MEIS transcription factors are key regulators of embryonic development and cancer. Research on MEIS genes in the embryo and in stem cell systems has revealed novel and surprising mechanisms by which these proteins control gene expression. This Primer summarizes recent findings about MEIS protein activity and regulation in development, and discusses new insights into the role of MEIS genes in disease, focusing on the pathogenesis of solid cancers.
Ever since their discovery, MEIS (myeloid ectopic viral integration site) genes have been linked to both embryonic development and cancer. They were independently identified as homeobox (HOX) co-factors in vertebrates and invertebrates, because they cause homeotic transformations when mutated in D. melanogaster and accelerate disease progression of HOX-dependent leukemias in humans (Moskow et al., 1995; Rieckhof et al., 1997). Recent studies have uncovered novel molecular details of MEIS protein function in different developmental settings, yet many of these new insights into the MEIS gene family have not yet been evaluated in the light of existing models. In addition, their relevance to cancer research, especially in solid tumors, has not been fully explored. In this Primer, we survey recent advances in understanding MEIS protein function during embryogenesis and compare these to the, often contradictory, reports about MEIS protein contribution to solid cancer. We first describe the MEIS family protein structure, MEIS protein complexes and interactions, and the function of MEIS proteins in regulating gene expression and chromatin dynamics. Next, we discuss the complex regulation of MEIS genes that gives rise to their versatile function. We then address MEIS protein function in development and disease, particularly in solid cancer. Finally, we briefly discuss remaining questions and potential approaches with which to answer them. As the oncogenic role of MEIS genes in leukemia has been well characterized and extensively reviewed elsewhere (Argiropoulos et al., 2007; Collins and Hess, 2016a,b; Eklund, 2011), it is not covered in detail here.
The MEIS protein family
MEIS protein structure, evolution and conservation
MEIS genes encode for transcription factors of the MEINOX family, a separate group within the atypical TALE (three amino acid loop extension) superclass of homeodomain (HD)-containing proteins. Like most HD-containing proteins, TALE-HD proteins use their homeodomain to contact DNA. Yet the TALE-HD differs from the classical HD by the insertion of three additional amino acids. These three amino acids are almost always a proline-tyrosine-proline (PYP) motif and are located between the first and second helix of the HD (Bürglin et al., 1997; Mukherjee and Bürglin, 2007). This ‘three-amino-acid-loop-extension’ gave the TALE-HD protein class its name; it is highlighted in red in the lower sequence box of Fig. 2A. The PYP motif forms a hydrophobic pocket, which participates in formation of TALE-HD-containing protein complexes on DNA (Dard et al., 2018, 2019; LaRonde-LeBlanc and Wolberger, 2003; Passner et al., 1999; Piper et al., 1999; Saadaoui et al., 2011). The TALE-HD superclass itself comprises five separate classes: MEINOX, PBC and the more distantly related TG-interacting factors (TGIFs), Iroquois (IRO) and Mohawk (Mkx) (Fig. 1) (Bürglin, 1997; Mukherjee and Bürglin, 2007). MEINOX genes are present in animals and plants, where they are known as MEIS and KNOX, respectively. In vertebrates, the MEINOX class is further divided into the MEIS and PKNOX/PREP families, and mammals possess three MEIS (MEIS1, MEIS2 and MEIS3) and two PREP (Prep1 and Prep2) genes. By contrast, only a single MEINOX gene is found in D. melanogaster (homothorax, hth) and in C. elegans (unc-62). The MEIS and PREP families thus seem to have split concomitantly with the emergence of vertebrates. The PBC class is more distantly related and comprises Pbx1, Pbx2, Pbx3 and Pbx4 in vertebrates, ceh-20, ceh-40 and ceh-60 in C. elegans, and extradenticle (exd) in D. melanogaster (Fig. 1). Here, we focus on MEIS genes in vertebrates, but do touch upon some invertebrate models for comparison. For a deeper understanding of the roles of other TALE genes, we refer the reader to excellent reviews on Prep1 by Blasi et al. (2017) and PBC genes by Selleri et al. (2019). We instead discuss recent findings on Meis1, Meis2 and Meis3, focusing on novel advances concerning their varying mechanisms of action and diverse modes of regulation. As all MEIS proteins are highly similar (with over 85% identity among MEIS1 and MEIS2 in mice and humans), it is experimentally difficult to discriminate between them in assays that involve overexpression systems or rely on antibodies. To take this into account, we use the term ‘MEIS’ to refer to all MEIS homologs, and use individual gene names only when distinct functions were attributed to individual MEIS family members.
