Enhancing Gonadotrope Gene Expression Through Regulatory lncRNAs.

The world of long non-coding RNAs (lncRNAs) has opened up massive new prospects in understanding the regulation of gene expression. Not only are there seemingly almost infinite numbers of lncRNAs in the mammalian cell, but they have highly diverse mechanisms of action. In the nucleus, some are chromatin-associated, transcribed from transcriptional enhancers (eRNAs) and/or direct changes in the epigenetic landscape with profound effects on gene expression. The pituitary gonadotrope is responsible for activation of reproduction through production and secretion of appropriate levels of the gonadotropic hormones. As such, it exemplifies a cell whose function is defined through changes in developmental and temporal patterns of gene expression, including those that are hormonally induced. Roles for diverse distal regulatory elements and eRNAs in gonadotrope biology have only just begun to emerge. Here, we will present an overview of the different kinds of lncRNAs that alter gene expression, and what is known about their roles in regulating some of the key gonadotrope genes. We will also review various screens that have detected differentially expressed pituitary lncRNAs associated with changes in reproductive state and those whose expression is found to play a role in gonadotrope-derived nonfunctioning pituitary adenomas. We hope to shed light on this exciting new field, emphasize the open questions, and encourage research to illuminate the roles of lncRNAs in various endocrine systems.

Transcriptional enhancers are defined by function, as cis regulatory DNA elements which increase basal levels of transcription and can mediate cell-specific transcription. They bind regulatory proteins, which, following DNA looping, are brought in close proximity with the target gene promoter to facilitate recruitment of the general RNAPII-associated machinery, leading to increased rates of transcription (1, 2). The role and importance of enhancers have been recognized for many years, but the massive advances in genome-wide RNA sequencing in recent years have shed new light on their actions. It is now clear that enhancers can act as transcriptional units and are transcribed to, and/or are associated with long non-coding (lncRNA) and enhancer RNAs (eRNAs), which share both similar and distinct characteristics (Table 1) (3, 4, 18, 19). This activity, together with the discovery that much of the genome is transcribed to lncRNAs, many of which affect gene expression and some directly associated with the chromatin, has opened up an entire new field in understanding gene regulation (5). Despite evidence regarding the common occurrence and crucial importance of these elements, functions have yet to be assigned to most lncRNAs and their distinct roles, as compared to eRNAs, is not yet clear. Furthermore, it is still not known exactly how these elements are activated, although key characteristics embedded in the regulatory DNA are beginning to emerge and may hold some clues.

Table 1.

Commonly used long non-coding RNA terminology

Term	Name in full	Features	Further information
LncRNA	Long non-coding RNA	>200 nt RNA that does not encode a protein, capped, with or without polyA; located in nucleus or cytosol. Multiple functions and subgroups.	(3-8)
LincRNA	Long intergenic non-coding RNA	As above, transcribed from genomic regions located between protein-coding genes.	(3-8)
eRNA	Enhancer RNA	Commonly 200-2000 nt, unstable, capped and no polyA; transcribed bidirectionally from central untranscribed region.	(9, 10)
ceRNA	Competing/competitive endogenous RNA	LncRNA of various origins that compete for the miRNAs and act as miRNA sponges.	(11, 12)
circRNA	Circular RNA	Produced by back-splicing circularization of several exons; very stable, located throughout the cell and sometimes secreted. Produced slowly and associated with fast rates of transcription.	(13, 14)
ciRNA	Circular intronic RNA	Derived from excised introns; formation depends on specific sequence elements; predominantly nuclear.	(13, 15-17)

The endocrine system is characterized by multiple hormonal signals and feedback mechanisms that converge on individual hormone-coding genes, most of which are expressed in a cell-specific or restricted manner. The pituitary gonadotropes form the center of the endocrine axis regulating reproduction. At puberty they are stimulated by the hypothalamic gonadotropin releasing hormone (GnRH) to synthesize luteinizing hormone (LH) and follicle stimulating hormone (FSH). These 2 gonadotropic hormones, comprising a common α subunit (encoded by Cga) and a hormone-specific β-subunit (encoded by Lhb or Fshb), then travel through the circulation to stimulate steroidogenesis and germ cell maturation. In the female mammal, these hormones are responsible for the estrous or menstrual cycle, involving regulation through feedback from the gonadal steroids and by activin/inhibin signaling pathways (20-22).

