COMMENTARY The Journal of Clinical Investigation β Cells led astray by transcription factors and the company they keep

Peter Thompson and Anil Bhushan Diabetes Center, UCSF, San Francisco, California, USA.

crine cell fates as a consequence of vari- ous stressors (2, 4, 17). Importantly, these Pancreatic β cells have one of the highest secretion burdens in the studies probe the different relationships body, as these cells must synthesize and secrete in proportion to of NKX2.2, PAX6, and the LIM domain– postprandial rises in blood glucose. Remarkably, it is now becoming clear binding protein 1–islet 1 (LDB1-ISL1) that adult β cells retain plasticity and can dedifferentiate into embryonic complex to the chromatin modification fates or adopt alternate islet endocrine cell identities. This property is machinery in adult β cells and show how especially important, because changes in cell fate alter β cell function and the expression and function of these TFs could form the basis for defects in insulin secretion that occur early in are sensitive to environmental and meta- the pathogenesis of the most prevalent form of β cell dysfunction, type bolic changes associated with T2D. 2 diabetes. In this issue, three different studies provide complementary perspectives on how the transcription factors NK2 2 (NKX2.2), TFs: maintenance of β cell paired box 6 (PAX6), and LIM domain–binding protein 1 (LDB1) serve to identity on different levels maintain mature adult β cell identity, revealing clues as to how adult β β cell–specific ablation of Nkx2.2, Pax6, cells can partially dedifferentiate or become reprogrammed into other islet or Ldb1 in adult animals resulted in a endocrine cells. similar suite of metabolic and physio- logic defects in β cell function, including loss of insulin content, defects in insu- lin secretion, and glucose intolerance, β Cell plasticity defects in the expression or regulation of with or without accompanying fasting or A growing body of evidence from stud- β cell TFs as well (4, 5). An understanding fed hyperglycemia (8–10). However, the ies in mice and humans indicates that of the mechanisms that drive β cell dedif- underlying cause of these related phe- dedifferentiation or partial transdiffer- ferentiation and partial transdifferentia- notypes was found to be quite different entiation/reprogramming of β cells into tion into α, (PP), in each case. Dominguez et al. showed alternate endocrine cell fates can alter or δ cell fates in response to various insults that Nkx2.2-KO β cells become partial- glucose homeostasis (1–5). β Cell dediffer- is only beginning to emerge. This knowl- ly transdifferentiated into α- and δ-like entiation is generally characterized by the edge will pave the way toward the devel- cells (8). This phenotype, along with the loss of insulin content and expres- opment of new therapeutic approaches ensuing hyperglycemia, is reminiscent of sion, as well as decreased expression of to prevent or reverse the progressive loss the β cell phenotype encoding β cell–specific transcrip- of functional β cell mass in patients with that is observed upon deletion of Pdx1 or tion factors (TFs), such as NK6 homeo- (T2D). In this issue, stud- Nkx6.1 in adult β cells, respectively (12, box 1 (NKX6.1), pancreatic and duodenal ies from the laboratories of Sussel (8), Dor 14). NKX2.2 functions primarily as an homeobox 1 (PDX1), forkhead box protein (9), and Stoffers (10) collectively rein- activator of β cell genes (including TFs O1 (FOXO1), neuronal differentiation 1 force the consensus view that the same and glucose metabolic genes) and as a (NEUROD1), and MAF bZIP transcription TFs that direct β cell specification are also repressor of metabolically disallowed factor A (MAFA) (6), and reactivation of required to maintain the gene activation genes and genes expressed by other islet endocrine progenitor cell markers such and repression programs that undergird endocrine cell types (8). Similarly, Swisa as neurogenin 3 (NGN3) (4, 7). Similarly, mature β cell identity (11–15). This unex- and colleagues demonstrated that adult β adult β cells can also become transdif- pected property of the murine and human cells lacking PAX6 also become partially ferentiated or partially reprogrammed to β cell transcriptional network provides an transdifferentiated by acquiring ghrelin adopt alternate islet cell properties, such additional explanation for the intrinsic expression (9). PAX6 was found to func- as the capacity for glucagon and/or soma- fragility, sensitivity (16), and propensity tion as both an activator and repressor, tostatin production, accompanied by of these cells to adopt alternate endo- co-occupying enhancers bound by the β cell–specific TFs PDX1, NKX6.1, MAFA, Related Articles: pp. 215, 230, and 244 FOXA2, and NEUROD1. Additionally, PAX6 also indirectly maintained repres- sion of the undifferentiated endocrine Conflict of interest: The authors have received investigator-initiated funding from Disease Interception Accelerator, Janssen Pharmaceuticals. progenitor program through regulation Reference information: J Clin Invest. 2017;127(1):94–97. doi:10.1172/JCI91304. of Foxa2 (9). Ediger et al. determined that