MEIS proteins share two common structural features, the MEINOX-homology domain consisting of two short alpha helices, MH-A and MH-B, and the TALE-HD. In addition, all MEINOX proteins possess N- and C-terminal unstructured regions, produced from alternatively spliced transcripts (Fig. 2A) (Geerts et al., 2005). Although the N terminus does not seem to possess any biochemical activity, the C terminus harbors a transcriptional activation domain (Huang et al., 2005; Hyman-Walsh et al., 2010). MEINOX family proteins also possess a nuclear export signal (NES) and the TALE-HD is N-terminally flanked by a stretch of basic amino acids that, as in many HD-containing transcription factors, can serve as nuclear localization signal (NLS) (Kolb et al., 2018) (Fig. 2A). We discuss the implications of the dual presence of NES and NLS motifs later.
MEIS protein complexes and interactors
A prominent feature of MEINOX proteins is their ability to heterodimerize with the structurally related TALE proteins PBX1-PBX4. This interaction occurs in the absence of DNA and is mediated by class-specific homology domains located N-terminal to the DNA-binding homeodomain: MH-A and MH-B in MEINOX proteins; PBC-A and PBC-B in PBX (Chang et al., 1997; Knöepfler et al., 1997; Shanmugam et al., 1999). MEIS/PREP and PBX family cooperativity thus appears to be largely achieved by direct protein-protein interaction, and not by the more common mechanism of binding to neighboring recognition sites on DNA (Jolma et al., 2015) (Fig. 2B). MEINOX and PBC genes are ubiquitously present in eukaryotes and MEIS/PREP-PBX heterodimerization occurs in all species tested to date (Joo et al., 2018). The MEINOX-PBC dyad thus represents an evolutionary ancient molecular scaffold, the primary function of which likely involves the formation of heterologous, DNA-bound protein complexes. Interestingly, the transcriptional activity of the MEIS C terminus is blocked in the monomeric protein, likely because the MEINOX domain folds back onto the C terminus, inhibiting its activity in the absence of PBX (Hyman-Walsh et al., 2010). PBX-MEIS dimerization elicits extensive conformational changes in the MEIS polypeptide, which relieve this constraint and provide a simple but effective control over MEIS protein activity.
MEIS proteins form higher order complexes and cooperatively bind DNA with additional transcription factors, usually as MEIS-PBX heterodimers. A prominent group of MEIS protein-binding partners facilitated by direct interactions are the HOX proteins, such as Abd-B (Abdominal B) class HOX proteins (posterior paralog groups 9-13) (Mann and Affolter, 1998; Ladam and Sagerström, 2014). Flies mutant for hth resemble posterior HOX gene ablation phenotypes without altering HOX gene expression itself (Rieckhof et al., 1997). hth, therefore, controls patterning of the body axis by modulating HOX protein activity rather than by controlling HOX gene activation, identifying MEIS family proteins as co-factors of posterior HOX genes. Indirect complex formation mediated by PBX also occurs between MEIS- and anterior paralog group Antp (Antennapedia) class proteins (Shen et al., 1997; Shanmugam et al., 1999; Williams et al., 2005) (Fig. 2C). In addition, a broad spectrum of other transcription factors also directly associate with MEIS family proteins and/or cooperatively bind DNA together with MEIS (summarized in Box 1).
MEIS family proteins can directly or cooperatively bind DNA with a diverse range of transcription factors. Among these are various HD-containing proteins, including all HD-bearing members of the PAX protein family, as well as OTX2, DLX1 and DLX2, the pancreas-specific protein PDX1, or engrailed (en) in D. melanogaster (Agoston and Schulte, 2009; Agoston et al., 2012, 2014; Heine et al., 2008; Kobayashi, 2003; Liu et al., 2001; Schulte, 2014; Swift et al., 1998). MEIS also cooperates with members of more distant transcription factor families, such as the T-box family member EOMES, basic helix-loop-helix (bHLH) proteins such as MAX, Ets transcription factors like ELF1, zinc-finger transcription factors like KLF4 or teashirt (TSH), GATA-transcription factors, NF-Y, and nuclear receptors like the androgen receptor (AR) in prostate cancer or the ecdysone receptor (EcR) in D. melanogaster (Bessa et al., 2002; Bjerke et al., 2011; Cui et al., 2014; Groß et al., 2018; Jolma et al., 2015; Knoepfler et al., 1999; Ladam et al., 2018; Neto et al., 2017). In addition, MEIS can interact with several other nuclear proteins, including epigenetic regulators, with the nuclear enzyme poly-ADP polymerase 1 (PARP1/ARTD1), and even some structural proteins of the cell nucleus (Choe et al., 2009, 2014; Groß et al., 2018; Hau et al., 2017).
