E-proteins and ID-proteins: Helix-loop-helix partners in development and disease

Abstract

The basic Helix-Loop-Helix (bHLH) proteins represent a well-known class of transcriptional regulators. Many bHLH proteins act as heterodimers with members of a class of ubiquitous partners, the E-proteins. A widely-expressed class of inhibitory heterodimer partners- the Inhibitor of DNA-binding (ID) proteins- also exists. Genetic and molecular analyses in humans and in knockout mice implicate E-proteins and ID-proteins in a wide variety of diseases, belying the notion that they are non-specific partner proteins. Here, we explore relationships of E-proteins and ID-proteins to a variety of disease processes and highlight gaps in knowledge of disease mechanisms.

E-proteins and Id-proteins are widely-expressed transcriptional regulators with very general functions. They are implicated in diseases by evidence ranging from confirmed Mendelian inheritance, association studies, and mouse models that resemble human disorders. Here we briefly recap the properties of E-proteins and Id-proteins in development and differentiation, cell proliferation and cancer that have been gleaned from decades of study (Massari and Murre, 2000; Murre, 2005; Belle and Zhuang, 2014; Lasorella et al., 2014; Ling et al., 2014; Nair et al., 2014), before describing disease associations where the mechanisms of E-proteins and Id-proteins contributions is incompletely understood.

HLH proteins in differentiation

The helix-loop-helix domain was first recognized in dimeric transcription factors that regulate the IgG enhancer in B cells. Since these bind to Ephrussi-box (E-box) sequences (CANNTG), the first HLH proteins were named E-proteins (Murre et al., 1989). Mammalian E proteins include E12 and E47, which arise through alternative splicing of the E2A (HGNC:11633) gene, E2-2 (HGNC:11634), and HEB (HGNC:11623). Renaming these proteins TCF3, TCF4 and TCF12 respectively has been suggested, but caution must be used to avoid confusion since the acronym TCF had already been assigned to unrelated proteins (Table 1). There is a single E-protein in Drosophila, the Daughterless (Da) protein.

Table 1.

HLH protein nomenclature

Protein name	Old gene name	New gene name
E12, E47	E2A	TCF3
E2-2	E2-2	TCF4
HEB	HEB	TCF12

E proteins have many functions as heterodimer partners of the tissue-specific bHLH proteins that are sometimes called Class II bHLH proteins. The diverse Class II bHLH proteins include the myogenic transcription factors exemplified by MYOD, a factor sufficient to transform cultured fibroblasts into myoblasts, the conserved proneural proteins such as Achaete and Scute from Drosophila, and hematopoiesis protein T-cell acute lymphocytic leukemia 1 (TAL1) (Massari and Murre, 2000; Powell and Jarman, 2008). The paradigm that was soon established is that tissue-specific Class II bHLH factors whose specific expression defines specification and differentiation programs such as myogenesis or neurogenesis act by conferring DNA binding specificity and transcriptional activation on heterodimers with the ubiquitous E-proteins (Figure 1).

Figure 1. Transcriptional modalities of HLH proteins.

(A) Transcriptional activation of Drosophila proneural genes requires the ubiquitously expressed E protein (Da) forming heterodimers with tissue-specific bHLH proteins (e.g. Ac/Sc) and binds to promoter E boxes. (B) Transcription of myogenic genes is controlled by the binding of E protein-myogenic specific bHLH protein (MyoD) heterodimers to promoter E boxes. (C) E protein (or Da)-ID (or Emc) protein heterodimers fail to bind DNA and can not activate transcription.

Another class of pervasive HLH proteins act in opposition to E-proteins. These Inhibitor of DNA binding (ID) proteins lack a basic DNA binding region, so that heterodimers of ID proteins with E-proteins or with Class II bHLH proteins are unable to bind DNA or activate transcription (Benezra et al., 1990). ID protein levels may be low in differentiated cells, but are abundant in proliferating, multipotent cells including stem cell populations (Ling et al., 2014). Like the E-proteins, ID-proteins are represented by a single homolog in Drosophila, the extra macrochaetae (emc) gene, which was identified in parallel with mammalian ID proteins. emc encodes a negative regulator of proneural bHLH proteins including Achaete and Scute, and is a negative regulator of neurogenesis so that emc mutants exhibit ectopic neural differentiation (Ellis et al., 1990; Garrell and Modolell, 1990). Figure 1 summarizes the basic outline of HLH function in cell type specification and differentiation, in which specific expression of particular class II bHLH proteins must compete with ID proteins to heterodimerize with E proteins and form tissue- and cell-type-specific transcription factors that drive specification and differentiation. bHLH proteins are now known to regulate differentiation in too many additional situations to be listed here. For example, E-proteins and ID-proteins regulate Sertoli cell differentiation (Chaudhary et al., 2005), possibly through the class II bHLH proteins POD1/Epicardin (TCF21) and Scleraxis (Muir et al., 2006; Bhandari et al., 2011).