Given the unique function of gonadotropes: production of the gonadotropin hormones at specific times and required levels, and in response to diverse hormonal signals, multiple transcriptional enhancers and other regulatory elements would be expected to play key roles in both the cell-specific expression and hormonal responsiveness of many of these genes. In this review, after briefly introducing enhancer elements, eRNAs, and other regulatory lncRNAs, we will discuss what is known about how they direct the cell-specific and/or hormone responsiveness of some of the key gonadotropic genes. We will also examine the growing number of reports of lncRNAs whose levels in the pituitary have been found to be associated with activity of the reproductive axis, or development of gonadotrope-derived nonfunctioning adenomas, in which regulatory functions of certain lncRNAs also in normal gonadotrope function have been indicated.

Enhancers and Associated Long Non-Coding RNAs

Enhancer Regulatory Elements

Transcriptional enhancers comprise regulatory elements that can activate genes nearby or at long distances, and their characteristics have been reviewed extensively (eg, (1, 23-26)). DNA looping, as shaped by the boundaries of topologically associated domains (TADs, Figs. 1 and 2A), allows these elements (Fig. 2B) to interact physically with their target gene promoters (25, 27). The large number of putative enhancers that have been detected suggests that most genes can likely utilize several of these elements in distinct contexts, such as in diverse cell types and in response to different signals (Fig. 1) (9, 28). These elements can function additively or synergistically (29), while some are associated with corepressors and act as silencers to repress transcription of target genes (30). Enhancer-like elements are also sometimes clustered into super-enhancers (Fig. 2C) (31, 32), spanning extensive regions, several kilobase pairs (kb) in length, in which they bind numerous transcription factors (TFs) and work additively or cooperatively to enhance gene expression to direct cell lineages (31, 33). It has further been proposed that in this context, they function by driving a liquid-phase separation due to the concentration of proteins and possibly RNAs (33-36).

Figure 1.

Differential usage of enhancer elements. Multiple enhancers, activated by various signals, can be used in distinct contexts to regulate expression of genes within the same topologically associated domain (TAD).

Figure 2.

Regulatory enhancers and lncRNAs discussed in this review. CTCF-Cohesin complex (A) closes the DNA domain into a loop containing various transcription units, which allows the regulatory elements to interact physically with their target gene promoters. Transcriptional enhancers (B) located up- or downstream of their target genes, sometimes clustered as super-enhancers (C) or enriched within the first introns (D), bind tissue-specific TFs to enhance or repress transcription. Enhancers are characteristically marked by H3K4me1, unlike promoters marked by H3K4me3. The central nontranscribed regions of active enhancers often contain noncanonical structures like G-quadruplexes (E) and produce bidirectional eRNAs (F) or lncRNAs (G) which might interact with the DNA to form R-loops (H). The lncRNAs enriched with G-rich sequences can also form noncanonical DNA structures which are often bound by PRC2 complex (I). Other lncRNAs can form circRNAs (J) which are found in the nucleus and cytosol and sometimes secreted from the cell.

Although enhancers generally function independently of their location, frequently being located far up- or downstream from the transcriptional start site (TSS), they are often found within the first 1 kb of first introns (Fig. 2D) (37). These intronic enhancers were noted commonly to regulate genes with tissue-specific functions rather than housekeeping genes, which tend to utilize intergenic enhancers; moreover, their usage often changes during development (38). Intronic enhancer elements activate transcription strongly and are thought to affect only the gene in which they are located (37). The mechanisms through which these enhancers function are not well understood and presumably differ from those of distal regulatory elements, especially if close to the TSS; they might include trapping and recycling of unproductive RNA polymerase II (RNAPII) to the gene promoter, as proposed for an intronic enhancer of Fgf5 (33).

Beyond their typically open chromatin and association with transcription factors and cofactors, characteristics that are shared by active gene promoters, enhancers are usually marked by monomethylation of histone H3 at lysine 4 (H3K4me1), which contrasts with its trimethylation (H3K4me3) at protein-coding gene promoters (26, 39). Given that many enhancers are transcribed to eRNAs by RNAPII, it is not clear why H3K4me3 is not required, but presumably this relates to distinct mechanisms of recruitment of the transcription machinery, possibly due to characteristics determined by DNA sequence (28, 40-42). Notably, while protein-coding gene promoters often include dense regions of CpGs, forming CpG islands, enhancers rarely contain CpGs, and those that do, form a distinct class with unique features of eRNA transcription and transcription-factor binding (42). Analysis with machine learning of thousands of transcribed enhancers and promoters from different cell contexts was able to distinguish promoters from enhancers based on their sequence, and predicted enhancer activity, while also indicating functionally relevant differences in enhancer and promoter GC content beyond the influence of CpG islands (41).

