1

Regulation of Expression and Activity of

the bHLH-PAS Transcription Factor NPAS4

David Christopher Bersten

B.Sc. (Biomedical Science), Honours (Biochemistry)

A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy

Discipline of Biochemistry

School of Molecular and Biomedical Science

University of Adelaide, Australia

June 2014

2

Contents

Abstract ................................................................................................................................................... 3

PhD Thesis Declaration ........................................................................................................................... 5

Acknowledgements ................................................................................................................................. 6

Publications ............................................................................................................................................. 8

Conference oral presentations ........................................................................................................... 9

Additional publications ....................................................................................................................... 9

Chapter 1: .............................................................................................................................................. 10

Introduction ...................................................................................................................................... 10

bHLH-PAS transcription factor structure and mechanism ............................................................ 10

bHLH-PAS dimerization, DNA binding and signal transduction mechanisms ............................... 16

bHLH-PAS transcription factor function ........................................................................................ 19

Neuronal PAS Factor 4 .................................................................................................................. 19

Drosophila Dysfusion .................................................................................................................... 20

NPAS4 Expression and Regulation ................................................................................................ 21

NPAS4–dependent Transcriptional Regulation ............................................................................. 24

NPAS4 Function ............................................................................................................................. 26

Chapter 2: Results ................................................................................................................................. 31

2.1 Elucidation of molecular mechanisms which restrict NPAS4 expression to the brain. .............. 31

2.2 Human variants in the NPAS4/ARNT2 transcriptional complex can disrupt function ................ 32

2.3 Generation and validation of Inducible and reversible systems for lentiviral and ES cell (Col1a1-RMCE) mediated manipulation of NPAS4 levels. ................................................................ 33

Chapter 3: Discussion ............................................................................................................................ 36

References............................................................................................................................................. 45

Appendix 1 .......................................................................................................................................... 103

Appendix 2 .......................................................................................................................................... 104

Appendix 3 .......................................................................................................................................... 105

Appendix 4 .......................................................................................................................................... 106

3

Abstract

Development of the Central Nervous System (CNS) relies on complex transcriptional programs

to specify distinct neuronal areas/cell types, and guide the formation of neuronal networks.

Synaptic activity during post-natal brain development dictates the number and strength of

synapses as well as promoting neuronal cell survival through activation of transcriptional

programs. The  establishment  of  synapses  during  this  “critical  period” of post-natal neuronal

development and the local rearrangement, fine tuning and maintenance of synaptic

connections into adulthood contributes to synaptic plasticity, memory, learning and cognitive

function, while dysfunction in these processes is thought to contribute to a number of

neuropsychiatric diseases. Studying transcription which underlies these events and disease

states has been technically challenging due to the lack of gain and loss of gene expression

systems to interrogate complex biological questions in primary neurons or the developing

nervous system of rodents. As a result, despite clinical and anatomical data, the molecular

mechanisms underlying neuropsychiatric disease or memory and learning remain poorly

understood.

The basic-Helix-Loop-Helix (bHLH) – Per/Arnt/Sim (PAS) (bHLH-PAS) homology domain

transcription factor Neuronal PAS factor 4 (NPAS4) is tightly coupled to neuron function by

homeostatically regulating neuronal activity via stimulating formation of inhibitory synapses.

NPAS4 expression is brain restricted and highly induced following neuronal depolarisation,

paradigms of learning, seizure or ischemia. NPAS4 null mice are prone to seizures,

hyperactivity, have defects in memory formation, social interaction, cognitive impairments, as

well as age related neurodegeneration.

This thesis shows that NPAS4 expression is highly restricted to the CNS, in particular the

cortex, by repressive activity of RE-1 Silencing Transcription Factor/Neuron-Restrictive Silencer

Factor (REST/NRSF) in non-neuronal cells and stem cells. In addition, we also provide evidence

4

that microRNA-224 targets  the  NPAS4  3’UTR,  which  may  contribute  to  regionalised  NPAS4  

expression in the brain. We identify human variants within NPAS4 and ARNT2 which disrupt

NPAS4 function, which may have implications for neuropsychiatric disease. Using structural

modelling and biochemical experiments we show that one of these variants disrupts

dimerisation, providing insight into bHLH-PAS dimerisation mechanisms.

We also describe a new system for knockdown and ectopic expression which is broadly

applicable for reliable, flexible and temporal control of gene expression to facilitate

investigating gene function. This system incorporates single gateway compatible vector

systems for lentiviral infection and Recombination Mediated Cassette Exchance (RMCE), the

latter targeting the Collagen 1a1 (Col1a1) locus in germline competent embryonic stem cells.

