The Nucleosome Remodeling Factor

Abstract

An essential component of the chromatin remodeling machinery is NURF (Nucleosome Remodeling Factor), the founding member of the ISWI family of chromatin remodeling complexes. In vertebrates and invertebrates alike, NURF has many important functions in chromatin biology including regulating transcription, establishing boundary elements, and promoting higher order chromatin structure. Since NURF is essential to many aspects of chromatin biology, knowledge of its function is required to fully understand how the genome is regulated. This review will summarize what is currently known of its biological functions, conservation in the most prominent model organisms, biochemical functions as a nucleosome remodeling enzyme, and its possible relevance to human cancer.

1. Introduction

The fundamental unit of chromatin is the nucleosome, which is composed of four basic histone proteins tightly associated with ~150 bp of DNA contained in 1.75 super helical turns [1]. In many cases, chromatin presents a significant barrier to the interaction of trans-acting factors with DNA. As such, chromatin regulates many biological processes like transcription, DNA replication, DNA repair, and DNA recombination [2]. Epigenetic mechanisms have evolved to regulate the structure of chromatin, and as a result, access to DNA. These mechanisms include the post-translational modification of histones, DNA methylation, incorporation of histone variants, and nucleosome remodeling activities [3]. Nucleosome remodeling and the incorporation of histone variants are largely accomplished through the action of ATP-dependent chromatin remodeling complexes. These complexes are a diverse family grouped into SWI/SNF, ISWI, CHD, or INO80 sub-families, based upon sequence homology of the associated ATPase [4].

Since its discovery, the ISWI family member NURF (Nucleosome Remodeling Factor) chromatin remodeling complex has been documented as a key regulator of development in many prominent model organisms. Evidence suggests that NURF is an ATP-dependent chromatin remodeling complex, specifically targeted to chromatin through interactions with sequence specific transcription factors and modified histones. To examine the evidence for this model, and come to a better understanding of NURF function, this review will: examine key points of its ATP-dependent remodeling reaction with emphasis on its interactions with the nucleosome, summarize known biological functions for NURF, identify potential NURF homologs and conserved functions by sequence conservation in major model organisms, and summarize available evidence for its role in human cancer.

2. The Nucleosome Remodeling Factor

NURF was first identified in Drosophila melanogaster as an ATP-dependent biochemical activity that enhanced GAGA factor (GAGAG binding factor)-mediated nuclease accessibility to reconstituted chromatin [5,6]. Biochemical purification of this activity revealed a four subunit complex composed of NURF301, the ATPase ISWI (NURF140), NURF55 and NURF38 polypeptides [7]. Purifications from human cells identified a complex highly homologous to D. melanogaster NURF, strongly suggesting that it has been conserved through evolution [8]. Homo sapiens NURF contains the NURF301 homolog BPTF, the ISWI homolog SNF2L, and a NURF55 homolog pRBAP46/48; however, a NURF38 homolog has not been identified [8] (Fig. 1A).

Fig. 1.

Diagram of the NURF remodeling complex and its associated subunits. (A) Cartoon showing the subunit composition of D. melanogaster and H. sapiens NURF complexes. D. melanogaster NURF contains 4 subunits: NURF301, the largest and essential subunit; the ISWI ATPase; NURF55, a WD repeat protein; and NURF38, a pyrophosphatase. H. sapiens NURF has homologs strongly related to 3 of these subunits. BPTF is closely related to NURF301, SNF2L to ISWI, and pRBAP46/48 to NURF55. Interestingly, H. sapiens NURF does not contain a homolog of the NURF38 pyrophosphatase. (B) Domain analysis of D. melanogaster NURF subunits. Domains are color coded (cf. Fig. 2).

2.1. NURF301

D. melanogaster NURF301 has a number of functional domains found in other chromatin associated proteins (Fig. 1B). The N-terminal HMGA (High Mobility Group) domain contains two AT-hook sequences and an acidic patch that interacts with nucleosomes [9]. These interactions are likely due to direct contacts with DNA because the AT hook has known affinity with the minor groove of AT rich DNA [10]. The N-terminal DDT domain (DNA-binding homeobox-containing proteins and the different transcription and chromatin remodeling factors in which they are found) and PHD finger (Plant Homeodomain Zinc Finger) have not been specifically characterized in NURF301; however, a similar DDT domain in ACF1 is essential for interactions with ISWI [11]. The NURF301 WAC and WACZ domains are not well characterized, but similar domains in ACF1 are important for its interactions with DNA [11]. The C-terminal domains include a poly-glutamate region which is intrinsically disordered, two PHD fingers, and a bromodomain [12]. The most C-terminal PHD finger (PHD2) and bromodomain compose a histone recognition module that binds di/trimethyl-K4 on histone H3 (H3K4me2/3) and acetyl-K16 on histone H4 (H4K16ac), respectively [13–15]. Additional domains include nuclear localization signals, poly-proline regions, and LXXLL motifs. The latter have been shown to be important for protein–protein interactions and could be important for interactions with nuclear hormone receptors [16].