The functional consequences of these diverse binding events are only beginning to be unraveled. Interaction with different protein partners may result in the recognition of different sequence motifs on DNA, ultimately favoring binding to distinct sites in the genome. Considering that MEIS proteins bind DNA mostly together with PBX family members, these additional interactions may change the MEIS transcription factor binding preferences among the pool of MEIS-PBX sites that are present in the genome. In addition, and as discussed in more detail below, MEIS protein binding to the same target sites may result in different functional outputs depending on the precise composition of the multimeric transcriptional complex that assembles around the MEIS family protein.
Mechanism of action
MEIS proteins regulate gene expression and orchestrate chromatin dynamics. ChIP-seq studies [chromatin immunoprecipitation (ChIP) followed by high-throughput sequencing] have demonstrated that the MEIS1 and MEIS2 proteins select two main binding motifs in DNA: the monomeric motif ‘TGACAG’, or close variants thereof, which is likely bound by MEIS1/2 directly; and a motif resembling the PBX-HOX binding sequence, ‘A/TGATTNAT’, to which both MEIS proteins bind indirectly (Amin et al., 2015; Penkov et al., 2013; Marcos et al., 2015). This indirect consensus binding sequence has been deduced from ChIP-seq analyses of the mouse embryonic trunk or isolated branchial arches. In these tissues, HOX proteins likely represent predominant MEIS- or MEIS/PBX-binding partners (Amin et al., 2015; Penkov et al., 2013). Yet, given their ability to cooperate with a wide spectrum of different transcription factor types, MEIS proteins can likely make indirect DNA contact through many other binding partners and associate with more diverse sequence motifs than are presently known. Further genome-wide binding studies from cells and tissues that do not express HOX genes are needed to address this issue.
MEIS transcription factors can modulate the epigenetic signature of DNA and chromatin of their target genes in remarkably diverse ways. Central to most of these is the high affinity of MEIS for PBX, enabling the MEIS protein to orchestrate the assembly of multi-protein transcriptional complexes surrounding its PBX partner. A prime example of this is the de novo activation of neuron-specific genes during differentiation of mouse adult neural stem cells (Fig. 3). In the undifferentiated progenitor cells, neuron-specific genes are transcriptionally silent and, although their promoter and enhancers are bound by PBX1, the loci are still compacted by the linker histone H1 (Grebbin et al., 2016; Hau et al., 2017). Upon initiation of neuronal differentiation, MEIS1 and MEIS2 recruit the nuclear enzyme PARP1 (also known as ARTD1) to these PBX1-primed sites and initiate PARP1-mediated eviction of H1 from the chromatin (Hau et al., 2017). Release of H1 then leads to local decompaction of the chromatin fiber, making these DNA elements accessible for additional, later-acting proteins. Interestingly, MEIS factors themselves can promote the loading of several such proteins to DNA. For example, in the sequential activation of HOX-dependent genes during zebrafish hindbrain development, MEIS proteins can facilitate the exchange of histone deacetylases for the CBP histone acetyl transferase (HAT) from PBX-bound genomic sites, initiate the deposition of activating histone modifications at these sites, and promote loading of the RNA polymerase II holoenzyme to the promoters of downstream genes (Choe et al., 2009, 2014). An attractive concept emerging from these studies is that MEIS transcription factors, by interacting with multiple components of the transcription machinery, guide their target genes through the consecutive regulatory steps that lead to full transcriptional activation and effective gene expression.