Using an optimised reverse tetracycline transactivator (rtTA) system with reduced background

expression and increased sensitivity to doxycycline, we have shown that we can rapidly

generate inducible overexpression and short hairpin RNA (shRNA) mediated knockdown cell

lines with homogenous, inducible expression.

The work encompassed within this thesis investigates the molecular mechanisms underlying

the restricted expression pattern of NPAS4, the contribution of human non-synonymous

variants to NPAS4/ARNT2 transcription factor function, and the development of flexible,

inducible and reversible gene expression systems for studying NPAS4 function in vitro and in

vivo.

5

PhD Thesis Declaration

This thesis contains no material which has been accepted for the award of any other degree or

diploma in any university or other tertiary institution to David Christopher Bersten and, to the

best of my knowledge and belief, contains no material previously published or written by

another person, except where due reference has been made in the text. I give consent to this

copy of my thesis when deposited in the University Library, being made available for loan and

photocopying, subject to the provisions of the Copyright Act 1968. The author acknowledges

that copyright of published works contained within this thesis (as listed below*) resides with

the copyright holder(s) of those works. I also give permission for the digital version of my

thesis  to  be  made  available  on  the  web,  via  the  University’s  digital  research  repository,  the  

Library catalogue, and also through web search engines, unless permission has been granted

by the University to restrict access for a period of time.

David Christopher Bersten

June 2014

6

Acknowledgements

Without the support, patience, generosity, friendship and love of colleagues, mentors, friends,

and family the PhD journey would not have been as enjoyable or rewarding. For these people I

must acknowledge.

Ass. Prof. Murray Whitelaw, you are truly an inspiring scientist and educator. I am forever

indebted to you for nurturing and supporting my interest in science, and guiding my

development as an independent researcher. I aspire to be as purely principled in scientific

pursuit, as kind, as patient and as enthusiastic as you are.

Dr. Daniel Peet, you have been a fantastic co-supervisor and mentor for me over the years.

Your door has always open for scientific discussion or a beer on a Friday afternoon, the former

being invaluable to me.

The biochemistry discipline – I would like especially thank all the past and present researchers

and educators in the biochemistry discipline for creating and maintaining an environment of

excellence and scientific rigor within the discipline.

For all those who have helped be along the way, colleagues and friends thank you for all the

help and support. There are many to name but I would specifically like to thank Adrienne

Sullivan, Pete McCarthy, Jo Wright, James Hughes, Paul Thomas, Sandy Piltz, Stephen Bent,

and Andrew (Nan) Hao.

Jayne, the love and support you have given me is above and beyond, thank you for being so

understanding. I would  also  like  to  thank  the  McConachy’s  and  my  siblings  for  their  support

and patience.

7

Lastly, I must acknowledge my parents Libby and Andrew. This was not possible without your

unfailing support and gentle guidance. Thank you for having confidence in me and allowing me

to pursue my interests.

8

Publications

This thesis is based on the following publications and referred to in the text:

I. Bersten DC, Sullivan AE, Peet DJ, Whitelaw ML. bHLH-PAS proteins in cancer. Nat Rev

Cancer. 2013 Dec;13(12):827-41.

II. Bersten DC, Wright JA, McCarthy PJ, Whitelaw ML. Regulation of the neuronal

transcription factor NPAS4 by REST and microRNAs. Biochim Biophys Acta. 2014

Jan;1839(1):13-24.

III. Bersten DC, Bruning JB, Peet DJ, Whitelaw ML. Human Variants in the Neuronal Basic

Helix-Loop-Helix/Per-Arnt-Sim (bHLH/PAS) Transcription Factor Complex

NPAS4/ARNT2 Disrupt Function. PLOS one. 2014 Jan, DOI:

10.1371/journal.pone.0085768

IV. Bersten DC, Sullivan AE, Bhakti V, Li D, Thomas PQ, Bent S, Whitelaw ML, Inducible

and reversible lentiviral and recombination mediated cassette exchange (RMCE)

systems for controlling gene expression. Manuscript submitted Nucleic Acids Research

9

Conference oral presentations

REGULATION OF EXPRESSION AND ACTIVITY OF THE NEURONAL TRANSCRIPTION FACTOR

NPAS4

Bersten D.C., Peet D.J. and Whitelaw M.L. Australian Society of for Biochemistry and

Molecular Biology annual conference (ComBio 2012)

Additional publications

Bonnefond A, Raimondo A, Stutzmann F, Ghoussaini M, Ramachandrappa S, Bersten DC,

Durand E, Vatin V, Balkau B, Lantieri O, Raverdy V, Pattou F, Van Hul W, Van Gaal L, Peet DJ,

Weill J, Miller JL, Horber F, Goldstone AP, Driscoll DJ, Bruning JB, Meyre D, Whitelaw ML,

Froguel P.