2.2. ISWI

Characteristic regions of ISWI are the ATPase domain, common to all remodeling proteins, and the C-terminal HAND, SANT, and SLIDE domains (Fig. 1B). ATPase domains are composed of a number of highly homologous motifs (Ia, Ib, II, III, IV, V and VI) separated by less conserved spacers that vary in length between ATPase family members [17]. The ISWI ATPase domain interacts with DNA in the nucleosome, ~20 bp away from the dyad axis [18]. Similar contacts have been observed for SWI/SNF chromatin remodeling complexes, suggesting an essential function in the remodeling reaction [19]. The C-terminal HAND, SANT, and SLIDE domains are highly conserved through evolution and are diagnostic of ISWI family members. The HAND, SANT, and SLIDE domains make essential contacts with the histone H4 tail and linker DNA, and are essential to the ISWI remodeling reaction [20,21]. In addition, the ISWI ATPase contains an N-terminal AT hook, and LXXLL motifs which likely interact with the nucleosome and facilitate protein–protein interactions, respectively (see Section 2.1).

2.3. NURF55

D. melanogaster NURF55 contains the highly conserved and widely utilized WD repeat domain (Fig. 1B) [22]. WD repeat-containing proteins are present in almost all organisms and are found in many chromatin associated complexes (for example Sin3, NuRD, CAF-1, PRC2, and pRB) [23]. They are named for the presence of four or more ~40 amino acid repeating units ending with conserved Gly-His (GH) and Trp-Asp (WD) residues [24]. Interestingly, NURF301 and ISWI alone are sufficient for ATP dependent remodeling of reconstituted chromatin suggesting that NURF55 is not essential for NURF activity in vitro [9]. NURF55 is proposed to indirectly interact with chromatin in vivo, likely through chromatin associated complexes [22]. In contrast to NURF55, the Xenopus laevis and H. sapiens homologs p48 and pRBAP48, respectively, directly interact with histone H4, suggesting that they have direct interactions with chromatin [22,25,26].

2.4. NURF38

NURF38 has a strong inorganic pyrophosphatase activity (Fig. 1B) [27]. Inorganic pyrophosphatases have been highly conserved through evolution because of essential functions in phosphate and nucleotide metabolism. Interestingly, NURF38 is not required for, and functions independently of, chromatin remodeling functions. One proposed function is to hydrolyze inorganic pyrophosphates, a byproduct of the RNA polymerase reaction, to increase the efficiency of transcription [27]. Human NURF does not have an associated pyrophosphatase homolog, making its significance to NURF function unclear.

3. The NURF remodeling reaction

The minimal substrate for the NURF remodeling reaction is a nucleosome with linker DNA [6]. D. melanogaster NURF mobilizes nucleosomes in 10 bp bidirectional step movements between stable positioning sequences with little unwrapping of the DNA from the nucleosome surface [28,29]. This is in contrast to activities of the SWI/SNF family, which can generate large loops of DNA from the nucleosome surface and can evict histones, and the INO80 family which are dedicated histone exchange complexes [30–33]. In vitro chromatin remodeling is influenced by linker length, strength of the DNA positioning sequence for the nucleosome, and biochemical properties of the remodeling complex [34,35]. NURF slides nucleosomes into a thermodynamically stable position or to the end of a DNA fragment, rendering them refractory to further remodeling [28,29,34]. DNA binding factors and adjacent nucleosomes can strongly influence the outcome of the remodeling reaction by providing barriers to the movement of nucleosomes in cis [36]. Thus it is widely assumed that the combination of DNA binding factors, adjacent nucleosomes, and physical properties of the DNA and histones significantly contribute to the outcome of NURF remodeling reactions in vivo.

Key elements of the nucleosome are required for NURF chromatin remodeling. Nucleosomes composed of histones lacking N-terminal tails are refractory to remodeling by NURF, stressing their importance to the reaction [37]. Mutagenesis has identified the histone H4 tail and its N-terminal proximal residues 16-KRHR-19 as the most important for NURF nucleosome remodeling functions. Site directed mutation of any of these residues, or acetylation of H4K16 or H4K8 by histone acetyltransferases, significantly inhibits the ATPase activity of ISWI [9,38–41]. Acetylation at residues other than H4K16 and H4K8 has little effect on NURF remodeling of mono nucleosomes, but acts synergistically with NURF to enhance transcription of chromatin in vitro [42]. In addition to the N-terminal tail, the H2A C-terminal tail is also essential for efficient chromatin remodeling by ISWI and SHF2H in vitro [43].

Histone variants can also regulate ISWI remodeling. H2A.Z is preferentially incorporated into nucleosomes at distal regulatory elements and nucleosomes at transcription start sites [44]. In vivo pull down assays indicate that human BPTF associates with the histone H2A.Z and in vitro SNF2L preferentially remodels H2A.Z-containing chromatin over H2A-containing chromatin [45]. The preference for H2A.Z chromatin as a substrate is dependent on residues in the extended acidic patch, a structure unique to H2A.Z [46]. These physical and functional interactions suggest that NURF may preferentially localize to, and remodel chromatin containing H2A.Z nucleosomes.