The above study by Colbran (41) trained on counts of 6-bp-long sequences and focused on TF binding sites. However, some G-rich sequences form noncanonical DNA structures which are enriched at regulatory regions of the genome (Fig. 2E) (43, 44). At transcriptional enhancers, these structures appear to direct cell-specific gene expression and differentiation (44-46). G-quadruplexes (G4s) are formed from stretches of guanines on a single DNA strand, which, depending on precise sequence, fold into various parallel or anti-parallel secondary structures through Hoogsteen hydrogen base-pairing (47). On the C-rich complementary strand, an i-motif (iM) can form through stacking of intercalating hemiprotonated C-neutral C base pairs, although this formation is highly sensitive to ambient pH (45). G4 and possibly iM structures appear to serve as beacons for moderating transcription: both bind diverse proteins including chromatin architectural and modifying proteins like nucleolin, high-mobility group (HMG) proteins, DNA, and histone methyltransferases, as well as numerous helicases and transcription factors which mediate some of their effects (44, 47, 48).

The Bidirectional eRNAs

Bidirectional transcription from the central nontranscribed region of active enhancers characteristically produces divergent eRNAs (Fig. 2F) (9, 10, 24, 49-52). These eRNAs are typically relatively short (800-2000 nucleotides), unstable, and neither spliced nor poly-adenylated. Their increased transcription following the same stimulation that upregulates the target gene, together with effects observed on target gene expression after eRNA knockdown, may indicate a functional role for these eRNAs. Accordingly, transcription of one of the eRNAs may be more favorable than the other for activation of the target gene (53-55). It is not yet clear how this directionality is regulated, but it was reported to be determined at the region where the RNAPII complex assembles (53).

Functions for the eRNAs have been demonstrated to include inducing activating histone modifications, changes in chromatin conformation and stabilization of enhancer-promoter looping (51, 54, 56-58). The eRNAs have been shown to interact with various proteins (eg, mediator (56), CBP (55, 59), cohesin (51), NELF (60)) which mediate some of their functions. However, the diversity of eRNA sequences and structures suggests that eRNAs have multiple different roles and, in some cases, the act of transcription itself might also play a role in enhancer activity and/or chromatin organization (24, 61).

Regulatory lncRNAs: Distinct Processing Endows Unique Features

Asides from the bidirectional eRNAs, other lncRNAs are often found in close proximity to active enhancers and can regulate gene expression via various mechanisms (Fig. 2G-2J), including by mediating activity of the enhancer on its target gene (6, 7, 62-64). These lncRNAs are divided into various classes according to the regions of the genome from which they are transcribed and how they are processed (3, 5, 7, 8). The promoters of the regions encoding these RNAs are usually marked with H3K4me3, and the lncRNAs are often both capped and polyadenylated, providing them the stability that characterizes these transcripts (6). However, distinct processing of lncRNAs can endow them with unique features and functions (4, 6, 65). For example, stabilization can be achieved by unusual mechanisms, such as RNase P cleavage to generate mature 3′ ends that fold on themselves to form a triple-helical knot (66, 67), and capping by small nucleolar RNA-protein complexes at the 5′ or both ends of the RNA (68-70) directs export of the lncRNAs into the cytosol. The processing of lncRNAs appears to be determined, at least in part, by their sequence: in lncRNAs with long introns, the processing is often weaker and the editing inefficient, so the lncRNAs remain in the nucleus where they act in a regulatory capacity (3, 6). The retention of lncRNAs in the nucleus was proposed to be due to their binding U1 small nuclear ribonucleoprotein (snRNP), which is via a specific binding motif (71, 72).

In the cytosol, lncRNAs are found in various cell compartments, including mitochondria, exosomes, and bound to ribosomes or to distinct RNA binding proteins, for their various functions which include regulating gene expression posttranscriptionally (6, 73-75). Some lncRNA can function as competing endogenous RNAs (ceRNAs) by acting as sponges for miRNAs (11, 76, 77). These include pseudogenes and also circular RNAs (circRNAs) derived from back-splicing of several spliced exons, or from excised introns, termed circular intronic RNAs (ciRNAs) (12, 15, 78). CircRNAs (Fig. 2J) are highly stable and have multiple functions in the cell, as well as sometimes being secreted (12, 13, 79, 80). They are also found in the nucleus and have been reported to regulate RNAPII transcription and can interfere with gene splicing (81-83). The circRNAs and circular intronic RNAs regulate expression of their parent genes while also exerting effects on other genes in trans (16).