Loss-of-function mutations in SIM1 contribute to obesity and Prader-Willi-like features.

J Clin Invest. 2013 Jul 1;123(7):3037-41.

All publications have been reproduced with the permission from the copyright holders

10

Chapter 1:

Introduction

NPAS4 is a member of the basic-Helix-Loop-Helix (bHLH) – Per/Arnt/Sim (PAS) (bHLH-PAS)

transcription factor family which is almost exclusively expressed within the central nervous

system (CNS) and is tightly regulated by neuronal activity. Like other members of this family it

must heterodimerise with a class II bHLH-PAS transcription factor (Aryl hydrocarbon receptor

nuclear translocator (Arnt) or Arnt2) to bind DNA and activate transcription. The heterodimer

binds to asymmetric E-BOX like elements (NNCGTG) within the promoters and enhancers of

target genes to modulate gene expression. The function of NPAS4 is primarily to

homeostatically regulate the activity state of excitatory neurons to keep a balance between

excitatory and inhibitory inputs within the neuron.

bHLH-PAS transcription factor structure and mechanism

bHLH-PAS proteins are a conserved family of proteins which contain two major functional

domain structures, the bHLH domain and the PAS domain (or PAS repeats)(Crews 1998; Yun,

Maecker et al. 2002; Kewley, Whitelaw et al. 2004; McIntosh, Hogenesch et al. 2010).

Transcription factors containing basic Helix-Loop-Helix (bHLH) motifs contribute to gene

expression profiles critical for many development processes such as myogenesis,

haematopoiesis, neurogenesis, heart and pancreatic development (Massari and Murre 2000).

This motif mediates dimerisation and DNA binding to a degenerate E-box CANNTG to regulate

transcription (Blackwell, Kretzner et al. 1990; Blackwell and Weintraub 1990; Baxevanis and

Vinson 1993). The N-terminal basic region of this motif contacts bases in the major groove of

DNA, while the HLH domain primarily contributes to dimmer stability (Ma, Rould et al. 1994).

The interaction between the two amphipathic helices of two proteins allows the formation

of homo- or heterodimers, which pairs the basic regions to facilitate binding to their cognate

11

half sites (Ferre-D'Amare, Prendergast et al. 1993; Ma, Rould et al. 1994). Over 240 HLH

proteins have been described in almost all eukaryotes ranging from Saccharomyces cerevisiae

to Caenorhabditis elegans, Drosophila melanogaster to mammals (Atchley and Fitch 1997;

Massari and Murre 2000).

There are two sub-families of bHLH transcription factors which utilize secondary dimerisation

domains adjacent to the bHLH domain; these are the bHLH leucine zipper (Zip) transcription

factor family and the bHLH-Per/Arnt/Sim (bHLH-PAS) homology domain family of

transcriptional regulators (Massari and Murre 2000). The bHLH-PAS proteins are characterized

by two N-terminal PAS domains denoted PAS A and PAS B. The PAS domain or the PAS repeats

are also ancient sensor and protein interaction domain structures found in organisms from

bacteria to humans (Crews and Fan 1999; McIntosh, Hogenesch et al. 2010; Henry and Crosson

2011). The PAS domain was initially defined as a ~275 amino acid motif with shared sequence

homology between the Drosophila melanogaster clock protein Period (Per), the midline

expressed neurogenesis factor Single Minded (Sim) and vertebrate ARNT (Reddy, Zehring et al.

1984; Crews, Thomas et al. 1988; Hoffman, Reyes et al. 1991). However, recent structural

insights have indicated that isolated PAS domains encompass an ~110 amino acid generic PAS

fold which is structurally highly conserved but poorly conserved in amino acid sequence

(Figure 1)(Huang, Chelliah et al. 2012; Bersten, Sullivan et al. 2013).