Complex regulation of MEIS factors allows for versatile function
Epigenetic regulation of MEIS gene expression
The examples described above give an impression of the transcriptional changes that MEIS proteins can elicit and their diverse physiological consequences. The biological activity of MEIS transcription factors must therefore be tightly regulated. Indeed, multiple layers of epigenetic, post-transcriptional and post-translational regulation apply to MEIS genes. For example, transcriptional activation of the Meis2 gene during mouse midbrain development involves extensive remodeling of the chromatin topology surrounding the Meis2 promoter, with successive association and dissociation of a promoter, a midbrain-specific enhancer and a genomic site located 3′ to the coding exons (Kondo et al., 2014). These stepwise contacts are orchestrated by Ring1B, a component of the polycomb repressive complex 1 (PRC1), suggesting a role for Ring1B in timing the onset of expression of developmentally regulated genes, like Meis2 (Kondo et al., 2014). Ring1A and Ring1B also mediate Meis2 repression in the distal limb bud in a process that is shaped by retinoic acid (RA) signaling (Yakushiji-Kaminatsui et al., 2016, 2018). In addition, the DNA methylation status of MEIS gene promoters has been linked to disease, and hypomethylation of the MEIS1 promoter is frequently observed in leukemia (Deveau et al., 2015; Ferreira et al., 2016). Similarly, the histone methyltransferase NSD1 (nuclear receptor SET domain containing protein 1), controls MEIS1 expression through histone tri-methylation of lysines 20 (H4K20me3) and 36 (H3K36me3) at the MEIS1 promoter regions in glioma and neuroblastoma cells, and epigenetic silencing of NSD1 causes MEIS1 upregulation (Berdasco et al., 2009).
Alternative splicing generates isoforms with different biological functions
Adding another level of complexity, MEIS genes can give rise to multiple splice isoforms with very different biological activities. In D. melanogaster, a splice isoform lacking the DNA-binding homeodomain of hth is expressed alongside the canonical, full-length isoform. Unexpectedly, many of the developmental activities that had been attributed to the full-length Hth transcript, such as segmentation or proximal-distal axis formation, are in fact carried out by the truncated HD-less isoform (Noro et al., 2006). Hth (and possibly vertebrate MEIS proteins too) can thus function without the need to directly bind DNA. In mammals, many different MEIS transcript splice variants can be formed. Of particular interest are the 3′ splice variants that result from differential splicing of protein-coding exons 11 and 12, which give rise to four major protein isoforms (called a-d) with different C termini (Fig. 4A). Although alternative splicing and tissue-specific expression of the MEIS2a-d variants were recognized when the gene was first cloned (Oulad-Abdelghani et al., 1997), different functions for individual MEIS2 isoforms have only recently been established in the context of cancer (discussed in more detail below; Groß et al., 2018). Given that the MEIS1 and MEIS3 genes share the same exon-intron structure, it is plausible to assume that variants with similarly divergent biological activities also exist for other MEIS genes (Geerts et al., 2005). Functional studies to test this hypothesis, however, are yet to be reported. Investigating the alternative splicing of MEIS transcripts in different physiological and pathophysiological settings is therefore of high priority, and can perhaps be achieved by CRISPR/Cas9-mediated removal of the relevant splice sites. Additional transcript variants are also generated by using alternative promoters, but these are not well described and most do not result in protein sequence changes. For a complete overview of MEIS transcripts see www.ncbi.nlm.nih.gov/gene/ and www.ensembl.org/index.html with MEIS1, MEIS2 and MEIS3 as search terms.
MEIS genes are under complex post-transcriptional control
Several miRNAs regulate MEIS transcripts in the embryo or during organ homeostasis (Fig. 4B). These include miR-204, which targets Meis2 in the developing retina of Medaka fish and the lens of mice, respectively (Conte et al., 2010; Hoffmann et al., 2012). In addition, miR-155 and miR-144 regulate MEIS1 and meis1 transcripts during hematopoiesis and vascular development, as shown in human cord blood samples in vitro and in developing zebrafish in vivo (Romania et al., 2008; Su et al., 2014), respectively. Finally, miR-9 and miR-134 target Meis2 in the mouse embryonic telencephalon and MEIS2 in human cells, which were isolated from fetal hearts by their expression of the Sca1 (stem cell antigene 1) epitope, respectively (Shibata et al., 2011; Wu et al., 2015). Meis1 is also among the transcripts that are modified by N6-methyladenosine in mouse embryonic cortical progenitor cells (Yoon et al., 2017). This epi-transcriptomic regulation determines the stability of selected transcripts during corticogenesis and synchronizes their translation, including that of Meis1, with the progression of forebrain development.