Roles of Chromatin-Associated lncRNAs

Nuclear lncRNAs are found both associated with the chromatin and in subnuclear compartments (4, 8, 73, 74). They affect nuclear and chromatin organization, regulate gene expression in cis or in trans, including through altering transcription factor activity, as well as via posttranscriptional regulation (6, 7, 8). At the chromatin, lncRNAs interact directly with various chromatin-modifying and chromatin-associated proteins to activate or repress transcription of nearby genes, and with CTCF proteins to mediate long distance interaction (7, 84, 85). Many lncRNAs interact with the polycomb complex 2 (PRC2, Fig. 2I), and this interaction was shown to facilitate gene expression by preventing the repressive PRC2-mediated methylation of H3K27 (86). However, other lncRNAs, such as XIST and HOTAIR repress gene expression through activating this same mechanism. The switch that determines whether PRC2 is activated or repressed by different lncRNAs is reportedly regulated by RNA-RNA bridging, which alters the lncRNA structure (87). Thus, the activity of PRC2, which preferentially binds to single-stranded RNA containing G-tracts and G-quadruplexes (88), is also determined by sequence through the potential for such bridging.

Some lncRNAs also interact with the DNA to form R-loops or RNA-DNA triplexes, which recruit protein complexes to regulate expression of nearby genes (Fig. 2H) (89). When R-loops form on the opposite strand from G-quadruplexes, they stabilize the structure and promote antisense transcription, and purportedly they can act as intrinsic Pol II promoters (89-92). RNA-DNA triplex formation was shown to mediate the effects of lncRNA MEG3 at distal elements on TGFβ pathway genes and was suggested to be a general characteristic of target gene recognition by chromatin-interacting lncRNAs (93). The bidirectional transcription at transcriptional enhancers might thus also be facilitated and directed by G4 structures and R-loops formed from chromatin-associated lncRNAs and eRNAs transcribed in these regions. The particularly high concentrations of lncRNAs and eRNAs at super-enhancers indicates that they might play additional roles, for example in the condensate formation that purportedly mediates super-enhancer activity. Thus, a positive feedback network might exist whereby the G4/iM promotes lncRNA transcription, resulting in stabilization of the G4/iM structures, while both the lncRNAs and the intrinsically disordered regions of proteins binding these structures contribute also to condensate formation (89-91).

Regulation of Key Gonadotrope Genes by Enhancers, eRNAs, and Regulatory lncRNAs

A Distal Enhancer and its eRNA Determine the Chromatin Landscape of Cga

The first distal gonadotropin gene enhancer whose function was shown to involve transcribed eRNAs was that of the chorionic gonadotropin alpha (Cga) subunit gene. Earlier studies showed that regulatory sequences positioned far upstream of the Cga gene control its cell-specific expression and were sufficient to drive Cga expression in gonadotropes (94). Deletion analysis in transgenic mice located a region between 4.6 and 3.7 kb upstream of the Cga gene as an enhancer responsible for high levels of gene expression specifically in gonadotropes and thyrotropes (95). This study revealed that the tissue-specific activity of the enhancer also requires a proximal region near the Cga promoter, implying direct protein-protein or protein-nucleic acid interactions.

Subsequently, this upstream regulatory region was seen to be enriched with histone marks typical of an enhancer (H3K4me1 and H3K27ac) and was found to drive bidirectional transcription to 2 eRNAs (54). The more distal eRNA, transcribed in the reverse direction from that of Cga, is responsible for chromatin remodeling at the enhancer and proximal promoter, involving looping of the DNA (Fig. 3). Knockdown of this distal eRNA led to reduction in activating H3K27 acetylation and its replacement with repressive H3K27 trimethylation at the enhancer and promoter (54). Promoter H3K4me3 was also reduced and the chromatin more compact following the eRNA knockdown, and the nucleosome depleted region was diminished, seemingly due to loss of the Chd1 chromatin remodeler (54, 97). The eRNA knockdown also led to a major decrease in Cga expression levels, which progressed over some weeks, reaching levels that were below 5% those in control cells (54). Thus, the Cga distal enhancer, through this eRNA, plays a crucial role in determining gene expression levels by shaping the chromatin landscape. Notably, however, the gene could still be stimulated by GnRH (54), suggesting the presence of additional, yet-to-be-identified, enhancers that might mediate this hormonal response.

Figure 3.

The Cga distal enhancer and its eRNA. A distal regulatory region located upstream the Cga gene is marked by typical histone marks of an enhancer (H3K4me1 and H4K27ac), and it drives bidirectional transcription of 2 eRNAs. The more distal eRNA (opposite) plays a role in the DNA looping which brings the enhancer close to the proximal promoter, to maintain the epigenetic landscape and high rates of Cga transcription (54, 96).