The PAS domains are present in thousands of proteins and are most highly represented in

bacterial proteins (Moglich, Ayers et al. 2009). A general mechanism of PAS domain sensing

coupled to transduction effector output domains, such as protein kinase, catalytic,

dimerisation, DNA binding or ion channel domains, has been proposed for PAS containing

proteins. Indeed PAS domain containing proteins are involved in sensing and responding to

light, voltage, oxygen, xenobiotics, redox potential and biological time keeping (McIntosh,

Hogenesch et al. 2010; Henry and Crosson 2011). A classic signal sensing mechanism exists in

12

Photoactive Yellow Protein (PYP), providing an example of a blue light sensing PAS protein

which acts negatively on phototaxis (Genick, Borgstahl et al. 1997; Henry and Crosson 2011).

In contrast to prokaryotes, few ligands or stimuli directly targeting the PAS domains have been

described for the mammalian PAS containing proteins. The best example lies with planar

aromatic hydrocarbons and dioxins, ligands which have been shown to directly bind to the PAS

B repeat in the Aryl hydrocarbon Receptor (AhR), thereby stimulating translocation to the

nucleus and activation of genes encoding xenobiotic metabolising enzymes (Whitelaw,

Gottlicher et al. 1993; Antonsson, Whitelaw et al. 1995; Whitelaw, McGuire et al. 1995). It has

also been reported that the NPAS2 PAS A domain can bind heme and carbon monoxide,

resulting in inhibition of an interaction with its bHLH/PAS partner, BMAL1 (Dioum, Rutter et al.

2002). Per2 has also been shown to bind heme, which mediates degradation of the protein.

As Per is an inhibitor of CLOCK/BMAL1, reduced Per levels result in increased CLOCK/BMAL1

dimer formation and function (Yang, Kim et al. 2008). CLOCK and NPAS2 have overlapping

function and can both bind nicotinamide adenine dinucleotide (NAD), which inhibits DNA

binding in vitro (Rutter, Reick et al. 2001). Recent structural resolution of the PAS B domain of

the hypoxia inducible factor HIF2 reveal its ability to bind small synthetic ligands and X-ray

crystallography of the bHLH-PAS A-PAS B regions of the CLOCK-BMAL1 dimer show a CLOCK

cavity which could potentially be targeted by small molecules (Scheuermann, Tomchick et al.

2009; Huang, Chelliah et al. 2012). While ligands or stimuli that directly act on the PAS

domains have not been described for most bHLH-PAS proteins there is generally a signal

regulated role for most PAS containing proteins.

Across all phyla, the PAS domain is frequently used as a protein-protein interaction interface

and ligand binding domain and forms a conserved structural module despite poor primary

sequence homology (Taylor and Zhulin 1999; Moglich, Ayers et al. 2009). Conformity of the

PAS domain structure has been elucidated over the last decade, mostly from crystallisation of

13

bacterial PAS containing proteins. The overall PAS fold and structure comprises five central

stranded anti-parallel -sheets (denoted A , B , G , H and I ) which are flanked by alpha

helices (denoted C , D , E , and F ) and connected by flexible linkers ( Figure 1). This forms

the PAS core or PAS fold which presents an ideal pocket for binding natural or synthetic

ligands (exploited by antagonist discovery for HIF2 (Scheuermann, Tomchick et al. 2009)).

When dual PAS repeats are present they are usually separated by a variable, unstructured

linker domain. Initially overlaying the structures from the Light, Oxygen and Voltages (LOV)

PAS domain containing sensors PYP, FixL and HERG demonstrated the conservation of the

overall PAS fold structure. As additional structures have been solved the overall conservation

of the PAS fold has held true even between kingdoms of life. However, the notion of signal

regulated domains and specificity of protein-protein interactions between PAS domains or PAS

domain contain proteins suggests a significant degree of conformation flexibility or surface

specific interactions. Recently the concept of surface specific interactions which confer

specificity has been described for the ARNT/AhR transcriptional complex in a screen for PAS A

domain residues critical for dimerisation (Hao, Whitelaw et al. 2011). Furthermore, recent

structures of the bHLH-PAS (PAS A and PAS B) of CLOCK and BMAL1 support the notion of

conserved PAS fold structure but divergent conformation. For example, while PAS A domains

of CLOCK and BMAL interact through reciprocal -helical and -sheet interactions, which is

also observed for NifL and Per/CLOCK PAS interactions, the CLOCK/BMAL1 PAS B domain

interactions fundamentally differ from those seen in HIF2 /ARNT PASB interactions (Huang,

Chelliah et al. 2012).