MEIS proteins are under complex post-translational control
A large body of evidence also exists for the post-translational regulation of MEIS proteins, where controlled nuclear import is of particular importance. The presence of NES and NLS motifs in the MEIS and PBX polypeptides suggests that both proteins can shuttle between nucleus and cytoplasm (Fig. 2B) (Kolb et al., 2018). Control over their subcellular localization thus provides a simple but effective mechanism to regulate the activity of this protein group. Interestingly, the NES in the MEIS polypeptide is fully embedded in the PBX-binding surface (Knoepfler et al., 1997). Consequently, binding of MEIS to either PBX or the nuclear export receptor chromosomal maintenance 1/exportin 1 (CRM1/XPO1) must be mutually exclusive events, which provides a parsimonious mechanism to regulate the subcellular localization of MEIS proteins (Kolb et al., 2018). Conversely, MEIS1 promotes nuclear localization and stabilizes PBX1 in the limb bud and hindbrain of vertebrate embryos (Mercader et al., 1999; Waskiewicz et al., 2001), and the PBX homolog extradenticle (exd) requires heterodimerization with Hth for nuclear import in D. melanogaster (Abu-Shaar et al., 1999; Rieckhof et al., 1997). Although co-dependency of PBX and MEIS transcription factors for nuclear accumulation is seen in many different developmental contexts, this process itself can be regulated by post-translational modification: serine phosphorylation of PBX1 circumvents its dependency on MEINOX proteins for nuclear localization, while arginine methylation of MEIS2 decreases its affinity for CRM1 (Kilstrup-Nielsen et al., 2003; Kolb et al., 2018). This modification interferes with MEIS2 nuclear export, favors its heterodimerization with PBX1 in the cell nucleus, and thereby ensures stable nuclear accumulation of the MEIS2 protein (Fig. 4C). MEIS protein activity can also be modulated by other forms of post-translational modification; MEIS1, for example, is a phosphoprotein in EGF-treated HeLa cells (Olsen et al., 2006), whereas MEIS2 becomes ubiquitylated by the E3 ubiquitin ligase complex CRL4, leading to proteasomal degradation of MEIS2 (Fischer et al., 2014). Notably, thalidomide [α-(N-phthalimido)glutarimide], a sedative that was taken off the market in the mid-1960s due to its highly teratogenic effects, targets the CRL4 component CRBN. This leads to inactivation of the CRL4 complex and stabilization of MEIS2 (Fischer et al., 2014; Ito et al., 2010). Thalidomide-induced MEIS2 degradation is remarkable, because a proximal-distal gradient of MEIS protein is required in the limb bud at early embryonic stages for the correct development of the skeletal elements of the limb (Table 1) (Mercader et al., 1999). This finding therefore not only provides a possible explanation for the limb deformities associated with the use of thalidomide during the first trimester of pregnancy, but may also be relevant for the recently re-evoked interest in thalidomide as an anti-cancer therapy (Pan and Lentzsch, 2012).
Versatile roles of MEIS family proteins in the embryo
MEIS transcription factors are widely expressed during embryogenesis, with expression in overlapping and sometimes complementary domains and highly dynamic spatial-temporal patterns. In chick and mouse embryos, for example, Meis1 and Meis2 transcripts are present in the primitive streak and lateral plate mesoderm during gastrulation, followed by the neural tube, eye, inner ear, mid- and hindbrain, proximal limb buds, somites, branchial arches, migrating neural crest, and cardiovascular system at late somite stages (Coy and Borycki, 2010; Cecconi et al., 1997; Sánchez-Guardado et al., 2011). The Meis3 gene (which seems to have been lost in the bird genome; Sánchez-Guardado et al., 2011) is prominently expressed in the hindbrain and pancreas, but appears not be expressed in the limbs or retina of zebrafish and X. laevis embryos (Manfroid et al., 2007; Salzberg et al., 1999). Consistent with these broad and complex expression patterns, MEIS family genes have emerged as key regulators of diverse developmental processes. Important examples of the developmental roles of MEIS are listed in Table 1. We now take the role of MEIS proteins in the developing eye, heart and limb as examples to link these to three common functions of the MEIS family: driving the proliferation of tissues, promoting cell cycle exit, and tissue patterning and differentiation (Fig. 5).