Fshb is Regulated in cis by a Distal Enhancer and Possibly in trans by miRNA-Sponging lncRNAs

A novel distal enhancer 26 kb upstream of the human Fshb gene was recently reported (98). This region comprises gonadotrope-specific open chromatin which was observed in a number of studies (eg, (99, 100)), strongly indicating its regulatory function. Moreover, genome-wide association studies (GWAS) revealed that this region contains a significant single nucleotide polymorphism (SNP), which is associated with polycystic ovary syndrome (101, 102). In mouse gonadotrope cell lines and even whole pituitaries, the region was seen to be enriched with typical enhancer histone modifications. This region was also found to contain several TF binding motifs that affected its activity in reporter assays in which a ~450 bp conserved sequence was placed upstream of the FSHB promoter. A functional binding site for Nr5a1 overlapped the SNP and was seen to play a role in enhancer activity in these reporter assays, while the SNP increased Nr5a1 binding and FSHB promoter activity (98). Although the reporter gene assays demonstrated that this element increases FSHB-activated transcription, similar effects were seen on other gene promoters, and it was concluded that its function is not promoter-specific, while its reversibility might be context-dependent (98). However, the study did not map the eRNAs/lncRNAs associated with the enhancer, and the DNA encoding these was presumably not included in the 450 bp sequence studied; these regions might well impart some of its target-specific function and could also explain the relevance of its orientation.

Several lncRNAs have been proposed to regulate Fshb expression, mostly by acting as “sponges” for various miRNAs. The lncRNA-m433s1 was seen to upregulate Fshb levels and FSH secretion in the rat pituitary by interacting with miR-433 and reducing its inhibition of Fshb mRNA (103). Another lncRNA, this time a circular RNA, circAkap17b, was seen to function similarly, by neutralizing miR-7 which also suppresses FSH secretion (104). Knockdown of this circRNA suppressed Fshb mRNA and FSH secretion, while its overexpression had the opposite effect (104). However, circAkap17b levels in the anterior pituitary were shown to drop quite dramatically during the first 2 weeks after birth (104), such that its function in regulating FSH might be limited. Additional circRNAs were found differentially expressed in pituitaries from immature and mature rats (see below) and were predicted to interact with various miRNAs reported to regulate Fshb (105), although their function has yet to be demonstrated.

Regulatory lncRNAs Upstream of the Lhb Gene?

Gene-specific transcriptional enhancers for Lhb have yet to be described, and the major regulatory elements required to drive Lhb expression have long been thought to be present in the proximal ~500 bp region upstream of the TSS, which is highly conserved across species (106-111). The genomic locus of the Lhb gene is complex in humans due to duplication events leading to multiple copies of the CGB genes, but immediately upstream of LHB there are several regions enriched with H3K4me1 (ENCODE data). This same region is reported in both human and mouse, to be transcribed to 1 or 2 lncRNAs, with only partial sequence similarity across species. The longer of the putative transcripts has an open reading frame (potentially encoding 310 [mouse] or 334 amino acids [human]), but it is annotated as a spliced lncRNA that is associated with the polycomb complex (112). As noted above, many lncRNAs interact with PRC2, but this association can lead to diverse outcomes depending on sequence and local context. Our previous studies showed that there is some H3K27me3 on the proximal Lhb promoter in αT3-1 cells (113, 114), although levels are considerably lower at the region encoding the 3′ end of this lncRNA (114). Work will be required to determine whether this and other lncRNAs regulate Lhb transcription, as well as identifying distal enhancer elements and eRNAs responsible for its tissue-specific and hormonally induced expression.

The Intronic Enhancers of NR5A1

The transcription factor NR5A1 (SF-1) is responsible for differentiation of the gonadotrope cell lineage, and its expression is restricted to steroidogenic tissues. In mice, Nr5a1 expression was shown to be regulated by several intronic and one upstream enhancer which work with the basal promoter to direct its cell-specific expression, and whose activities are regulated by DNA methylation (115, 116). A pituitary gonadotrope-specific enhancer which binds pituitary homeobox 2 (Pitx2), is located in the sixth intron of the gene (117), and is hypomethylated in gonadotropes, but hypermethylated in ventromedial hypothalamus neurons and in the adrenal cortex (116). Interestingly however, in nongonadotrope pituitary cells (ie, not Gnrhr-expressing cells or their progeny (118)), it was found to be less hypermethylated than in other tissues (116), perhaps underlying its potential expression in these cells which might lead to the transdifferentiation between distinct hormonal cell types and/or multihormonal pituitary cells that have been reported (119-121).