14

Figure 1 | Roles and structures of class I and class II bHLH–PAS family members. a | Examples of heterodimeric basic

HLH (helix-loop-helix)–PER–ARNT–SIM (bHLH–PAS) transcription factors are shown. Dimers are formed between a class I

factor and a class II factor (AHR nuclear translocator (ARNT), ARNT2, brain muscle ARNT-like 1 (BMAL)). A class I factor may

be tissue-restricted in expression (such as single-minded homologue (SIM) and neuronal PAS domain-containing protein

(NPAS)) or active in response to a stimulus (such as a ligand for aryl hydrocarbon receptor (AHR) and hypoxia for hypoxia-

inducible factor (HIF)). The heterodimers bind to class I-specific variations of the canonical E-box sequence and regulate

specific target genes, thereby mediating various developmental or homeostatic processes and/or responses to

environmental or physiological stresses. Class II factors are capable of forming dimers with more than one class I factor,

although some combinations may be limited in vivo; for example, ARNT2 expression is mostly restricted to neurons, and it

functions as the obligate partner in vivo for SIM1 and probably for NPAS4. HIF1 can form active dimers with either ARNT

or ARNT2 (REF. 37), but the HIF–ARNT dimer is more abundant overall owing to both proteins being ubiquitously

expressed. Conversely, ligand-bound AHR can dimerize with either ARNT or ARNT2 in vitro, but only AHR–ARNT dimers can

activate target genes171. b | The HLH and PAS folds constitute the interface of dimerization between class I and class II

bHLH–PAS factors. Comparison of the limited available structural information indicates that the tertiary structure of the

PAS fold is well conserved. However, this does not preclude different orientations and surfaces from being involved in

15

PAS–PAS interactions throughout the family. The PASB regions of BMAL1 and circadian locomotor output cycles kaput

(CLOCK) associate in roughly parallel orientations, whereas isolated PASB folds of HIF2 and ARNT show antiparallel

orientation and a -sheet interface8 65. These different interfaces may contribute to the specificity of binding and partner

selection, and they present different surfaces for specific interactions with accessory signalling proteins or transcription

co-regulators. The period proteins, which lack a bHLH, show promiscuous binding with other family members and, similar

to ARNT, can homodimerize172. Nuclear receptor co-activator 1 (NCOA1) is one of a small class of co-activator proteins that

contain typical PAS folds but that do not seem to function as obligate heterodimers. Structures shown: HIF2 PASB, ARNT

PASB (Protein Data Bank (PDB) ID: 4GHI); CLOCK PASB, BMAL1 PASB (PDB ID: 4F3L); PER2 PASB (PDB ID: 3GDI); NCOA1

PASB (PDB ID: 1OJ5). ARC, activity-regulated cytoskeleton-associated protein; BDNF, brain-derived neurotrophic factor;

CME, central midline element; CRY, cryptochrome; CYP1A1, cytochrome P4501A1; EGR1, early growth response 1; EPO,

erythropoietin; GLUT1, glucose transporter 1; GST, glutathione S-transferase; HRE, hypoxia response element; PER,

period circadian protein homologue; ROR, retinoic acid-related orphan receptor; VEGF, vascular endothelial growth factor;

XRE, xenobiotic response element.

Reproduced from Paper I Bersten et al, Nat Rev Cancer. 2013.

16

bHLH-PAS dimerization, DNA binding and signal transduction mechanisms

The bHLH-PAS class of proteins can be generally separated into transcription factors or

coactivator proteins which contain a shared domain structure (Kewley, Whitelaw et al. 2004;

Partch and Gardner 2010). bHLH-PAS proteins are characterised by an invariant N-terminal

bHLH domain which is involved in primary dimerisation and DNA binding (Reisz-Porszasz,

Probst et al. 1994; Fukunaga, Probst et al. 1995) followed by a PAS domain which

encompasses two PAS repeats denoted PAS A and PAS B, and a PAC domain (Ponting and

Aravind 1997) or S2 boxes (Zhulin, Taylor et al. 1997) which are structurally associated with the

PAS domain (Hefti, Francoijs et al. 2004). The bHLH-PAS transcription factors are further

divided into class I or class II factors, to form functional transcriptional complexes (Figure 1).

Class I factors (AhR, HIF1-3 , Sim1-2, NPAS1-4, and CLOCK) must heterodimerise with class II

factors (ARNT, ARNT2, BMAL1, BMAL2) to enable DNA binding and transcriptional activation.

Unlike the bHLH and bHLH/Zip transcription factors which bind the classic CANNTG E-box, the

bHLH-PAS transcription factors utilize an atypical E-box which  has  variable  5’  sequence

(Massari and Murre 2000; Kewley, Whitelaw et al. 2004). It has been shown in vitro and in

cultured cells that class II factors such as ARNT can homodimerise to activate canonical E-box

elements, although the biological significance of this has not been addressed (Sogawa, Nakano

et al. 1995; Swanson and Yang 1999).