The activities of NURF in vitro have been studied extensively on the nucleosome, but the physiological substrate in vivo is the chromatosome [47]. The chromatosome is composed of the nucleosome core particle and a linker histone. Neither the ATPase activity of NURF, nor its ability to slide nucleosomes in cis, is influenced by linker histones [48]. This result is in contrast to other reports, which show that the SWI/SNF, CHD, and ISWI family members are negatively influenced by linker histones [49]. A direct comparison between these studies is difficult because of distinct differences in the source of the chromatin, the type of linker histones used, and the method of assembly; however, current evidence suggests that NURF can remodel the chromatosome as efficiently as the nucleosome.

In addition to H2A.Z, other regulators of NURF activity include post-translational modifications of ISWI. The post-translational modification of D. melanogaster ISWI by poly-ADP-ribosylation reduces its ATPase activity and nucleosome binding both in vivo and in vitro [50]. Poly-ADP-ribosylation of ISWI is thought to contribute to apoptosis through loss of essential chromatin remodeling activities of NURF. In addition to poly-ADP-ribose, D. melanogaster ISWI is acetylated on lysine753 in vivo and in vitro by the acetyl-transferase GCN5. GCN5 acetylated ISWI is found exclusively in the NURF complex in vivo; however, its biological functions have not been determined [51].

Small molecule inhibition of NURF has not been extensively characterized. The only small biomolecules known to affect NURF function are the inositol polyphosphates. Biochemical assays demonstrate that NURF and ISWI are negatively regulated by IP6, but not by IP5 or IP4. In Saccharomyces cerevisiae, known ISWI gene targets are deregulated in mutants defective for IP6, consistent with a regulatory function for the small molecule during in vivo chromatin remodeling [52]. It is not completely understood how NURF is regulated by inositol polyphosphates and the physiological significance of this control remains unclear.

4. NURF and its interaction with chromatin

Many studies show interactions between components of the NURF complex, or the highly related FAC1 protein (Fetal Alz-50-reactive Clone 1), and sequence specific transcription factors (Table 1) [9,53–65]. FAC1 has sequence identity to the N-terminus of the human NURF301 homolog BPTF, strongly suggesting the two are related [66,67]. These interactions are thought to recruit NURF to specific regions of the genome. In many cases the exact nature of the NURF transcription factor interaction has not been characterized. The most comprehensive study of NURF transcription factor interactions has been in D. melanogaster. From these studies, the authors show that HSF (Heat Shock Factor), GAGA and the artificial activation domain VP16 interact with multiple surfaces on NURF301 and weakly with ISWI [9]. Similar interactions have been observed between the SMAD transcription factors and the BPTF and SNF2L subunits of the human NURF complex (Landry, J., unpublished data). These results suggest interactions between NURF and transcription factors could occur over multiple surfaces. Whether these interaction surfaces are important in vivo needs to be explored [15].

Table 1.

Summary of published NURF interactions with transcription factors.

Organism	Interaction partner	Biological function	Molecular function	Reference
D. melanogaster	GAGAa,b	Heat shock, embryo patterning	Regulator of chromatin structure at insulators and promoters	[9]
D. melanogaster	HSFa	Stress response	Regulator of heat shock response	[9]
M. musculus/H. sapiens	Smad2a	Gastrulation	Nodal signaling	[53]
M. musculus/H. sapiens	SRFa,b	Thymocyte development	T cell receptor signaling	[54]
M. musculus	AP-1b	Thymocyte development	T cell receptor signaling	[54]
H. sapiens	MAZa	Tissue development	Cell type specific regulation of GC containing promoters	[55]
H. sapiens	hKeap1a	Stress response	Redox signaling	[56]
D. melanogaster	EcRa	Larval metamorphosisc	Ecdysteroid signaling	[57]
D. melanogaster	Kena	Innate immune system development	Cytokine signaling	[58]
D. melanogaster	Armadilloa	Wing development	Wingless signaling	[59]
D. melanogaster	HP2a,b	Heterochromatin	Gene silencing	[60]
H. sapiens/D.  melanogaster	PRa,b	MMTV transcription/female reproduction	Expression of MMTV viral genome/regulator of hCGH response	[61,62]
D. melanogaster	TRF2 and TAF4bb	Cell proliferation	Cell type specific regulation of genes encoding DNA replication factors	[63]
D. melanogaster	Pzgb	Wing developmentd	Notch, ecdysone, JAK/STAT signaling	[64]
G. gallus/H. sapiens	Usf1b	Globin gene regulation	Insulator function	[65]

^aInteraction by in vitro pull down.

^bInteraction by in vivo pull down.

^cInteraction implied during female germline stem cell self-renewal [97].

^dInteraction implied during larval metamorphosis and motility [95].