Studies in mouse gonadotrope-derived cell lines thought to represent gonadotropes at various stages of differentiation, found that methylation of this gonadotrope-specific intronic enhancer correlates with levels of gene expression, being unmethylated in LβT2 cells and heavily methylated in αT1-1 cells, while partially methylated in the αT3-1 gonadotrope precursor-derived cell line and the AtT20 corticotrope cell line (122). In these cells, an additional putative Nr5a1 enhancer was also identified in the fourth intron, which was suggested to play a transient role in early gonadotrope differentiation, and its temporary chromatin accessibility was evident in 12.5-14.5 d mouse embryonic pituitary cells (123). Activity of this enhancer in the gonadotrope cell lines was seen to be directed by estrogen receptor α (ERα)-mediated chromatin remodeling, and ERα protected the enhancer from repressive DNA methylation (123). Given that chromatin compaction, histone modifications and DNA methylation/hydroxymethylation can vary between model cell lines and primary cells (124, 125), genome editing approaches will be useful to confirm the functions of these elements in vivo.

Distal Regulation of the Gnrhr Gene Is Not Conserved

Several distal regulatory elements have been reported to drive expression the Gnrhr gene in the gonadotropes where it regulates all 3 gonadotropin genes, although some of these are highly species-specific (126). The rat Gnrhr gene is activated by Lhx2, Isl1, and Gata2 at a distal enhancer ~1 kb upstream of the TSS that interacts with the Nr5a1-bound proximal promoter (127). However, this element is not conserved in other species. The human GNRHR contains several silencer elements around ~1 to 1.7 kb upstream of the TSS that appear to function in a cell-specific manner, as well as more distal activating elements that are utilized differently in gonadotrope and placental cell lines (126, 128). In ovarian cancer cell lines, an upstream alternative promoter was also reported (129). The ability to utilize these diverse sites to enhance GNRHR transcription presumably underlies its aberrant expression in many cancers (130, 131), where misregulated open chromatin permits access to these potential regulatory elements.

Although its direct regulation by lncRNAs has yet to be elucidated, the GNRHR gene is regulated posttranscriptionally by miRNAs, which reportedly mediate up-regulation of GnRH protein by leptin (132). Thus “sponging” lncRNAs, including possibly circRNAs, might well play an indirect role to stimulate GnRHR expression by reducing effects of these repressive miRNAs. Elucidation of how GnRHR expression is controlled by these various, as yet elusive, lncRNAs is important for understanding its regulation in the pituitary gonadotrope and its aberrant expression in tumor cells There is clearly still much to explore in how diverse classes of lncRNAs regulate this gene.

Screening for Functional Pituitary lncRNAs Associated With Sexual Maturation and Reproductive Function

The identification of lncRNAs that are specific to gonadotrope function is hampered by the small numbers of these cells in the pituitary. One approach to identify them and determine their possible roles has been to examine lncRNA levels in whole pituitaries at various stages of sexual maturation and the estrous cycle. In one such study, RNA sequencing of the anterior pituitary of immature and mature rats, led to the identification of 7039 lncRNAs, including long intergenic noncoding RNAs (lincRNAs) (58.9%), antisense lncRNAs (12.7%), intronic lncRNAs (22.5%) and sense lncRNAs (5%), which together corresponded to 4442 lncRNA genes. Of these, 1181 transcripts were differentially expressed (DE), a similar number being up- and downregulated. Possible functions were inferred by searching for protein-coding genes within 100 kb of the lncRNAs, and 3 lncRNAs (MSTRG.80236.1, MSTRG.80236.2 and MSTRG.80236.3) that were predicted to interact with Fshb showed similarly increased expression (133). This group also screened for functional circRNAs, using a similar approach, and detected 32 DE circRNAs in the sexually immature and mature rat anterior pituitaries (105). They predicted interactions of some of the circRNAs and miRNAs, which lead to the discovery of several circRNAs with potential to regulate Fshb by acting as miRNA sponges as detailed in the previous section (“Fshb is Regulated in cis by a Distal Enhancer and Possibly in trans by miRNA-Sponging lncRNAs”) (104, 105).

Several comparable studies in sheep pituitaries were performed to explore the functions of lncRNAs in sexual maturation and reproductive function. Comparison of the lncRNAs in pituitaries of immature and mature rams revealed 2417 known lncRNAs and 1256 new lncRNAs, including 193 that were DE (134), while 1407 DE mRNAs were identified. For the DE protein-coding genes related to growth, reproduction, or steroid synthesis, interactions with the lncRNAs were predicted and networks constructed. Short-interfering RNA knockdown of one of these lncRNAs (TCONS_00066406) in sheep pituitary cells led to a decrease in expression of its predicted target Hsd17b12, as well as in Lhb and Fshb mRNA levels, although the specificity of this effect was not reported (134). The same group used strand-specific RNA-seq to profile lncRNAs and mRNAs in highly and poorly prolific sheep, which revealed 57 DE lncRNAs and 298 DE mRNAs (135). Coexpression networks of these lncRNAs and their putative target genes highlighted Smad2, and the inhibitory effect of one of the lncRNAs (MSTRG.259847.2) on Smad2 as well as Lhb, though not Fshb, mRNA levels was shown in cultured sheep pituitary cells (135).