Expression of the only E protein, Da, in Drosophila seems to be ubiquitous, whereas particular mammalian E-protein and Id-protein genes have various overlapping patterns of expression. Id1 and Id3 in particular are often upregulated by BMP signals, and Id1 by cytokines (Ling et al., 2014). TGFβ,FGF and TCR signaling also regulate Id gene expression (Lasorella et al., 2014; Ling et al., 2014). Both Drosophila and mammals exhibit cross-talk between Id proteins and Notch signaling, consistent with their common regulation of bHLH-mediated differentiation (Baonza et al., 2000; Adam and Montell, 2004; Bhattacharya and Baker, 2009; Lasorella et al., 2014; Ling et al., 2014). It is increasingly recognized that Id genes can be transcriptional targets of E proteins, in Drosophila and in mammals. This makes Id proteins feedback inhibitors of E protein activity, which may be particularly important as a brake on E protein autoregulation (Bhattacharya and Baker, 2011; Schmitz et al., 2012) (Figure 2). Id proteins are also regulated at the level of protein stability (Lingbeck et al., 2005; Lasorella et al., 2006; Kim et al., 2009; Williams et al., 2011b; Lasorella et al., 2014).

Figure 2. Autoregulation and feedback inhibition in normal and disease conditions.

(A) Emc/ID expression requires Da activity. Once Emc/ID level is upregulated, it will in turn block autoregulation of da and Da activity (Bhattacharya and Baker, 2011). This regulation may ensure the steady state of E protein and ID protein levels in normal progenitor cells. (B) ID proteins are transcriptional targets of E proteins in Burkitt lymphoma cells. Both gain of function mutations of E protein (indicated by+symbol) and loss of function mutations of ID protein (indicated by X symbol) found in Burkitt lymphoma interrupt the feedback network between E and ID proteins. Cyclin D3 is also a target gene of E proteins in Burkitt lymphoma cells, and may contribute to the excessive proliferation of tumor cells (Schmitz et al., 2012).

Since mouse knock outs of each of the E-proteins and Id proteins have mutant phenotypes, neither E-proteins nor Id-proteins can be completely redundant. The example of Id1 and Id3 double null (Id1^−/−; Id3^−/−) mice, which exhibit premature neuronal differentiation, forebrain vascular abnormalities, cardiac defects and embryonic lethality not seen in the single mutants (Lyden et al., 1999; Lasorella et al., 2014), illustrates significant functional overlap that may also occur for other Id-proteins, and similarly between E-proteins (Table 2). Non-overlapping expression accounts for some non-redundancy, although unique protein functions may also exist, especially in the case of E-proteins that differ in basic region sequence and DNA-binding preference. It is thought that any unique functions of individual proteins are superimposed on the overall ratio of total E-proteins to total Id-proteins within a cell that is an important index of competence in differentiation and other processes. For example, elevating Da levels, which are controlled by feedback through Emc, affect the capacity of Ato or Sc to induce neurons in Drosophila (Bhattacharya and Baker, 2011), and elevated E protein levels are thought to act similarly in the postnatal mouse brain (Fischer et al., 2014).

Table 2.

Phenotypes of HLH protein knockout mice

Genotype	Phenotype	References
E2A−/−	Disrupted B cell development; lymphopoietic developmental defects with high frequency of T-cell lymphomas	Bain et al., 1994; Bain et al., 1997; Yan et al., 1997
HEB−/−	Reduced pro-B cells; Developmental arrest of thymocyte development;	Zhuang et al., 1996
E2-2−/−	Reduced pro-B cells; Normal mammary gland development; abnormal hindbrain development in the pontine nucleus;	Zhuang et al., 1996; Flora et al., 2007; Itahana et al., 2008
Id1−/−	Altered lipid and glucose metabolism; osteoporotic phenotype; normal mammary gland development; impaired cancer stem cell properties in glioma	de Candia et al., 2004; Chan et al., 2009; Satyanarayana et al., 2012; Lasorella et al., 2014
Id2−/−	Altered lipid and glucose metabolism; hydronephrosis (smooth muscle hypertrophy at the ureteropelvic junction); Parkinson’s disease-like features (impaired dopaminergic system); reduced proliferation in mammary epithelia (lactation defect); reduced body size; absence of lymph nodes and Peyer’s patches; absence of NK cells and Langerhans cells	Aoki et al., 2004; de Candia et al., 2004; Havrda et al., 2008; Park et al., 2008; Hou et al., 2009; Mathew et al., 2013; Tripathi et al., 2012; Havrda et al., 2013; Lasorella et al., 2014
Id3−/−	symptoms of Sjögren’s syndrome (impaired tear and saliva secretion due to defective autoimmune system); normal mammary gland development	de Candia et al., 2004; Li et al., 2004
Id4−/−	premature neuronal differentiation; altered lipid and glucose metabolism; reduce osteoblast differentiation; enhance adipocyte differentiation; reduced proliferation and increased apoptosis in mammary cells	Dong et al., 2011; Murad et al., 2010; Tokuzawa et al., 2010; Lasorella et al., 2014;
Id1−/− Id3−/−	premature neuronal differentiation; cardiac developmental defects; brain vascular abnormalities; embryonic lethality	Lyden et al., 1999; Lasorella et al., 2014
Id1−/+ Id3−/−	decreases of proliferation and mineralization in osteoblasts; vascular defects and growth failure in tumor cell xenografts genetic models of cancer	Maeda et al., 2004; Lasorella et al., 2014
Id3−/− ApoE−/− or Id3−/− Ldlr−/−	Enhanced atherosclerosis	Doran et al., 2010; Lipinski et al., 2012