MEIS proteins have a conserved role in regulating progenitor cell proliferation, particularly during eye development. Indeed, D. melanogaster that are mutant for hth exhibit impaired proliferation of eye progenitor cells during larval life (Bessa et al., 2002; Lopes and Casares, 2010), while mice lacking Meis1 are microphthalmic, with mild developmental defects in the lens and retina by mid-gestation (Hisa et al., 2004; Marcos et al., 2015). Meis genes share a high degree of sequence similarity, indicative of extensive functional redundancy in regions of co-expression. Reflecting such redundancy, Meis1 and Meis2 are co-expressed in the presumptive lens surface ectoderm and simultaneous deletion of both genes abolishes lens development at pre-placodal stages and thus causes earlier and more profound defects than mutation of either gene alone (Antosova et al., 2016). Multiple components of the cell cycle machinery are under direct or indirect transcriptional regulation by MEIS transcription factors. In the zebrafish and chick embryonic neural retina, MEIS1 (together with MEIS2 in the avian retina) promotes the expression of cyclin D1 (ccnd1) and c-myc (myca) (Bessa et al., 2008; Heine et al., 2008) (Fig. 5A). Remarkably, Ccnd1, besides its well-established role in controlling G1-S phase transition of the cell cycle, also directly drives expression of Meis2 in mouse retinal progenitor cells (Bienvenu et al., 2010), forming a self-reinforcing regulatory circuit that promotes proliferation. Meis1, in turn, is an integral part of the genetic network that regulates growth and patterning of the vertebrate eye and controls the expression of several components of the Notch signaling pathway (including Notch2, Jag1, Hes2 and Hes5) in the early stage mouse eye cup (Marcos et al., 2015). In the developing compound eye of D. melanogaster, Hth maintains the proliferative and undifferentiated state of eye progenitor cells, and delays cell cycle exit and differentiation by repression of proneural genes such as atonal (ato) and string (stg) – the D. melanogaster homolog of the Cdc25 phosphatase, which triggers entry into mitosis by dephosphorylation of the CDC2 protein kinase (Kumagai and Dunphy, 1991; Bessa et al., 2002; Lopes and Casares, 2010, 2015). However, more evidence is necessary to demonstrate direct regulation of cell cycle genes by MEIS proteins. In fact, a recent study combining transcriptome analysis with open chromatin profiling and motif analysis suggests that, although expression of several cell cycle genes is sensitive to altered hth levels during D. melanogaster eye development, these cell cycle genes are likely to be indirect Hth targets, and are instead regulated through an intermediary layer of nuclear receptors (Neto et al., 2017).
Several studies have established a regulatory loop between MEIS proteins and the Wnt signaling pathways, both canonical and non-canonical, which suggest other ways MEIS factors might regulate proliferation (Fig. 5B). In the zebrafish retinal margin stem cell niche, Meis1 acts upstream of – or parallel to – Wnt/β-catenin signaling, which is necessary to maintain the undifferentiated progenitor/stem cell compartment (Stephens et al., 2010). In the mouse embryonic branchial arches, MEIS factors bind in close proximity to the Wnt5a gene (Amin et al., 2015), whereas in X. laevis embryos, Xmeis3 acts in a regulatory circuit with Xwnt3a to orchestrate hindbrain specification (Elkouby et al., 2012). As Wnt signaling influences cell proliferation by regulating (among others) Ccnd1 expression, these findings further support a proliferation-promoting role of MEIS factors in stem or embryonic progenitor cells.
In sharp contrast to a role of MEIS factors in promoting proliferation, MEIS protein activity has also been linked to cell cycle exit in the adult murine heart. The cyclin-dependent kinase inhibitors encoded by the Ink4b-Arf-Ink4a locus, which include p16, p15 and p19ARF, and p21CIP1 are under positive control by MEIS1 in murine adult cardiomyocytes and targeted deletion of the Meis1 gene is sufficient to induce mitosis in these cells (Fig. 5C) (Mahmoud et al., 2013).
Finally, MEIS proteins can promote differentiation, patterning and segmentation in the embryo. A close connection hereby exists between MEIS and RA signaling, notably again in the form of reciprocal regulatory circuits. RA activates Meis1 and Meis2 expression in the developing vertebrate limb bud (Freemantle et al., 2002; Mercader et al., 2000; Oulad-Abdelghani et al., 1997). MEIS proteins, in turn, inhibit the expression of the RA-degrading enzyme Cyp26b1, ultimately forming a RA concentration gradient in the developing limb, which is instrumental for correct patterning along the proximal-distal axis of the limb (Mercader et al., 1999, 2000; Rosello-Diez et al., 2014). Likewise, in the mouse embryonic hindbrain and in neuroblastoma cell lines, expression of the Aldh1a2 gene encoding the RA-producing enzyme RALDH2 is under direct, positive regulation by MEIS2, while transcription of Cyp26a1, which encodes a RA-degrading enzyme, is decreased (Fig. 5C) (Groß et al., 2018; Vitobello et al., 2011). Given that RA is a well-established inducer of cell cycle exit and cellular differentiation in the nervous and hematopoietic systems, MEIS proteins can trigger irreversible differentiation processes by generating a local RA-source within a proliferating tissue or cell mass (Collins, 2002; Janesick et al., 2015). In contexts where RA functions in regionalization or pattern formation, such as the hindbrain or limb, MEIS proteins emerge as key regulators of segmentation and morphogenesis (Dibner et al., 2001; Mercader et al., 2000; Vitobello et al., 2011).