In addition to interactions with transcription factors, NURF301 has two well characterized domains which bind specific histone post-translational modifications. The PHD finger juxtaposed to the bromodomain (PHD2) interacts with H3K4me2/3, and the adjacent bromodomain binds H4K16ac (Fig. 1A) [13,14,68]. These adjacent domains may serve as a binding module specific for regions of active chromatin.

In addition to interactions with chromatin associated proteins, NURF likely directly interacts with DNA and some evidence suggests that it could be sequence specific. The AT hooks and N-terminal acidic patch on D. melanogaster NURF301 are important for NURF interactions with nucleosomes and for its remodeling activity [9]. AT hooks interact with AT rich DNA sequences, suggesting that this domain could foster direct contacts between NURF and DNA [10]. Sequence specific binding has been reported for FAC1, suggesting that NURF could also have this activity. Using a PCR assisted binding site selection assay, the consensus sequence CACAA-CAC was obtained for the N-terminal ~400 amino acids of FAC1. The consensus sequence represses luciferase reporter constructs when located near a SV40 promoter or enhancer, suggesting that it has activity in vivo [69]. Sequence specific binding activity is not unprecedented in remodeling complexes. For example, RSC in yeast has sequence specific binding activities through its Rsc3 subunit [70]. It will be important to investigate whether the sequence specific binding identified for FAC1 has relevance to DNA binding in vivo.

In vivo studies on the human NURF subunit SNF2L have revealed a uniform distribution throughout the nucleus at the same density as chromatin [71]. SNF2L is mobile throughout the nucleus with an average residence time of 1–2 ms, and only a small fraction (~3% in G1/2 phase and ~10% during S phase) is immobile with a residence time of ~150 ms. Depletion of cellular ATP or the expression of an ATPase inactive SNF2L + 13 splice variant (see below for explanation) results in increased mobility for the mobile fraction and decreased mobility for the immobile fraction. From these experiments a sampling model of chromatin remodeling is proposed, in which SNF2L continuously samples a large number of nucleosomes without chromatin remodeling (i.e., low residence time of 1–2 ms). Chromatin remodeling is proposed to occur (i.e., conversion to high residence time of ~150 ms) when a SNF2L-containing remodeling complex encounters chromatin marked by histone modifications, transcription factors or histone variants which have high affinity for the complex [71].

Consequences of NURF recruitment to chromatin have been observed both in vivo and in vitro. GAGA factor recruits D. melanogaster NURF to the HSP70 and Ftz promoters, as shown using reconstituted systems [6,72]. At HSP70 NURF remodeling activities are required for the heat shock-induced binding of HSF, an essential transcription factor for stress induced transcription [73]. NURF-dependent enhancement of transcription is not observed with naked DNA templates, suggesting that it functions to relieve the inhibitory effects of chromatin on transcription [74]. In a similar study, PR (Progesterone Receptor) was shown to target NURF to the MMTV promoter. Using this system NURF targeting and chromatin remodeling was shown to be important for subsequent NF1 binding and synergistic PR binding, two essential requirements for efficient transcription of the MMTV promoter [62].

In addition to having key regulatory roles during induced transcription, D. melanogaster NURF is also a general regulator of chromatin structure. One of the more striking phenotypes of NURF301 and ISWI mutants is the dramatic decondensation of the male X chromosome [73,75]. The DCC (Dosage Compensation Complex) and H4K16ac adjust for X chromosome haploinsufficency in males (females have 2 X chromosomes) by increasing its transcription 2-fold relative to autosomes. Mutation of the NURF components NURF301 and ISWI, or the expression of a dominant negative ATPase defective ISWI mutant, results in a drastically decondensed chromatin structure, increased expression of some genes on the X chromosome, and chromosome-wide loss of histone H1, but interestingly no significant defects in DCC association or histone acetylation [76,77]. NURF effects could be direct, through nucleosome remodeling and/or binding of histone H1, or indirect, through the regulation of unknown genes important for chromatin structure of the male X chromosome [78]. One possible mechanism for maintaining X chromosome structure is through NURF-dependent localization of the ATAC acetyltransferase. ATAC is a Gcn5 and Atac2-containing acetyltransferase complex important for regulated transcription [79,80]. Mutations in ATAC phenocopy the male X chromatin defects observed with NURF301 and ISWI mutants. It has been proposed that NURF-dependent localization of ATAC is required for the acetylation of histones and likely non-histone proteins, to regulate the chromatin structure of the male X chromosome [81].

In addition to regulating higher order chromatin structure of the male X chromosome, NURF has been characterized as a regulator of chromatin insulator elements in a number of contexts. In D. melanogaster, NURF is essential for the function of insulator elements at homeotic gene clusters including Fab7, SF1, and Fab8 in both S2 cells and animals. Defects in insulator function correlate with increased nucleosome occupancy, suggesting that NURF regulates chromatin structure to facilitate insulator function. NURF is proposed to be recruited to insulators by GAGA factor, a known sequence specific binding protein important for Fab7 and SF1b function, to position nucleosomes and regulate insulator activity [82]. Similar roles for NURF in insulator function were reported using the Gallus gallus and H. sapiens β-globin locus as a model. These studies show that the β-globin insulator associated factor USF1 co-purifies with a number of chromatin modifying proteins, including subunits of NURF and the hSET1 complexes. Knockdown of BPTF results in increased nucleosome occupancy at the endogenous 5’HS4 insulator element and is required for insulator function using a reporter assay. BPTF localization to the 5′ HS4 requires the hSET1 complex suggesting collaboration between the two complexes is required for insulator function [65]. In combination these studies clearly show that NURF has important roles in regulating insulator function, likely as a nucleosome remodeling complex.