HLH Proteins in the cell cycle and senescence

The commitment of progenitor cells to begin cell differentiation is often associated with cell cycle withdrawal. MYOD provided the first example connecting bHLH proteins to cell cycle arrest, inhibiting cell cycling as myoblasts differentiate (Gu et al., 1993; Halevy et al., 1995; Guo et al., 1995; Zhang et al., 1999). Subsequent studies have uncovered a common relationship between E-and ID-proteins and cell cycle progression. In many cells, E12 or E47 homodimers activate transcription of the CDK inhibitors (CDKIs) p15, p16 (also known as INK4a), p21, p27 and p57 (Sloan et al., 1996; Prabhu et al., 1997; Pagliuca et al., 2000)(Figure 3). Through this mechanism, E proteins inhibit cell cycle progression, and are also implicated in cellular senescence (Ling et al., 2014). In fact E47 is necessary for senescence of human fibroblasts, which is prevented by E-protein knockdown (Zheng et al., 2004).

Figure 3. Mammalian HLH proteins and cell cycle control.

Cell cycle regulation involving HLH proteins in multiple ways. E protein-mediated activation of the CDK inhibitors p16, p21, p27 and p57 is shown. Cell cycle arrest is promoted by these CDK inhibitors in G1-S transition, S and G2-M phases. This regulation is negatively modulated by ID proteins through the association with E proteins. Rb can antagonize ID2 by direct interaction.

As one might expect, the role of E proteins is blocked by ID proteins (Peverali et al., 1994). ID proteins can promote RB inactivation and activate E2F-mediated transcription and cell-cycle progression indirectly because of their inhibition of E proteins and consequently inhibition of CDKIs’ expression, while ID2 can be sequestered by RB directly (Lasorella et al., 2014; Ling et al., 2014) (Figure 3). ID1 and ID3 are degraded during senescence (Kong et al., 2011). ID proteins regulate proliferation in many important tissues, for example during the expansion of mammary epithelia during pregnancy (de Candia et al., 2004). Because of their dual effects on differentiation and proliferation, Id protein expression is associated with stem cell maintenance, and, in tumors, anaplasia (Nam and Benezra, 2009; Anido et al., 2010; Barrett et al., 2012; Niola et al., 2012; Lasorella et al., 2014; Ling et al., 2014; Nair et al., 2014).

Although this general relationship is seen in many cells there are examples of different effects. Id1 is reported to protect mammary epithelia cells from senescence independently of CDKI expression (Swarbrick et al., 2008), and rapidly-dividing transit amplifying cells surprisingly express less Id1 than their quiescent neural stem cell progenitors (Nam and Benezra, 2009). The poor growth of Drosophila cells lacking emc is also due to Da hyperactivity, not attributed to CDKI expression but to repression of the phosphatase Cdc25/string or to overexpression of the gene expanded within the Salvador-Warts-Hippo pathway of tumor suppressors (Bhattacharya and Baker, 2011; Andrade-Zapata and Baonza, 2014; Wang and Baker, 2015) (Figure 4). It is also envisaged that ID functions independent of dimerization with HLH proteins may exist (Ling et al, 2014).

Figure 4. Drosophila HLH proteins in growth control.

When Drosophila ID protein (emc) is mutated, excess Drosophila E protein (Da) homodimers activate expanded (ex) transcription via binding to E boxes in the cis-regulatory element. The subsequent activation of Salvador-Warts-Hippo signaling pathway inhibits Yki/YAP activity, thereby blocking cell growth and survival. High levels of Da are also involved in blocking cell cycle progression in G2-M transition through inhibiting stg/cdc25 transcription (Wang and Baker, 2015).

E-proteins and ID-proteins in cancer

There is a well-established connection between cancer and E-proteins (Murre, 2005; Belle and Zhuang, 2014). In addition to lymphoma predisposition in E2A mutant mice, E2A mutations have been found in many human lymphomas and leukemias (Murre, 2005; Tijchon et al., 2013; Belle and Zhuang, 2014; Cozen et al., 2014). As would be expected, ID proteins mostly have the opposite properties (Lasorella et al., 2014; Nair et al., 2014). ID levels are elevated in human tumors ranging from melanoma to neuroblastoma (Perk et al., 2005). ID1 and ID3 are clearly implicated in lung cancer initiation, and ID4 is overexpressed by a t(6;14)(p22;q32) chromosomal translocation in leukemia and amplified in ovarian cancer, qualifying it as a proto-oncogene (Perk et al., 2005; Lasorella et al., 2014; Nair et al., 2014).

