Collected blood was split into serum-or EDTA-treated plasma, incubated at room temperature for 30 min, followed by centrifugation at 6000 rpm for 12 min (twice) to obtain serum or EDTA plasma. Serum calcium and phosphate levels were measured using QuantiChrom TM calcium assay kit (DICA-500, BioAssay Systems, Hayward, CA) and QuantiChrom TM phosphate assay kit (DIPI-500, BioAssay Systems). Circulating intact FGF23 and PTH were measured in EDTA plasma via mouse/rat FGF-23 (intact) ELISA kit (60-6800, Immutopics, San Clemente, CA) and mouse PTH(1-84) ELISA kit (60-2305, Immutopics), respectively.

BMD and CT analysis
At 8 -9 weeks of age, BMDs of the CRISPR-generated mice and their WT littermates were measured and analyzed by dual X-ray absorptiometry with a PIXImus densitometer (GE-Lunar Corp., Madison, WI), as described previously (63). CT analyses were performed using a high-resolution benchtop Scanco uCT50 (Scanco Medical, Bassersdorf, Switzerland) at the Vanderbilt Center for Bone Biology. Tomographic images were acquired at 70 peak kilovoltages and 200 A with an isotropic voxel size of 12.0 mm and at an integration time of 400 ms with 1000 projections per 360°rotation. Reconstructed CT images

Endocrine regulation of Cyp27b1
were binarized, using a Gaussian noise filter with ϭ 0.3 and support ϭ 1 and a threshold of 470 mg of hydroxyapatite/ccm for trabecular bone or 0.3, 1, and 733 mg of HA/cm 3 for midshaft cortical bone. The volume of interest for trabecular bone in the distal femoral metaphysis was the entire medullary area located 0.360 -1.560 mm proximal to the distal growth plate. The midshaft volume of interest was a 1.2-mmlong segment of cortical bone centered between the femoral head and surface of the distal condyles. All volumetric structural parameters were calculated from the 3D renderings using Scanco Evaluation software following standard methods and nomenclature (69).

Immunohistochemistry
Mice were perfused with PBS followed by 4% paraformaldehyde. The collected tissues were immersed overnight in 4% paraformaldehyde and then overnight again in 30% sucrose. Thyroparathyroid tissues were processed to prepare frozen tissue blocks using Neg-50 (Richard-Allan Scientific) and cut into 10-m sections with an HM5050E microtome (Microm). The tissue sections were stained with hematoxylin stain solution (3530-32; Ricca Chemical Co.) and eosin-Y (Fisher). The stained tissue sections were mounted with Permount (Fisher). Images for histological assay were taken using an ECLIPSE Ti-S microscope (Nikon) and QICAM 12-bit Mono Fast 1394 cooled camera (QImaging). PTG sizes were determined using ImageJ software. After CT analysis, undecalcified tibiae were embedded in methyl methacrylate, sectioned on an automatic retractable Microtom 355 with a D-profile, tungsten carbide steel knife at 4 m. Adjacent sections were stained with Masson trichrome (70) or with von Kossa to visualize mineralized bone.

T cell activation
Spleen CD4 ϩ or CD8 ϩ T cells were isolated by negative selection using EasySep mouse CD4 ϩ or CD8 ϩ T cell isolation kit (Stemcell Technologies, Vancouver, British Columbia, Canada) per the manufacturer's instructions. Isolated CD4 ϩ or CD8 ϩ T cells were plated in 24-well plates with a density of 1 ϫ 10 6 cells/well and treated 3 h with vehicle (PBS) or 25 l/well of CD3/CD28 mouse T-activator Dynabeads (Gibco, Life Technologies, Inc.). In addition, isolated CD4 ϩ or CD8 ϩ T cells were plated in 24-well plates with a density of 1 ϫ 10 6 cells/well and treated 3 h with vehicle (EtOH) or 100 nM 1,25(OH) 2 D 3 .

Statistical evaluation
Data were analyzed using GraphPad Prism 7 software (GraphPad Software, Inc., La Jolla, CA) and in consultation with the Statistics Department, University of Wisconsin. All values are reported as the means Ϯ S.E., and differences between group means were evaluated using one-way ANOVA, two-way ANOVA, or Student's t test as indicated in the figure legends.