In summary, although the requirement of MEIS proteins for proliferation and maintenance of undifferentiated cells in many biological contexts is undisputed, and their ability to induce cellular differentiation or pattern formation in others is equally well established, a unifying model does not emerge from these studies.
Versatile roles of MEIS proteins in disease
MEIS genes also have important functions in congenital disease (Box 2), leukemias and solid cancers. Indeed, Meis1, the founding member of the family, was discovered when its gene locus was first identified as a neoplastic viral integration site in a mouse model for acute myeloid leukemia (AML), and MEIS1 dysregulation was subsequently recognized as pivotal driver of disease progression in several human leukemia types (Imamura et al., 2002; Kawagoe et al., 1999; Moskow et al., 1995; Nakamura et al., 1996; Rozovskaia et al., 2001). In particular, in leukemia with MLL gene rearrangements [chromosomal aberrations including the KMT2A (‘mixed-lineage leukemia’) histone methyltransferase, representing aggressive forms of the disease], a wealth of published evidence supports a tumor-promoting role for MEIS family proteins; this has been reviewed extensively (Argiropoulos et al., 2007; Collins and Hess, 2016a,b; Eklund, 2011). Reports on the contribution of MEIS genes to solid tumors, by contrast, are mostly incidental. Aggravating this situation, expression of the same MEIS gene can be elevated in some cancer types, but downregulated in others. Even within one tumor entity, presumptive oncogenic and tumor-suppressive roles have been attributed to the same MEIS transcription factor (Table 2).
MEIS1 mutation is an important risk factor for restless legs syndrome (RLS), a neurological condition characterized by nocturnal painful sensations and with a neuro-developmental component (Hammerschlag et al., 2017; Lane et al., 2017; Schormair et al., 2017; Spieler et al., 2014; Winkelmann et al., 2007). Additional MEIS1 mutations have been linked to atrial fibrillation, while mutations in the MEIS2 gene involve cardiac defects, facial dysmorphism and intellectual disability (Crowley et al., 2010; Douglas et al., 2018; Erdogan et al., 2007; Fujita et al., 2016; Giliberti et al., 2019; Johansson et al., 2014; Louw et al., 2015; Pfeufer et al., 2010). Thus, human congenital defects associated with MEIS mutations reflect the developmental expression of MEIS family members in the heart, brain and neural crest (Frank and Schulte, 2014; Maeda et al., 2001; Stankunas et al., 2008; Toresson et al., 2000).
A good example for such discrepancies is in prostate cancer: this malignancy, usually adenocarcinomas arising in the prostate bladder, is one of the most frequent types of cancer in men. Its etiology is complex, but hereditary components are known and include a germline mutation in HOXB13 (84G→E) in early onset familial disease (Ewing et al., 2012). This mutation lies within the domain by which HOXB13 contacts its co-factor MEIS1, suggesting a role for MEIS1 in prostate cancer initiation. Functional studies testing this hypothesis, however, have been largely inconclusive (Bhanvadia et al., 2018; Chen et al., 2012; Cui et al., 2014; Jeong et al., 2017). Why these discrepancies exist remains unresolved.
Another solid cancer with contradictory evidence regarding the role of MEIS genes is neuroblastoma: a pediatric, highly metastasizing malignancy that arises from the sympathoadrenal lineage of the neural crest. Early studies revealed abundant expression of MEIS genes and occasional amplification of the MEIS1 gene locus in neuroblastoma cell lines and tumor samples, which are indicative of an oncogenic function for MEIS genes in this cancer (Fischer and Berthold, 2003; Geerts et al., 2003, 2005; Spieker et al., 2001). Supporting this view, overexpression of X. laevis MEIS1 in frog embryos induced the formation of tumor-like cell masses, and MEIS2 knockdown in neuroblastoma cell lines arrested tumor cell proliferation due to mitotic defects (Maeda and Ishimura, 2002; Zha et al., 2014). However, a significant positive correlation exists between high MEIS2 expression and favorable disease outcome according to publicly available mRNA-expression profiles of samples from thousands of individuals with neuroblastoma. This correlation indicates that a purely oncogenic role for MEIS genes in neuroblastoma is unlikely (Groß et al., 2018). An initial explanation for these discrepancies came from analyzing alternative splicing of exon 12 and the resulting isoforms, MEIS2A and MEIS2D (Fig. 4B). It has been shown in neuroblastoma that expression of the MEIS2A isoform induces proliferation arrest and neuronal differentiation, whereas expression of the MEIS2D isoform or MEIS2A knockdown increases the aggressiveness of already highly malignant neuroblastoma cells and tumors. The differential roles for MEIS2A and MEIS2D in neuroblastoma cells most probably stem from their C-terminal domains, which can assemble largely non-overlapping interactomes, allowing them to recruit different proteins to the regulatory regions of target genes (Groß et al., 2018).