No doubt many oncogenic and tumor suppressor roles of ID proteins and E proteins reflect their roles in cell proliferation, stemness, and senescence, but there are also contributions to metastasis and angiogenesis(Slattery et al., 2008; Lasorella et al., 2014; Ling et al., 2014; Nair et al., 2014). ID proteins are required to maintain glioma stem cells in the perivascular niche by inhibiting E-protein transcription of the RAP1GAP gene, since RAP1 controls cancer stem cell adhesion to the niche through integrin signaling (Niola et al., 2013). Unexpectedly, ID3 and ID4 sometimes behave as tumor suppressors (Yu et al., 2005; Chen et al., 2011; Love et al., 2012; Richter et al., 2012; Schmitz et al., 2012). For example, recurrent mutations in the E47 (TCF3) and ID3 genes occur in Burkitt’s lymphoma and may allow unrestrained E47 expression to activate both cyclin D3 and PI3K survival signaling (Love et al., 2012; Richter et al., 2012; Schmitz et al., 2012) (Figure 2B).

Post-mitotic ID functions: Neuronal morphogenesis

In addition to roles in cell fate specification and in the regulation of the cell cycle and senescence, in postmitotic neurons ID1, ID2 and ID4 are ubiquitinated by the Anaphase Promoting Complex/Cyclosome (APC/C) and their subsequent turnover affects axon and dendrite development (Lasorella et al., 2006; Kim et al., 2009). Mutations of the APC/C interaction motif of ID2 enhance axonal growth in mouse cerebellar granule neurons both in vitro and in the cerebellar cortex, suggesting that ID2 degradation maintains axon morphology by restraining E protein-dependent axon growth (Lasorella et al., 2006). On the other hand, knockdown of ID1 in the cerebellar cortex or primary hippocampal neurons stimulates dendrite morphogenesis, indicating an inhibitory role of ID1 in dendrite development (Kim et al., 2009). APC/C is important in cell cycle exit (Peters, 2006), and therefore might couple cell cycle exit to dendrogenesis in differentiating neurons. It remains unknown at present whether ID-protein downregulation contributes to terminal cell cycle withdrawal in neurons, and how E proteins affect dendrites. These questions are of considerable interest, as neurological disorders are associated with ID- and E-proteins (see below).

Beyond developmental defects and cancer: Connections to disease

E-proteins and ID-proteins have been implicated in a number of human diseases, which may be related to the roles of E- and ID-proteins in differentiation and cycle regulation, or perhaps reflect other mechanisms (Table 3). It is interesting to catalogue these potential disease relationships because their wide range argues persuasively that E- and ID-proteins represent central regulators active in most or all cells, rather than ubiquitous heterodimer partners of little intrinsic interest, and also so that gaps in knowledge can be identified. The following summaries are ordered according to the strength of the evidence linking to human disease, beginning with unambiguous Mendelian genetic disease in humans and progressing through other human data to mouse models that are suggestive but for which direct evidence in human disease is wanting.

Table 3.

HLH protein diseases and mechanisms

Disease	Affected organ or tissue	HLH protein	evidence	Process affected
Pitt-Hopkins syndrome	Brain, face	E2-2	Human mutation	Differentiation? Cell cycle? Neuronal morphogenesis?
Fuchs corneal dystrophy	eye	E2-2	Human gene association	senescence?
schizophrenia	brain	E2-2	Human gene association	Differentiation? Cell cycle? Neuronal morphogenesis?
Rett syndrome	brain	E2A, ID1-4	Gene expression	Differentiation? Cell cycle? Neuronal morphogenesis?
atherosclerosis	arteries	E2A, ID3	Human, mouse gene association	Differentiation?
Diamond Blackfan anemia	Bone marrow	E2A, E2-2, ID2	Gene expression	Differentiation?
Diabetes, glucose and lipid metabolism	multiple	ID1–2,4	Mouse mutant models	Differentiation?
Osteoporosis	bone	ID1, ID4	Mouse mutant models	Differentiation?
Arterial vascular disease	arteries	ID1, ID3	Mouse mutant models	Cell cycle?
Congenital hydronephrosis	kidney	ID2	Mouse mutant models	Differentiation?
Polycystic kidney disease	Kidney	E2A, ID2	Gene expression	Cell cycle?
Parkinson’s disease	Brain, nerves	ID2	Mouse mutant models	Differentiation?
Sjogren’s syndrome	Lachrymal, salivary glands, skin	ID3	Mouse mutant models	Differentiation?

Pitt-Hopkins Syndrome

Haploinsufficiency for the E protein transcription factor E2-2 (TCF4) due to heterozygous mutations or deletions is responsible for Pitt-Hopkins Syndrome (PTHS), an autosomal dominant disorder characterized by severe intellectual disability, global developmental delay, recurrent seizure and hyperventilation (Zweier et al., 2007; Amiel et al., 2007; Giurgea et al., 2008; de Pontual L. et al., 2009). Most of the mutations found in PTHS patients disrupt E2-2 transcriptional activity (Forrest et al., 2012; Sepp et al., 2012). In the developing and mature central nervous system, E2-2 is a heterodimer partner for Class II bHLH proneural and neuronal precursor proteins. Hence, heterozygous loss of function affects a large number of target genes, including the neuronal adhesion receptor genes contactin associated protein-like 2 (CNTNAP2) and neurexin 1 (NRXN1) (Forrest et al., 2012). Both CNTNAP2 and NRXN1 mutations have been associated with PTHS-like disorder, schizophrenia and autism (Alarcon et al., 2008; Zweier et al., 2009; Kirov et al., 2009). PTHS could reflect defective expression of such neuronal differentiation genes that depend on E2-2/proneural gene heterodimers, defective maintenance of cell cycle withdrawal (since unscheduled cell cycle re-entry can be associated with defective neuronal differentiation (Ruggiero et al., 2012)) or be related to the roles of ID1 and ID2 in dendrite and axon morphogenesis, or multiple of these mechanisms. It is also noteworthy that a particular E2-2 mutation has been found in Rett Syndrome (see below), suggesting that related molecular mechanisms lead to PTHS and Rett Syndrome.