94 jci.org Volume 127 Number 1 January 2017 The Journal of Clinical Investigation COMMENTARY

Figure 1. Roles of the islet cell TFs PAX6, NKX2.2, and LDB1-ISL1 in the maintenance of adult β cell identity. (A) An example of how PAX6 might regulate the activity of a β cell enhancer. PAX6 binds to its DNA sequence with a combination of different β cell–spe- cific TFs, represented by PDX1, NKX6.1, and MAFA. Alongside the β cell TFs, PAX6 could modulate the activity status of the enhancer by promoting recruitment of histone methyl- transferases (HMTs) that deposit H3K4me1 and histone acetyltransferases (HATs) required for H3K27ac, while inhibiting recruitment of histone deacetylases (HDACs). (B) In a scenario similar to that depicted in A, NKX2.2 could regulate the activity state of a poised enhancer by binding to its sequence along with β cell–specific TFs and recruiting HMTs to maintain H3K4me1 and HDACs to maintain deacetylation of H3K27. (C) LDB1-ISL1 multimeric complexes bind sequences in the enhancer along with β cell TFs to form looped chromatin structures, which may be nec- essary for high-level enhancer activity toward active (H3K4me3-marked) target promoters. Whether the looping structure is necessary for the active enhancer marks H3K4me1 and H3K27ac, or occurs as a consequence of these marks, remains to be determined.

LDB1, in concert with its DNA-binding TFs and chromatin modifiers: toward an answer to this question may lie interactor ISL1, is required to maintain keeping company in adult β cells in the relationship between TFs and chro- the active state of enhancers, thereby During development and endo- matin modifiers in terminally differentiat- supporting the function of the β cell– crine lineage specification, a sequential cas- ed cells, such as β cells, as compared with specific TFs PDX1, NKX6.1, and FOXA2 cade of TFs is known to control the ordered other more proliferative, nonsecretory cell (10). It should be noted, however, that the differentiation process to a pan-endocrine types. Indeed, the β cell–specific deletion metabolic and physiologic consequences fate that ultimately culminates in the pro- of TFs or chromatin modifiers does not of the adult β cell–KO phenotype for Ldb1 duction of the six different islet endocrine result in β cells adopting fates outside the and Isl1 were less severe than the conse- cell types, including β cells. At each stage, endocrine lineage (8–10, 12, 14, 15, 17, 18), quences observed in β cells lacking the epigenetic mechanisms, including DNA suggesting that plasticity is limited to fates TFs that functionally interact with LDB1- methylation and histone modifications, within the endocrine lineage and that both ISL1, suggesting that the β cell TFs may provide so-called “memory” for partic- TFs and chromatin modifiers maintain the perform some of their gene-regulatory ular gene functional states (e.g., active, mature endocrine cell fate. Rather than functions independently of LDB1-ISL1. repressed, poised). Thus, while TFs may dictating an establishment versus main- Therefore, while different in their adult establish a cell’s identity, epigenetic mech- tenance paradigm, TFs and chromatin β cell–KO phenotypes, NKX2.2 and PAX6 anisms are generally thought to maintain modifiers may cooperate on an ongoing may have similar functional roles in pro- this identity, such that the perpetual activi- basis to maintain appropriate endocrine viding activator and repressor activities ty of the pioneering TFs may not always be cell phenotypes (Figure 1, A and B). This at enhancers and a small subset of pro- necessary. Why, then, are the same devel- mechanism may lend itself to the ability moters. In contrast, LDB1-ISL1 serve as opmental TFs necessary to maintain par- to rapidly adapt to changes in homeostasis chromatin structural factors to maintain ticular promoter/enhancer activity states at the transcriptional level, as is necessary β cell enhancer activity. in fully matured adult β cells? Some clues for professional, nondividing secretory