The involvement of lncRNAs in seasonal estrus was also examined in sheep pituitaries and 995 lncRNAs were identified, of which 335 were DE during estrus and anestrus (136). Prediction of the target genes of these lncRNAs, together with the function of the encoded proteins in hormone synthesis and metabolism, led the authors to suggest possible functions for some of these lncRNAs in estrous (136). The same group profiled also the circRNAs expressed in estrus and anestrus sheep pituitary (137) and those expressed in prenatal and postnatal sheep pituitaries (138). Both studies revealed a large number of DE circRNAs, and the latter study predicted that some of these interact with pituitary-specific miRNAs (138).

These screens in the model animal pituitaries have produced large datasets of lncRNA and circRNAs that might play a role in pituitary and possibly gonadotrope function. However, this data needs to be integrated and the functions analyzed, and it is not yet clear how many of these RNAs are conserved across species. Accessing human pituitary and gonadotrope specific lncRNAs is obviously more challenging, although a useful resource and online tool is available in the long non-coding RNA Knowledgebase (lncRNAKB; http://www.lncrnakb.org/), which comprises a catalog of lncRNAs expressed in human pituitary as well as 30 other normal human tissues (139). The database contains coexpression modules that point to possible lncRNA functions, and expression quantitative trait loci with tissue-specific lncRNA-trait associations from 323 GWAS studies, and thus it should comprise a highly useful resource for study of lncRNAs particularly in the context of human disease.

Regulatory lncRNAs in Nonfunctioning Pituitary and Gonadotrope Adenomas

Altered expression of lncRNAs is likely to underlie diverse cell phenotypes and behavior including oncogenesis, and several studies have examined their differential expression and possible functions in gonadotrope or nonfunctioning adenomas (NFPAs) which are primarily of gonadotrope origin (140, 141). Analysis of the coexpression networks for lncRNAs and mRNAs in these tissues, compared with normal pituitary tissue, suggested that each lncRNA might have a large number of mRNA targets (142, 143) and exposed lncRNAs that appear to play a role in tumor formation or invasiveness, while some of their targets are known to play roles in normal gonadotrope function.

The maternally expressed gene 3 (MEG3) locus is characteristically strongly repressed in NFPAs. This region encodes multiple imprinted genes, which are expressed from only one of the alleles, depending on its parent of origin, the second allele being repressed by DNA methylation (144). MEG3 encodes a noncoding RNA expressed only from the maternal allele, and this locus also encodes 3 genes expressed only from the paternal allele, which include protein delta homolog 1 (DLK1). Both MEG3 and DLK1 are virtually silenced in NFPAs, though not in functioning adenomas (142, 143, 145, 146); this is associated with increased DNA methylation at an intergenic control region upstream of MEG3 and at the MEG3 promoter (146-148). DLK1 is expressed highly in the pituitary, and mutations on the paternal allele are associated with central precocious puberty (149), while Dlk1 knockout mice have reduced FSH and LH levels (150, 151). Although known to encode a receptor for Notch signaling, its precise function in the gonadotropes and a role in suppressing tumor formation or invasiveness have yet to be determined.

MEG3 is a particularly large lncRNA found as multiple isoforms, some of which induce cell cycle arrest and apoptosis via interactions with p53 protein (152, 153), but MEG3 can also target the chromatin directly. It was found that in breast cancer cells it binds, together with PRC2, to distal GA-rich regulatory elements involving RNA-DNA triplex formation. In this way it represses expression of various genes in the TGFβ/activin pathway, including SMAD2 in an apparent feedback loop (93). Thus, its silencing in NFPAs might activate these pathways. However, other studies found reduced SMAD3 expression in NFPAs compared to normal pituitaries (154-156), and one of these projected that this is due to a subset of miRNAs targeting the TGFβ signaling pathway (156). These findings point to diverse mechanisms through which MEG3 might affect TGFβ and activin signaling. Elucidation of its role in modulating these signaling pathways, which regulate gonadotropin gene expression and normal gonadotrope function (157-159), awaits further study.