Fuchs corneal dystrophy

Multiple SNPs in E2-2 are associated with Fuchs corneal dystrophy (Baratz et al., 2010; Li et al., 2011; Thalamuthu et al., 2011; Eghrari et al., 2012; Stamler et al., 2013; Lau et al., 2014), a common, dominant, progressive, late onset disease in which endothelial cells are gradually lost from the internal surface of the cornea. The hallmark of this disease is the increasing density of tiny bumps, termed guttae, that form on the cornea, loss of the fluid-pumping function of the endothelium cells, decrease in corneal transparency and resultant loss of vision. The lack of Fuchs corneal dystrophy in PTHS patients that have heterozygous loss-of-function E2-2 mutations, suggests that Fuchs corneal dystrophy might involve E2-2 gain-of-function. Interestingly, endothelial samples from Fuchs corneal dystrophy patients exhibit reduced ID1 expression, which could be another route to elevate E2-2 activity (Matthaei et al., 2014)(Figure 1). Since corneal endothelium is normally non-proliferative, cell cycle arrest is unlikely to cause disease, but altered expression of senescence-related genes has been reported (Matthaei et al., 2014). In Drosophila, elevated Da expression can be toxic and activates the Salvador-Warts-Hippo pathway of tumor suppressors (Wang and Baker, 2015), but it is not known whether this occurs in Fuchs corneal dystrophy. A different view is suggested, however, by studies of SNP rs613872 associated with an intronic trinucleotide repeat expansion (Breschel et al., 1997; Wieben et al., 2012). This expansion causes mis-splicing and RNA toxicity, similar to other trinucleotide expansion disorders such as myotonic dystrophy type 1 and 2, fragile X syndrome, and frontotemporal dementia (Du et al., 2015). RNA toxicity affects expression of many other genes, and the effects of this SNP might have little to do with E2-2 function.

Schizophrenia

In addition to linkage to PTHS and Fuchs corneal dystrophy, several SNPs in E2- 2 are significantly associated with schizophrenia (Stefansson et al., 2009; Forrest et al., 2014). E2-2 is also a target of micro-RNA 137, which has been mapped as a schizophrenia susceptibility locus (Schizophrenia Psychiatric Genome-Wide Association Study (GWAS) Consortium, 2011; Wright et al., 2013), and predicted to bind other miRNAs associated with schizophrenia, autism and other CNS disorders (Perkins et al., 2007; Talebizadeh et al., 2008). However, none of the variants affect the E2-2 protein, so how E2-2 activity is affected in schizophrenia remains uncertain (Williams et al., 2011a). This has not been resolved by mouse studies. E2-2 homozygous null (E2-2^−/−) mice displayed abnormal hindbrain development in the pontine nucleus (Flora et al., 2007) (Table 2). Pontine nucleus development requires heterodimers of E2-2 with Atoh-1 (Flora et al., 2007). Thus, schizophrenia could have a developmental origin caused by reduction of E2-2 function. On the other hand, E2-2 overexpressing transgenic mice have defects in fear conditioning and sensorimotor gating (Brzozka et al., 2010). Sensorimotor gating has been used for a long time in animal model studies because it resembles the oversensitivity to sensory stimulation seen in schizophrenic patients (Braff and Geyer, 1990). Clearly, it would be very informative to resolve how schizophrenia-associated SNPs affect E2-2 function, and how they differ from those associated with Fuchs Corneal Dystrophy.

Rett Syndrome

Rett syndrome is an X-linked dominant disorder primarily caused by loss-of-function mutations in the methyl-CpG-binding protein 2 (MeCP2) gene, which lead to intellectual disability including language deficits, epileptic seizures and respiratory dysfunction (Bedogni et al., 2014). Microarray and chromatin immunoprecipitation analyses identified ID1, ID2, ID3 and ID4 genes as primary targets of MeCP2 (Peddada et al., 2006). MeCP2 is a repressor, and upregulation of all four IDs has been found in MeCP2 null mouse brain (MeCP2-/Y hemizygotes) as well as brain tissue from Rett Syndrome patients (Peddada et al., 2006). The MeCP2 null mutant mouse also has reduced expression of the neural precursor gene NeuroD1 (Peddada et al., 2006). Since E47 stimulates NeuroD1 transcription (Sharma et al., 1999), high ID-protein levels could reduce NeuroD1 expression. Interestingly, a frameshift mutation of E2-2 was recently discovered in a variant Rett syndrome patient who did not have mutations in MeCP2 (Armani et al., 2012). E2-2 loss of function is expected to mimic effects of ID-protein overexpression. Other E2-2 loss of function mutations cause Pitt-Hopkins syndrome (see above), which shares symptoms with Rett syndrome. Although the cellular basis of Rett Syndrome is not yet clear, altered neural differentiation as a consequence of reduced NeuroD expression, defects in neuronal cell cycle withdrawal, or altered axon or dendrite morphogenesis are all plausible candidates.