jci.org Volume 127 Number 1 January 2017 95 COMMENTARY The Journal of Clinical Investigation cells like the β cells that are wired to mod- showed that NKX2.2 directly represses the MAFA, FOXO1, and NKX6.1, are vulnera- ulate output on the basis of environmental Arx promoter sequence (20). Conversely, ble to dysregulation in T2D due to metabol- signals. Notably, both epigenetic “writ- whether chromatin modifications them- ic stressors, such as chronic hyperglycemia ers” (e.g., histone methyltransferases and selves feed back on the functions of TFs and oxidative damage (1, 3, 6). These obser- acetyltransferases and DNA methyltrans- in β cells remains to be determined. Thus, vations dovetail nicely with the intrinsic ferases) and “erasers” (histone demethy- PAX6 and NKX2.2 are examples of islet fragility of mature β cells when challenged lases, deacetylases, and ten-eleven trans- cell TFs that modulate the activity status with various stressors (4, 16) and suggest location [TET] ) are expressed in of enhancers and promoters as the result that the sensitivity of these core β cell TFs adult mouse and human β cells; therefore, of sequence context and combinatorial to stress is an explanation for β cell fragili- a balance between the activities of these interactions with other TFs and chromatin ty. The studies on PAX6 and LDB1 indicate chromatin modifiers must be maintained modifiers (Figure 1, A and B), as has been that they too fall into this category. Indeed, at active versus repressed loci throughout recently demonstrated for other groups of PAX6 mRNA and protein levels were down- the genome. TFs (21). LDB1 and ISL1 function as chro- regulated in hyperglycemic insulin recep- TFs can activate or repress gene matin structural/looping factors for TF tor antagonist–injected mice and in leptin expression via a wide variety of direct complexes at active β cell enhancers (Fig- –deficient (db/db) mice in a man- and indirect mechanisms (19) including: ure 1C). Active enhancers are also known ner that correlated with the age-dependent regulation of of other to produce short noncoding transcripts severity of hyperglycemia (9). Similarly, TFs; differences in binding partners and known as enhancer RNAs (eRNAs), and LDB1 mRNA and protein levels were also chromatin modifiers of these TFs; interac- this looping might be required for eRNA downregulated in db/db mice starting at 8 tions with noncoding RNAs; alternatively production (22). Enhancer structural loop- weeks of age (10). Whether NKX2.2 is sim- spliced TF isoforms (which was suggest- ing is necessary to enforce the fully dif- ilarly disrupted in db/db mice or in human ed for PAX6; ref. 9); and decoy/blockade ferentiated β cell transcriptional program, T2D awaits further inquiry. NKX2.2, PAX6, of other TFs. Thus, it is not surprising to since KO of Ldb1 or Isl1 in adult β cells led and LDB1 seem to play similar roles in find that NKX2.2 and PAX6 can serve as to dedifferentiation into Ngn3+ endocrine human β cells, as evidenced from the ChIP- both activators and repressors of different progenitors (10). In this case, LDB1 and seq analyses of human islets (8–10), and genes in adult β cells (Figure 1, A and B). A ISL1 are likely to be recruited with the β NKX2.2 and PAX6 mutations have been further layer of complexity for deciphering cell–specific TFs themselves (ref. 10 and found to cause neonatal diabetes in humans the functions of NKX2.2 and PAX6 in the Figure 1C) and provide an