There is also some evidence that NURF associates with heterochromatin. Using a biochemical approach, NURF301 was found to co-elute with heterochromatin-associated protein HP1 and HP2. These interactions were confirmed using in vitro pull downs, indicating direct interactions between NURF and HP2 [60]. Assays for NURF301 or ISWI as a dominant suppressor of position effect variegation did not reveal a biological connection between NURF and heterochromatin [73]. Observations supporting a functional role for NURF in facultative heterochromatin may require more sophisticated biological assays.

5. A sequence and phylogenetic analysis of NURF subunits

With the exception of S. cerevisiae, NURF301 homologs exist across all eukaryotic species investigated, suggesting that the NURF complex has largely been conserved through evolution (Fig. 2A and B). Many D. melanogaster NURF301 domains have been conserved in homologs from other species; these domains include the N-terminal HMGA domain or acidic patch, N-terminal DDT and PHD finger, and in the C-terminal region a polyglutamate region, PHD1 and PHD2 fingers, and a bromodomain. Two C-terminal PHD fingers and a bromodomain in D. melanogaster are conserved in Caenorhabditis elegans; however, only PHD2 has been conserved in every species investigated except Arabidopsis thaliana, alluding to its likely importance for NURF301 function (Fig. 3). While the AT hooks of the HMGA domain are conserved in D. melanogaster, C. elegans, and A. thaliana, only its acidic patch has been largely conserved throughout evolution. The N-terminal DDT, PHD finger, and surrounding residues are the most highly conserved domains, making these sequences diagnostic of the NURF301 protein family (Fig. 4). These regions have been shown to interact with the transcription factors MAZ, hKeap1, and GAGA [9,55,56].

Fig. 2.

NURF301 homologs are found in many widely studied model organisms. (A) Alignment of NURF301 homologs showing the positions of conserved domains. Color codes for domains are the same as in Fig. 1. Sequences used for analysis; H. sapiens NP_872579.2, M. musculus NP_789820.2, T. rubripes ENSTRUG00000008386, X. tropicalis XP_002942992.1, G. gallus ENSGALP00000005731, D. rerio XP_001920272.1, D. melanogaster NP_728507.1, C. elegans NP_001022117.1, A. thaliana AT5G12400.1. (B) Phylogenetic analysis of NURF301 homologs showing relatedness between species. Comparisons were made using the default settings for the Cobalt alignment program [116].

Fig. 3.

Strong sequence conservation at C-terminal PHD fingers and bromodomain of NURF301 homologs. (A) Hydrophobic residues important for H3K4me2/3 binding were identified from Ref. [68] and are designated by black arrows. These residues are highly conserved in all species except A. thaliana, making it quite likely that binding to H3K4me2/3 is conserved in most species. (B) Alignment and comparison of PHD1 and PHD2 domains of D. melanogaster, and C. elegans NURF301 and C-terminal PHD finger of A. thaliana. (C) The C-terminal PHD finger of A. thaliana is significantly different from PHD1 and PHD2 as shown by nearest neighbor analysis using default settings for the Cobalt alignment program [116]. (D) Comparison of bromodomains from NURF301 homologs showing strong sequence conservation through evolution. All alignments were made using default settings for the Clustal W program [117].

Fig. 4.

Conserved domains and sequences diagnostic for the NURF301 protein family. (A) Alignment and comparison of N-terminal DDT and PHD fingers from NURF301 homologs showing strong sequence conservation through evolution. (B) Alignment and comparison of conserved sequences near the N-terminal DDT and PHD fingers from NURF301 homologs showing strong sequence conservation through evolution. This region, encompassing the N-terminal DDT domain and PHD finger, is diagnostic of NURF301 homologs. Alignments were made using default settings for the Clustal W program [117].

One or more homologs of D. melanogaster ISWI exist in each of the eukaryotes investigated (Fig. 5). Each homolog shares extensive sequence similarity within the ATPase and HAND, SANT, and SLIDE domains. D. melanogaster, C. elegans and Takifugu rubripes have a single homolog, which has been named ISWI. In contrast, S. cerevisiae, Xenopus tropicalis, H. sapiens, Mus musculus, G. gallus, and Danio rerio have two distinct homologs which differ at amino acids in the N and C-terminal tails. These tails are important for the assembly of the ATPase subunit into specific remodeling complexes [83]. Members from each of these groups were first identified in H. sapiens and named SNF2L (SNF2-Like) and SNF2H [84,85]. Some chromatin remodeling complexes preferentially assemble with either the SNF2L or SNF2H variant, whereas others do not have a preference [86–89]. NURF has been shown to selectively assemble with SNF2L [8].