Expression of another lncRNA, C5orf66-AS1, was also found reduced in NFPAs as compared to normal pituitary tissue, was lower in invasive adenomas compared with those that were noninvasive, and its expression levels negatively correlated with maximum tumor diameter (160). This spliced lncRNA is transcribed from a region immediately upstream of Pitx1 whose role in pituitary development and gene expression, including of gonadotropes, is well recognized (164-164). C5orf66-AS1 is predicted to interact with the Pitx1 proximal promoter and first intron, and these RNAs show similar patterns of expression which is in a limited number of tissues (Fig. 4). Moreover, they were differentially expressed in an additional cohort of NFPAs (160). Supporting a possible regulatory role, a correlation was also seen in head and neck carcinoma between increased methylation at this locus and lower expression of the lncRNA, with reduced Pitx1 expression in tumor tissue compared to surrounding normal tissue (166). The first exon of C5orf66-AS1 and the first intron of Pitx1 contain G-quadruplex forming sequences (predicted by pqsfinder), and the genomic region is enriched for PRC2 (ENCODE; Fig. 4). Together, these findings point to this lncRNA as a putative regulator of Pitx1, possibly via its interactions with the PRC2 complex, and whose activity might be controlled by DNA methylation in normal tissues as well as tumors.

Figure 4.

The expression of lncRNA C5orf66-AS1 and its connections with Pitx1. The lncRNA C5orf66-AS1 is expressed in the pituitary from the region located immediately upstream (~4.5 kb) of the Pitx1 gene (the lncRNA C5orf66 transcribed from the opposite strand is also expressed in the pituitary). Gene expression from GTEx RNA-seq (from UCSC Genome Browser: http://genome.ucsc.edu) shows a similar pattern of expression for Pitx1 and C5orf66-AS1. C5orf66-AS1 is predicted by GeneHancer to interact with 3 regions, one of which is at the Pitx1 proximal promoter and the second in Pitx1 first intron. The first exon of C5orf66-AS1 and the first intron of Pitx1 contain G-quadruplex forming sequences (as predicted by pqsfinder). ChIP seq data from ENCODE reports high occupancy of EZH2 and SUZ12 proteins (ie, the PRC2 complex) at this genomic region (165).

Additional screens of lncRNAs differentially expressed in gonadotrope adenomas compared with normal pituitaries revealed high levels of several lncRNAs that regulate expression of HMGA1 and HMGA2, whose roles in pituitary development are established (167, 168). The lncRNAs (ceRNAs) block the inhibitory effects of miRNAs that target the HMGA genes. RPSAP52, which is antisense to HMGA2, was found to be highly upregulated in gonadotrope and prolactin-secreting pituitary adenomas and its expression was correlated with that of HMGA2 in the tumors (169). This lncRNA was seen to regulate also other HMGA family members, and possibly plays a role in the cell cycle (169). Another study found that 2 pseudogenes HMGA1P6 and HMGA1P7 also act as ceRNAs, protecting HMGA mRNAs from miRNAs, and their expression also correlated with HMGA1 levels in tumors (170). Moreover, overexpression of these pseudogenes caused enhanced proliferation and migration of a tumor cell line, suggesting that they might contribute to the tumor behavior (170).

Several studies have looked at circRNAs in NFPAs, making comparisons between invasive vs noninvasive tumors, and in tumors that are recurrent vs at first surgery. Putative circRNA-miRNA networks were then produced to predict function (171-174). These large datasets on the differential expression of lncRNAs are not easy to interpret, both because lncRNAs have multiple targets mediated via numerous mechanisms, and because of the heterogeneity of NFPAs tissues. Single cell–RNA-seq should help decipher which lncRNAs are associated with the individual cell phenotype and behavior and lead to a better understanding of their function. More extensive work will then be required to understand their roles in regulating gene expression and tumorigenesis.

Outlook and Concluding Comments

The regulation of gonadotropin gene transcription has been studied for many years, promoter elements defined, and more recently the role of the chromatin addressed. Now, however, the field of regulatory biology has undergone a paradigm shift. RNAs are transcribed not only from classical enhancer elements, but potential regulatory regions have expanded almost infinitely to include massive inter- and intragenic regions of the genome that can function to regulate transcription. While the magnitude of this is clear, we have barely begun to understand the intricacies of these mechanisms and how they are implemented. Considerable work lies ahead to understand the mechanisms through which these diverse classes of enhancing and repressing lncRNAs are controlled and, in turn, how they regulate their target genes. This will lead to an understanding of hitherto unexplained aspects of gene expression and could well also provide novel targets for manipulating gene expression in health and disease.