additional lay- (25, 26), underscoring the functional impor- β cell transcriptome is evidenced by the er of epigenetic information in the form tance of these TFs for human β cell forma- finding that these TFs also exert regulatory of chromatin looping to sustain enhancer tion and function. In addition, it was sug- effects on the activity and/or expression of activity. Whether these LDB1-ISL1 looped gested that the enhancer state maintained other β cell–specific TFs, including PDX1, domains are needed at all active β cell by LDB1-ISL1 becomes progressively com- NKX6.1, NEUROD1, MAFA, and FOXA2 enhancers, or just at those promoting promised with age and in T2D individuals (8, 9). The sequence context and, thus, the high-level gene expression, remains to be (10). Whether the observed changes in presence of other TFs bound to the same determined. Future investigations could expression and function of these TFs in the enhancer are likely to be major factors in use conformation capture face of metabolic stressors, aging, and T2D determining the functional consequences carbon copy–based (5C-based) approach- cause the functional decline of β cells, or are of PAX6 and NKX2.2 binding (Figure 1, A es to identify genes controlled by LDB1- merely a consequence of the pathogenesis, and B). These context-dependent effects ISL1–bound enhancers in β cells. More- requires further investigation. Reversal of β were evidenced by the ability of PAX6 over, recent work in hematopoietic and cell oxidative stress by crossing db/db mice to activate or repress luciferase reporter neurodevelopmental models has demon- with glutathione peroxidase 1 (Gpx1) trans- constructs in Min6 cells in the same way strated that LDB1-enforced chromatin genic mice led to restoration of the appro- these sequences are regulated in vivo (9). looping at enhancers can dictate a vari- priate expression and localization of β cell The regulatory effect of PAX6 on different ety of functional states, including POL TFs (6). In addition, upon dedifferentiation enhancer constructs in Min6 cells also cor- II–paused repression and developmen- in other contexts, β cells were also found to related with the degree of the histone mark tally controlled transcriptional activation redifferentiate appropriately upon removal H3K27ac at each region, as determined (23, 24). Therefore, it will be of interest of metabolic stress (2), suggesting that this from ChIP-sequencing (ChIP-seq) data to investigate how LDB1-ISL1 complex- may be the case. Therefore, therapeutic sets on these cells (9), thereby affirming the es, which play general roles in enhancer approaches aimed at restoring islet cell TF interplay between TF binding and chroma- organization in multiple lineages, orches- expression and/or localization by mitigat- tin modification status (Figure 1, A and B). trate enhancer-promoter relationships in ing stress could improve β cell function and Similarly, in a previous study, NKX2.2 was adult β cells. prevent deterioration of functional β cell found to bind and repress the aristaless- mass in T2D. related homeobox (Arx) promoter in β cells TF sensitivity to stress by recruitment of DNA methyltransferase underlies β cell fragility Address correspondence to: Anil Bhu- 3a (DNMT3a) and histone deacetylase 1 The emerging view from recent studies shan or Peter Thompson, Diabetes Cen- (HDAC1) complexes, and reporter assays suggests that β cell TFs, including PDX1, ter, UCSF, 513 Parnassas Avenue, San