An additional twist in this context is provided by the related protein PREP1, an established tumor-suppressor that governs a distinct group of downstream target genes and is highly expressed in neuroblastoma (Dardaei et al., 2014; Geerts et al., 2003; Longobardi et al., 2010; Penkov et al., 2013). PREP1 competes with MEIS1 for dimerization with the shared co-factor PBX1, as demonstrated using several cancer cell lines, including some from individuals with neuroblastoma (Dardaei et al., 2014). When MEIS1 is co-expressed with PREP1, the relative intracellular concentration of both proteins thus determines MEIS1 protein function (Blasi et al., 2017). In addition, like other central oncogenic transcription factors, such as MYC, MEIS1 has recently been linked to activation of a super-enhancer (a region comprising multiple enhancers that bind several transcription factors, driving cell-identity transcription). In Ewing's sarcoma, MEIS1 is a super-enhancer-associated gene itself and cooperates with the EWS-FLI oncogene to activate the super-enhancer of a cancer-associated downstream gene, linking MEIS protein activity to enhancer-mediated gene activation in cancer (Lin et al., 2019).
Taken together, the examples described above show that RNA and protein levels can have limited predictive value for the functionality of TALE-HD transcription factors, which should be considered when evaluating expression data for MEIS or PBX genes in patient tumor samples. In addition, although MEIS gene expression is frequently regarded as an oncogenic biomarker, caution needs to be exercised in any generalization. The varying and often opposing activities of MEIS factors in different solid cancers rather appear to be a reflection of their diverse, wide-reaching roles in controlling developmental programs in the embryo (discussed above). As we have explained, the mechanisms for such divergent activities can likely be found in the complex modes of regulation of MEIS protein activity, in the differential cell type-specific alternative splicing patterns of MEIS genes and in the diverse spectrum of binding partners that MEIS proteins can assemble with.
Concluding remarks and future perspectives
Initially described as little more than ‘helping hands’ of HOX proteins, MEIS proteins by themselves now emerge as important and highly versatile regulators of cellular behavior. In embryonic development or tissue homeostasis by adult stem cells, MEIS factors can recruit a broad spectrum of epigenetic modulators to chromatin. In addition, MEIS gene activity is tightly regulated on multiple levels, from RNA expression, splicing and stability to post-translational modification of the protein and controlled subcellular availability, all of which respond to extracellular cues. This raises the intriguing possibility that this highly conserved protein family has evolved to synchronize the interaction between chromatin and chromatin-modifying enzymes on one hand with the temporal, spatial or cell-type specific requirements of the cell on the other hand. In simpler terms: MEIS proteins may serve as sensors for extracellular signals and rapidly translate this information into a transcriptional outcome. The selection of target genes, which are regulated in response to these events, likely depends on the splice isoforms of MEIS transcripts generated and the nature of MEIS-interacting proteins that are prevailing in the respective cells. This concept still requires systematic experimental testing, yet if proven correct, it implies that MEIS protein activity is likely regulated on a cell-by-cell basis and dictated by the cellular context in which these proteins act. A challenge faced by cancer researchers will therefore likely be the realization that MEIS protein function may have to be assessed separately for different cell and tumor types, quite possibly even for different subtypes of the same cancer. The greatest challenge for developmental biologists will be to integrate the regulatory networks that are driven by MEIS factors with the complex extracellular stimuli and intracellular signaling pathways that impinge on these proteins. The wealth of genomic, transcriptomic and proteomic data, which were generated for the MEIS family in recent years, together with ever rapidly developing single cell technologies will make this a not too distant goal.
We thank members of our laboratories for helpful discussions and critical reading of the manuscript. We acknowledge the many excellent scientists that have made invaluable contributions to the MEIS field and apologize to those whose work we could not cite due to space limitations.
Collaboration leading to this review was made possible through a European Cooperation in Science and Technology grant BM0805 (HOX and TALE transcription factors in Development and Disease), through the Deutsche Forschungsgemeinschaft (DFG) (SCHU1218/3-3), through KWF Kankerbestrijding and through Wilhelm Sander-Stiftung.
The authors declare no competing or financial interests.