A further connection between HLH proteins and autism spectrum disorders is that ID3 has been reported as a target of miR-29b, a brain-specific miRNA that is expressed more highly in autism spectrum disorders. ID3 expression is significantly downregulated in miR-29b-transfected cells (Sarachana et al., 2010). Whether ID3 is really involved in autism, whether its function in this regard is related to that of E2-2, and if differentiation, cell cycle, neuronal morphology or other defects are involved remains to be elucidated. Intriguingly, one trait that is strongly associated with autism is circadian rhythm dysfunction (Hu et al., 2009). ID proteins play essential roles in controlling circadian rhythm by sequestering the bHLH transcription factors CLOCK and BMAL that are central components of the circadian pacemaker (Duffield et al., 2009).

Atherosclerosis

ID3 maps to atherosclerosis susceptibility loci in both humans and mice (Lusis et al., 2004). A single-nucleotide polymorphism (SNP rs11574) substitutes Thr for Ala105 in the human ID3 protein and is associated with increased carotid intima-media thickness (IMT) (Doran et al., 2010). This mutation decreases ID3 function as measured by binding to E12 in NIH 3T3 cells (Doran et al., 2010).

Atherosclerosis is strongly enhanced when Id3 is deleted in ApoE or Ldlr knockout mice (ie in Id3^−/− ApoE^−/− or Id3^−/− Ldlr^−/− genotypes) (Doran et al., 2010; Lipinski et al., 2012) (Table 2). Bone marrow transplantation from Id3^+/+ ApoE^−/− donors is atheroprotective, pointing to a role in bone marrow-derived cell types (Doran et al., 2012). These mice have reduced B-cell numbers but increased macrophage numbers in the aorta (Doran et al., 2012; Lipinski et al., 2012). Chemokines, and vascular cell adhesion molecule 1 (VCAM-1) are elevated, and may promote these changes in leukocyte adhesion and recruitment into the arterial wall. The VCAM-1 promoter is regulated by E12, and could be activated by E12 following a reduction in ID3 (Lipinski et al., 2012). One model, therefore, is that ID3 protects against atherosclerosis by altering leukocyte recruitment to artery walls (Doran et al., 2010; Lipinski et al., 2012; Doran et al., 2012).

The effect of Id proteins on the vasculature is complicated, however (Yang et al., 2014). ID3 inhibits adiponectin expression by preventing E47 from binding to E-boxes in the adiponectin promoter (Doran et al., 2008), Adiponectin is an adipocyte-derived cytokine that attenuates plaque formation: low adiponectin levels are associated with coronary artery disease, whereas adiponectin over-expression reduced atherosclerosis in ApoE-deficient mice. Atherosclerosis might also be impacted by the roles of Id proteins in vascular smooth muscle cells and endothelial cells (see below). For example, alternative splicing of Id3 in vascular lesions may produce a protein with a uniquely atheroprotective role (Forrest et al., 2004).

Arterial Vascular Disease

Mutations in the bmpr2 gene that encodes the BMP type II receptor are found in most cases of heritable pulmonary arterial hypertension (PAH), in which smooth muscle cells proliferate to occlude the vessel (Lane et al., 2000; Machado et al., 2006). ID1 and ID3 are targets of BMP signaling in pulmonary artery smooth muscle cells, as in other cell types, and a recent study suggests that Id proteins mediate anti-proliferative effect of BMP signaling (Yang et al., 2013). The mechanism is unclear, however, because both Id2 and Id3 are also reported to inhibit p21 expression in proliferating vascular smooth muscle cells, as in many other cell types (Matsumura et al., 2002; Forrest et al., 2004; Taylor et al., 2006)(Figure 3). Id proteins also have roles in endothelial cells (Ling et al., 2014; Yang et al., 2014). Id1^−/−;Id3^−/− mice exhibit cardiac developmental defects, brain vascular abnormalities, and embryonic lethality (Lyden et al., 1999) (Table 2).

Bone Marrow Failure

A connection between HLH proteins and bone marrow failure has been suggested based only on ID2 upregulation and E47 and HEB downregulation in Diamond Blackfan Anemia (DBA) patients (Zhang et al., 1997). E47, TCF4, and a tissue-restricted expressed bHLH protein, SCL, regulate early erythroid differentiation and are reduced during maturation when ID2 is upregulated, so that elevated ID2 would be expected to decrease the expression of erythroid specific genes and erythroid differentiation. This could be re-evaluated now that it is known that approximately 65% of patients harbor mutations in ribosomal protein genes (Farrar et al., 2011), and that many aspects of DBA have been attributed to p53 hyperactivity caused by nucleolar stress (Ellis, 2014). Although p53 can regulate ID1 and ID2 transcription in neural stem cells, repression is observed, contrary to the notion that ID2 is upregulated in DBA patients (Paolella et al., 2011). It is not impossible that differentiation defects can contribute to DBA, however, as suggested by mutations in the hematopoietic transcription factor GATA1 that were recently found in some DBA cases (Sankaran et al., 2012).