96 jci.org Volume 127 Number 1 January 2017 The Journal of Clinical Investigation COMMENTARY

Francisco, California 94134, USA. Phone: 9. Swisa A, et al. PAX6 maintains β cell identity by 19. Spitz F, Furlong EE. Transcription factors: from repressing genes of alternative islet cell types. enhancer binding to developmental control. Nat 310.206.5750 (A. Bhushan), 415.691.1484 J Clin Invest. 2017;127(1):230–243. Rev Genet. 2012;13(9):613–626. (P. Thompson); E-mail: anil.bhushan@ 10. Ediger BN, et al. LIM domain–binding 1 main- 20. Papizan JB, et al. Nkx2.2 repressor complex ucsf.edu (A. Bhushan), peter.thompson@ tains the terminally differentiated state of pan- regulates islet beta-cell specification and pre- ucsf.edu (P. Thompson). creatic β cells. J Clin Invest. 2017;127(1):215–229. vents α-to-β-cell reprogramming. Genes Dev. 11. Piccand J, et al. Rfx6 maintains the functional 2011;25(21):2291–2305. 1. Talchai C, Xuan S, Lin HV, Sussel L, Accili identity of adult pancreatic β cells. Cell Rep. 21. Stampfel G, Kazmar T, Frank O, Wienerroither S, D. Pancreatic β cell dedifferentiation as a 2014;9(6):2219–2232. Reiter F, Stark A. Transcriptional regulators form mechanism of diabetic β cell failure. Cell. 12. Schaffer AE, et al. Nkx6.1 controls a gene reg- diverse groups with context-dependent regula- 2012;150(6):1223–1234. ulatory network required for establishing and tory functions. Nature. 2015;528(7580):147–151. 2. Wang Z, York NW, Nichols CG, Remedi MS. Pan- maintaining pancreatic identity. PLoS 22. Kim TK, Hemberg M, Gray JM. Enhancer RNAs: creatic beta-cell dedifferentiation in diabetes Genet. 2013;9(1):e1003274. a class of long noncoding RNAs synthesized and redifferentiation following insulin therapy. 13. Gu C, et al. Pancreatic β cells require NeuroD to at enhancers. Cold Spring Harb Perspect Biol. Cell Metab. 2014;19(5):872–882. achieve and maintain functional maturity. Cell 2015;7(1):a018622. 3. Spijker HS, et al. Loss of β-cell identity occurs in Metab. 2010;11(4):298–310. 23. Krivega I, Dale RK, Dean A. Role of LDB1 in the type 2 diabetes and is associated with islet amy- 14. Gao T, et al. Pdx1 maintains β cell identity and transition from chromatin looping to transcrip- loid deposits. Diabetes. 2015;64(8):2928–2938. function by repressing an α cell program. Cell tion activation. Genes Dev. 2014; 4. Cigliola V, Thorel F, Chera S, Herrera PL. Metab. 2014;19(2):259–271. 28(12):1278–1290. Stress-induced adaptive islet cell identity chang- 15. Nishimura W, Takahashi S, Yasuda K. MafA 24. Zhang F, et al. Enhancer-bound LDB1 reg- es. Diabetes Obes Metab. 2016;18(suppl 1):87–96. is critical for maintenance of the mature ulates a corticotrope promoter-pausing 5. Remedi MS, Emfinger C. Pancreatic β-cell β cell phenotype in mice. Diabetologia. repression program. Proc Natl Acad Sci U S A. identity in diabetes. Diabetes Obes Metab. 2014;58(3):566–574. 2015;112(5):1380–1385. 2016;18(suppl 1):110–116. 16. Dooley J, et al. Genetic predisposition for β cell 25. Flanagan SE, et al. Analysis of transcription 6. Guo S, et al. Inactivation of specific β cell tran- fragility underlies type 1 and type 2 diabetes. Nat factors key for mouse pancreatic development scription factors in type 2 diabetes. J Clin Invest. Genet. 2016;48(5):519–527. establishes NKX2-2 and MNX1 mutations as 2013;123(8):3305–3316. 17. Bramswig NC, et al. Epigenomic plasticity causes of neonatal diabetes in man. Cell Metab. 7. Brereton MF, Rohm M, Ashcroft FM. β-Cell dys- enables human pancreatic α to β cell reprogram- 2014;19(1):146–154. function in diabetes: a crisis of identity? Diabetes ming. J Clin Invest. 2013;123(3):1275–1284. 26. Solomon BD, et al. Compound heterozygosity Obes Metab. 2016;18(suppl 1):102–109. 18. Dhawan S, Georgia S, Tschen SI, Fan G, Bhushan for mutations in PAX6 in a patient with com- 8. Domínguez Gutiérrez G, et al. Pancreatic β cell A. Pancreatic β cell identity is maintained by plex brain anomaly, neonatal diabetes mel- identity requires continual repression of non–β DNA methylation-mediated repression of Arx. litus, and microophthalmia. Am J Med Genet. cell programs. J Clin Invest. 2017;127(1):244–259. Dev Cell. 2011;20(4):419–429. 2009;149A(11):2543–2546.

jci.org Volume 127 Number 1 January 2017 97