Fig. 5.

ISWI homologs are found in all widely studied model organisms. (A) The domain structure of D. melanogaster ISWI. Because the sequence of D. melanogaster ISWI and its homologs is so highly conserved, a single homolog is presented showing the position of conserved domains. Color codes for domains are the same as in Fig. 1. (B) Phylogenetic analysis of ISWI homologs showing relatedness between species. Species with two ISWI homologs can be segregated into SNF2H and SNF2L variants. Comparisons were made using the default settings for the Cobalt alignment program [116]. Sequences used for analysis; A. thaliana NP_850847.1 and NP_187291.2, S. cerevisiae NP_009804.1 and NP_014948.1, G. gallus XP_420329.2 and XP_001234486.1, D. melanogaster NP_523719.1, C. elegans NP_498468.2, X. tropicalis XP_002931866.1 and NP_001007993.1, T. rubripes ENSTRUP00000036063, H. sapiens NP_003060.2 and NP_003592.2, M. musculus NP_444353.3 and NP_444354.2, D. rerio NP_001075098.1 and NP_001093467.1.

6. The NURF family of remodeling complexes

As observed for the SWI/SNF family of remodeling complexes, the ISWI family varies its subunit composition as a means of regulating its in vivo functions [90]. Work in D. melanogaster and C. elegans indicates that NURF is not a single complex, but rather a family of related complexes (the elements of which largely differ as a consequence of alternative splicing) [13,91]. NURF301 is transcribed as four isoforms in D. melanogaster (NURF301A, B, C, or D) and five isoforms in C. elegans (NURFA, B, C, D, or E, none of which are a full length NURF301 isoform). Specific functions of these isoforms are unknown, with the exception of D. melanogaster NURF301C. Studies on a D. melanogaster NURF301 mutant allele that approximates the NURF301C short isoform show that the C-terminal PHD fingers and bromodomain of full length NURF301 are not required for viability. Pull downs from these mutants indicate that the truncated NURF301 mutant protein can assemble into a complex with ISWI, NURF55, and NURF38, suggesting that it has remodeling activities [13]. Similar truncated isoforms likely exist in mammals. A short form of BPTF called FAC1 was identified as a protein recognized by the ALZ50 antibody [67]. Antibodies specifically raised to FAC1 identify a protein of 125 kDa found in human brain, and independent studies have identified a similar protein in M. musculus [67] [53]. It is unknown whether FAC1, like the NURF301C isoform in D. melanogaster, associates in a complex with SNF2L and pRBAP46/48 to form an active chromatin remodeling complex, or if its functions are completely unrelated to chromatin remodeling.

Like NURF301, ISWI is expressed in different splice forms and in some cases the splice variant has significance to chromatin remodeling activities [61,92]. In humans, but not mice, a novel SNF2L splice variant is expressed which incorporates a 13th exon into the mature transcript (SNF2L + 13 variant), interrupting the ATPase domain [92]. Pull down experiments show that it assembles into the NURF remodeling complex but is catalytically inactive and unable to remodel chromatin. These results are significant because they show for the first time that chromatin remodeling complexes can be regulated by incorporating a catalytically inactive ATPase, thus providing a novel method of regulating activity.

7. Biological functions for NURF as a regulator of development

7.1. D. melanogaster

The genomic interval containing Nurf301 was identified as E(bx) (Enhancer of Bithorax) in screens to identify regulators of body patterning in D. melanogaster [93]. These results assigned the genomic interval containing Nurf301 as a trithorax member, a positive trans-regulator of homeotic genes in D. melanogaster [94]. Mutants in Nurf301 and Iswi are lethal at the third instar larval stage [73,75]. Nurf301 mutants, which likely reflect biological functions for NURF more specifically than Iswi mutants, have defects in regulated expression of homeobox, heat shock, JAK/STAT and ecdysone responsive genes [57,58,73]. NURF is likely a direct regulator of ecdysone and STAT responsive genes, as shown by its interactions with the ecdysone receptor and Ken, a repressor of STAT signaling [57,58]. Subsequent work showed that the regulation of ecdysone, homeobox, heat shock, and JAK/STAT pathways do not require the full length NURF301, but rather that the short splice form NURFC is sufficient to regulate the pathways [13].

The role of NURF during adult D. melanogaster development has been investigated using the tissue specific expression of dominant negative alleles and synthetic genetic studies. NURF positively regulates both the canonical Wingless and Notch pathways, as shown in genetic screens to identify regulators of wing development [59,64]. These genetic interactions were supported by in vitro or in vivo pull downs suggesting that NURF acts as a co-regulator of Armadillo or Pzg function. More recent studies show that Pzg mutants phenocopy many of the Nurf301 mutant phenotypes, and suggest that interactions between NURF and Pzg are important for many NURF functions during development [95]. A more general role for Pzg in regulating NURF functions could be through its interactions with TRF2 and DREF, two proteins that interact with NURF and act as selective regulators of transcription during development and cell cycle progression [63].