Diabetes, glucose and lipid metabolism

Insulin affects glucose and lipid metabolism and both are disrupted in diabetes. β-cells in the islets of Langerhans of diabetic mice, or isolated β-cells exposed to hyperglycemia or lipid, have elevated ID1/ID3 expression (Wice et al., 2001; Busch et al., 2002; Kjorholt et al., 2005; Billestrup, 2011). Consistent with the notion that ID proteins impair glucose and lipid metabolism, islets from Id1^−/− mice are protected against high-fat diet-induced repression of β-cell genes, such as pancreatic duodenal homeobox-1, NeuroD1 (also called Beta2), Glut2, pyruvate carboxylase, and Gpr40 (Akerfeldt and Laybutt, 2011). Id1^−/− mice show less insulin resistance in skeletal muscle, liver and white adipose tissue, and increased Uncoupling Protein 1 and Peroxisome proliferator-activated receptor gamma coactivator 1 in brown adipose tissue (corresponding to increased energy expenditure) (Satyanarayana et al., 2012). Mice over-expressing E-proteins revealed similar phenotypes (Zhao et al., 2014).

Although Id1 expression is down-regulated and the protein apparently degraded during adipocyte differentiation, adipogenesis appears normal in Id1^−/− mice, suggesting that it is effects on oxygen consumption and thermogenesis that protect them against a high-fat diet (Satyanarayana et al., 2012). E2A and E2-2 proteins can also directly increase insulin expression in β-cells (Vierra and Nelson, 1995).

Other ID proteins also regulate metabolism. Mice with homozygous deletions of Id2 or Id4 have less body fat and gain much less weight on a high-fat diet (Murad et al., 2010; Mathew et al., 2013). Id2^−/− mice have altered lipid metabolism, impaired adipogenesis, and reduced gonadal white adipose deposits and liver lipid content (Park et al., 2008; Hou et al., 2009; Mathew et al., 2013) (Table 2). Male Id2^−/− mice also reveal increased insulin sensitivity, increased glucose uptake by skeletal muscle and brown adipose tissue, and reduced intramuscular triacylglycerol and diacylglycerol levels (Mathew et al., 2013). Apparently, therefore, ID1, 2, and 4 play a variety of roles in insulin secretion and energy metabolism.

Osteopenia and osteoporosis

Id1^−/− mice have an osteoporotic phenotype ie reduced bone density and increased bone fragility (Chan et al., 2009) (Table 2). Healthy bone volume is controlled by the balance between bone deposition by osteoblasts and bone resorption by osteoclasts. Excess of bone resorption over bone formation results in osteoporosis and its milder potential precursor, osteopenia. Osteoclasts and osteoblasts are derived from distinct lineages: osteoclasts differentiate from hematopoietic stem cell (HSC) precursors (Suda et al., 1992), while osteoblasts arise from mesenchymal stem cells (MSCs). The bone phenotype of Id1^−/− is due to excess osteoclastogenesis (Chan et al., 2009). Genes required for osteoclastogenesis are up-regulated including Trap, Oscar and Ctsk; over-expressing ID1 downregulates these genes (Chan et al., 2009). It would be reasonable to anticipate a positive role of one or more E-proteins in osteoclast differentiation and maturation, although none seems to have been reported.

Osteoporosis and age-related osteopenia also result from differentiation of MSCs into bone marrow adipocytes rather than osteoblasts. Adipocytes additionally suppress osteogenesis because secreted adipogenic signals influence osteoclast differentiation by HSCs. Id4^−/− mice reduce osteoblast differentiation and enhance adipocyte differentiation (Tokuzawa et al., 2010) (Table 2). ID4 competes with the bHLH factor Hes1 to dimerize with Hey2, allowing Hes1 to activate osteoblast-specific gene transcription and to promote osteogenesis through the key osteoblast differentiation factor Runx2 (Tokuzawa et al., 2010). It is not known how Hes1 regulates Runx2 levels (Ikawa et al., 2006). Other ID proteins also affect osteoblasts, which show reduced proliferation and mineralization in Id1^+/−; Id3^−/− mice (Maeda et al., 2004). Taken together, these findings make both ID- and perhaps E-proteins potential targets for the prevention and treatment of osteoporosis, because of their roles in osteoclast and osteoblast specification.

Renal disease

Id2^+/− or Id2^−/− mice develop congenital hydronephrosis, which is common in humans and may result in early-onset renal failure (Aoki et al., 2004; Tripathi et al., 2012). The cause of hydronephrosis is frequently obstruction at the ureteropelvic junction. Id2 mutant mice display irregular, hypertophic muscle layers at the ureteropelvic junction, suggesting that smooth muscle hypertrophy at the ureteropelvic junction could be the cause of hydronephrosis (Aoki et al., 2004). E protein-dependent myogenesis could be stimulated when ID2 levels are decreased, but the usual role of E-proteins is to inhibit proliferation, not promote it (Figure 1 and 3).

A role for ID proteins was also suggested in autosomal dominant polycystic kidney disease (PKD), which is mostly caused by mutations in the PKD1 and PKD2 genes (Peters and Sandkuijl, 1992). In this disease bilateral renal cysts replace the normal renal parenchyma, often resulting in end-stage renal disease. Cysts are lined by a single layer of epithelium that is characterized by increased cellular proliferation and decreased differentiation (Nadasdy et al., 1995). It was reported that Polycystin-2, the product of the PKD2 gene, sequestered ID2 in the cytoplasm in a PKD1-dependent manner, suppressing inappropriate proliferation through E47-dependent p21 transcription (Li et al., 2005; Zhou, 2009). More recent publications focus on defects in Ca²⁺ regulation at the cilium as a cause of PKD, however, and the relationship to cell cycle regulation by ID- and E-proteins is unclear (Fliegauf et al., 2007; Fedeles et al., 2011).

Parkinsonism

Id2 homozygous null mutant mice display features of Parkinson’s disease, such as fewer dopaminergic neurons in the olfactory bulb and reduced olfactory discrimination (Havrda et al., 2008). Id2^−/− mice also have reduced dopamine transporter expression, activated caspase-3, as well as glial infiltration in the substantia nigra pars compacta of brain (Havrda et al., 2013) (Table 2). These observations suggest that ID2 is required for the development of midbrain dopaminergic neurons and raise the question of how ID2 might be involved in Parkinson’s disease and other disorders of the dopaminergic system such as attention deficit hyperactivity disorder, schizophrenia, and drug abuse. Intriguingly, significant inductions of ID1, ID2 and ID3 expression by decreased dopamine levels have been detected in rodents under continuous stress conditions (Konishi et al., 2010). At present, it is not known whether E-proteins play specific roles in dopaminergic neurons, other than their general role in neuronal differentiation (Figure 1).

Sjogren’s Syndrome

ID3 homozygous null (Id3^−/−) mice develop many symptoms similar to those found in Sjögren’s Syndrome, an autoimmune disease in which immune cells chronically attack the lachrymal and salivary glands, resulting in impaired tear and saliva secretion (Li et al., 2004) (Table 2). The effect of ID3 gene inactivation may be on lymphocyte differentiation because adoptive transfer of Id3^−/− bone marrow cells is sufficient to induce symptoms of Sjögren’s syndrome (Li et al., 2004; Guo et al., 2011), and because CD20 antibody treatment, which depletes B cells, relieves the Sjögren’s symptoms of Id3^−/− mice (Hayakawa et al., 2007). Since ID3 is required to downregulate E proteins during T cell development, defective T cells may also contribute to the disease (Engel and Murre, 2002; Kim et al., 2002a). Although no changes in ID3 mRNA levels have been seen in T cells from human patients, it remains an interesting possibility that defects in E-protein-dependent lymphocyte differentiation contribute to Sjögren’s Syndrome (Ling et al., 2014)

Conclusions and Prospects

Although genetic evidence in humans directly implicates E- and ID-proteins in Pitt-Hopkins Syndrome, schizophrenia, Rett Syndrome, atherosclerosis, and perhaps Fuchs Corneal Dystrophy and pulmonary arterial hypertension, how these diseases are related to the well-studied roles of E- and ID-proteins in normal development is not yet clear. More information would be useful to substantiate the contributions of E- and ID-proteins to Diamond Blackfan Anemia and Polycystic Kidney Disease. At the moment changes in E-proteins or ID-proteins are not implicated as direct causes of defects in metabolism or diabetes, bone fragility, hydronephrosis, Parkinsonism, or Sjogrens’ Syndrome in humans, but their mouse mutations provide models of these disorders and suggest they may be involved. This variety of diseases and models reinforce the notion that E-and ID-proteins, and their relative expression levels, constitute an important aspect of the cellular regulatory environment that remains important throughout life, not only during early development. The mechanisms established for HLH protein regulation in normal development provide a useful starting point for disease mechanisms (Figures 1–2). Do, for example, ID-proteins regulate dendrite growth through E-proteins and transcription, and is this relevant to neurological conditions? More still remains to be learned concerning normal E-and ID-protein expression and function, especially at the post-transcriptional level. Post-translational modification and changes in subcellular localization have all been reported (Lingbeck et al., 2005; Kurooka and Yokota, 2005; Sun et al., 2005; Rollin et al., 2009; Nio-Kobayashi et al., 2013; Lasorella et al., 2014). Protein turnover and stability may be a regulatory factor and therapeutic target. Recent work shows that the Drosophila E-protein Da is destabilized by Notch signaling, and this may be one of the mechanisms by which Notch regulates so many differentiation processes (Kiparaki et al., 2015). These studies exemplify the broad effects, and potential therapeutic opportunities, expected to accrue from modulating these very widely-expressed and important HLH proteins.