Special attention has been given to investigating NURF function during germ stem cell (GSC) development in the ovary. Mutation of Iswi in the female germ line results in reduced proliferation and self renewal of GSC. These defects are similar to mutations in the BMP signaling pathway, suggestive of a role for ISWI in regulating this pathway [96]. Subsequent work with Nurf301 mutants confirmed that these phenotypes are likely due to the NURF complex; however, a new role was uncovered for NURF as a regulator of ecdysone signaling in GSC self renewal [97]. It is currently unknown if NURF works exclusively through BMP or ecdysone signaling pathways, or a combination of the two, to regulate GSC self renewal in the D. melanogaster ovary.

Studies in the male germline show that NURF is required for self-renewal of GSC and cyst progenitor cells (CPC) by activating JAK/STAT signaling. Germ line clones of Nurf301 mutants lacking either the full length or short isoforms of NURF301 show a progressive loss of GSC and CPC, likely due to premature differentiation into early daughter cells (i.e., defective self renewal). Defects were observed for mutants of Iswi and Nurf38, and not for Acf1 or Dmi-2, documenting that the NURF complex, but not ACF, CHRAC or NURD complexes, are required for GSC and CPC stem cell self-renewal [98]. Similar experiments looking at sperm development showed that full length NURF301 is selectively required for late stage spermatogenesis. In flies lacking full length NURF301, there is a pronounced block in sperm development after the first meiotic division. The block in meiosis correlates with reduced levels of cyclin B, a known regulator of cell cycle progression. Analysis of gene transcription profiles shows that NURF regulates the Bruno family of translational inhibitors, which are known inhibitors of cyclin B protein levels [13]. These studies show that both the full length and short NURF301 isoforms are required for GSC and CPC self renewal, but conversely, only full length NURF301 is required for late stage spermatocyte development.

7.2. M. musculus and H. sapiens

Bptf knockout mice do not gastrulate due to defects in the differentiation of extra-embryonic tissue lineages. Bptf-dependent lineages include the distal visceral endoderm, an essential precursor to one of the first organizers of the mammalian embryo, and the ectoplacental cone, an essential tissue at the interface between the embryo and decidua [53,99]. This could be due in part because of a requirement for Bptf in Nodal signaling through the Smad transcription factors, an essential signaling pathway for the specification of embryonic and extra-embryonic tissue differentiation in the early embryo. Characterization of Bptf knockout M. musculus embryonic stem cells shows a greater NURF requirement for mesoderm and endoderm differentiation than for ectoderm differentiation [53]. The same Bptf knockout embryonic stem cells have significant defects in the expression of homeobox genes, supporting similar observations from D. melanogaster, X. laevis, and H. sapiens [8,14,53,73,98].

As with D. melanogaster, studies on NURF function in the adult mammal have been limited because Bptf is essential for embryonic development. Cre-loxp conditional knockout technology revealed that Bptf is essential for adult thymocyte development [54]. In vitro and in vivo pull downs show specific interactions between NURF and SRF and AP-1, two transcription factors essential for T cell receptor-dependent gene regulation. In this study defects in chromatin structure and transcription were observed prior to and after β-selection; consistent with a role for NURF having functions very early in thymocyte development prior to the positive selection process.

As is true for D. melanogaster, mammalian NURF likely has important functions during differentiation of adult germ line stem cells. During ovarian follicle development the Snf2l transcript and protein levels are elevated with hCG stimulation. hCG in concert with progesterone regulates the development of oocytes in mammals. In vivo pull down from granulosa cells show stimulation-dependent interactions between Snf2l and the PR, a key transcription factor which regulates the transcription of gene targets in response to hCG. ChIP studies show stimulation-dependent recruitment of Snf2l to the promoter of a Snf2l-dependent granulsoa cell differentiation marker gene StAR, suggesting direct functions in its regulation [61].

7.3. X. laevis

Biochemical purifications of ISWI-containing complexes from egg extracts did not identify a NURF complex homolog [100]. These results suggest that the NURF complex does not exist in X. laevis oocytes. A morpholino knockdown of NURF301 at the 2 cell stage of development reveals defects in axial development, gut formation and blood cell development, as well as aberrant regulation of homeobox gene expression [14]. A similar morpholino knockdown of ISWI at the one cell stage results in pronounced defects in gastrulation, neural fold closure, and eye development. ISWI knockdown mutants have defects in many neural specific gene targets including Bmp4, which is proposed to be a direct target of ISWI [101].

Some very interesting chromatin remodeling-independent functions for ISWI were discovered using the X. laevis model. ISWI was identified biochemically as a microtubule-associated protein whose interaction with microtubules is independent of its ATPase activity. ISWI knockdowns in both X. laevis egg extracts and in D. melanogaster S2 cells results is dramatic decondensation of microtubules during anaphase, suggesting that ISWI has activities independent of chromatin remodeling in promoting microtubule stability during anaphase [102].

7.4. C. elegans

In C. elegans both the NURF301 homolog NURF-1 and ISW-1 are essential for the synMuv (Synthetic Multivulva) phenotype. Syn-Muv genes are negative regulators of vulvae cell development and frequently encode novel nuclear components, repressors of transcription, and factors that remodel chromatin. Nurf-1 or Iswi-1 mutations suppressed all synMuv combinations tested, demonstrating that NURF acts as an important positive regulator of the vulval cell fate. SynMuv alleles regulated by NURF include gain of function mutations in the ras signaling pathway, suggesting NURF may have important regulatory functions in this pathway [91].

7.5. A. thaliana

Mutants of NURF301 in A. thaliana have not been reported; however, a knockout of the ISWI homolog CHR11 has been characterized. Loss of function of CHR11 results in defects during the diploid and haploid phases of plant development. Diploid phase knockdown resulted in reduced plant height and small cotyledonary embryos, whereas knockdown during female gametogenesis resulted in arrested megagametogenesis. These phenotypes correlate with reduced cell expansion possibly causing the reduction in the size of the embryos [103].

8. Roles for NURF in human cancer

Primary human cancers, and cancer cell lines from many tissues including brain, breast, lung, liver, and prostate, frequently duplicate the 17q distal chromosome arm containing the BPTF gene [104–107]. Many genes have been proposed to be responsible for selection of this chromosomal duplication; however, none of the reports considered BPTF as a possible candidate [108–111]. As a result, it is unknown if BPTF gene duplication provides an advantage to cancer cells or is an indirect consequence of 17q21 distal chromosome duplications. One report identified a non-reciprocal translocation (der(X)t(X;17)) in the BPTF gene when human embryonic lung cells were maintained in continuous culture [112]. The chromosomal translocation resulted in increased BPTF mRNA levels and correlated with increased cellular proliferation. Consistent with the literature on 17q distal duplications, the authors report frequent BPTF duplications in neuroblastomas, lung tumors, leukemia and colon cancers.

In a targeted study, SNF2L expression was found to be equivalent in tumor tissues and normal tissue controls [113]. The authors observed that a number of human cancer cell lines are sensitive to knockdown of SNF2L but not SNF2H. Knockdown of SNF2L resulted in reduced proliferation, increased DNA damage and increased apoptosis, suggesting that the selective inhibition of SNF2L could be a viable therapy for some cancers. Why these cancer cells are sensitive to SNF2L, but not SNF2H knockdown is unknown and likely a topic for future study. There is evidence that SNF2 protein expression is elevated in some human cancers. In a targeted study of prostate cancer, the SNF2 protein was significantly elevated in prostate neoplasm compared to benign prostatic hyperplasia [114]. However, the antibody used in this study reacts with both SNF2H and SNF2L, making it difficult to determine which of the variants are increased during the progression of prostate cancer.

9. Concluding remarks

The literature clearly documents that the NURF complex has critical functions in many aspects of chromatin molecular biology. Most importantly, NURF has critical functions regulating higher order chromatin structure and gene expression output of many prominent signaling pathways to the nucleus. While this ground-work has established much in the way of understanding NURF function, many important questions remain to be answered. Some of these questions include

1. What are the direct targets of NURF chromatin remodeling in vivo? Comprehensively identifying direct NURF targets from multiple cell types and model organisms is an essential step towards understanding its in vivo functions. Use of high throughput methods to map changes in chromatin structure with Bptf/NURF301 knockout and determine NURF localization to the genome will begin to identify these sites.

2. How is NURF directed to specific regions of the genome? NURF is likely recruited to specific regions of the genome by interactions with transcription factors, histone modifications, and possibly specific DNA sequences. Identifying all of these elements will make a significant contribution towards identifying NURF dependent genes a priori.

3. What is the outcome of the NURF remodeling reaction in vivo? Germane to this line of inquiry is understanding how physiologically relevant elements of chromatin, like transcription factors, linker histones and histone modifications influence the outcome of the in vivo remodeling reaction.

4. Is the ATP-dependent remodeling activity of NURF its only function in vivo? Recent reports show that the Brg1 remodeling enzymes have gene regulatory functions independent of their ATPase activity [115]. It will be essential to our study of the complex to investigate if NURF, like Brg1-containing complexes, has similar ATP-independent functions.

5. Are remodeling reactions by NURF stable or transitory in vivo? Knowing how dynamic the events of chromatin remodeling are in vivo will help us to understand their short and long term effects on chromatin structure.

6. Is NURF301/BPTF exclusive to the NURF complex? It will be essential to know if the largest and essential subunit of NURF functions exclusively as a component of the NURF complex in both mammals and other model organisms.

7. Will regulating NURF activities have any therapeutic benefit to humans? It has been well documented that NURF functions as a critical component of signaling pathways relevant to human disease (TGFβ and JAK/STAT, as two examples). Exploring NURF as a therapeutic target to regulate these pathways for the treatment of human diseases will be an important avenue of research.

The challenge for the future will be to address these outstanding questions and learn much about this essential chromatin remodeling complex with the objective of advancing both basic and translational science.