The PAS A domain of bHLH-PAS transcription factors also contributes to dimerisation strength

(Erbel, Card et al. 2003; Chapman-Smith, Lutwyche et al. 2004) as wells as mediating

dimerisation specificity between bHLH-PAS family members (Pongratz, Antonsson et al. 1998).

Interestingly, the PAS A domain has also been implicated in direct DNA contact and DNA

bending, downstream of the primary bHLH-major groove contact (Chapman-Smith and

Whitelaw 2006). While PAS B also provides a dimerisation interface between bHLH-PAS

17

proteins (Card, Erbel et al. 2005; Huang, Chelliah et al. 2012), in selected cases it also directly

binds ligands (Whitelaw, Pongratz et al. 1993; Scheuermann, Tomchick et al. 2009),

coactivator proteins (Partch, Card et al. 2009; Partch and Gardner 2010; Partch and Gardner

2011) and chaperone proteins to control signal transduction (Pongratz, Mason et al. 1992;

Coumailleau, Poellinger et al. 1995).

Importantly, the PAS domains can also contribute to target gene specificity, for example, in

Drosophila the bHLH/PAS members Trachealess (Trh) and Single Minded (SIM) are able to bind

to the same DNA recognition sequence in collaboration with the obligate bHLH-PAS partner

factor, Tango (Tgo; equivalent to Aryl Hydrocarbon Nuclear Translocator (ARNT) in mammals).

While the DNA response elements of these two bHLH/PAS heterodimers are identical, target

gene induction can be switched by swapping the PAS domains between Trh and SIM,

suggesting that in this case target gene specificity is conferred by the PAS domain (Zelzer,

Wappner et al. 1997). How they achieve this is unclear, but may involve ancillary factors

recruited by the PAS domains or direct contact between the PAS domain and DNA elements

adjacent to the primary bHLH contact (Kimura, Weisz et al. 2001; Pearson, Watson et al. 2012).

This is somewhat supported by the ability of some bHLH-PAS transcription factors to bind to

the same DNA binding motifs yet have distinct target gene profiles or the capacity to bind to

several variants of the asymmetric E-box core (Swanson, Chan et al. 1995; Woods and

Whitelaw 2002; Ooe, Saito et al. 2004; Jiang and Crews 2007; Farrall and Whitelaw 2009;

Bersten, Sullivan et al. 2013).

The DNA binding specificity of bHLH-PAS transcription factors is conferred by the basic region

within the bHLH (Davis and Weintraub 1992). As with other bHLH containing transcription

factors, basic residues make direct contact with nucleotides to guide DNA binding specificity

(Ma, Rould et al. 1994). bHLH-PAS transcription factors utilising ARNT or ARNT2 bind to a

asymmetric E-Box like element where Class I and Class II factors contribute to three nucleotide

18

half site specificity. For example AhR/ARNT heterodimers bind to a Xenoboitic Response

Element (XRE) (TNGCGTG) where the AhR binds to NGC half site and ARNT binds to GTG half

site. In this context the ARNT DNA binding specificity is fixed to GTG whereas the class I

partner protein sequence specificity can vary. Although ARNT2 can also bind to many of the

class I factors the DNA binding specificity appears to be the same as for ARNT. This is in

contrast to various heterodimers formed with either BMAL1 or BMAL2, or BMAL homodimers

or ARNT homodimers, all of which bind to canonical E-Box CACGTG (Hogenesch, Gu et al.

1998; Reppert and Weaver 2002).

Many of the class I bHLH-PAS transcription factors dimerise with a common class II factor,

which can create competition for that Class II factor within the cell (Chan, Yao et al. 1999;

Woods and Whitelaw 2002). In addition, overlapping DNA binding specificities can allow for

further competition between heterodimeric complexes (Woods and Whitelaw 2002; Farrall

and Whitelaw 2009). Transcriptionally inactive or DNA binding defective forms of Class I bHLH-

PAS transcription factors can also compete for Class II factors to regulate the output

transcriptional pathways (Mimura, Ema et al. 1999; Makino, Cao et al. 2001; Makino, Kanopka

et al. 2002). The extent to which cross talk occurs in vivo or in pathological scenarios is yet to

be explored, however it is expected that ARNT levels will be limiting, supporting the concept of

competition for ARNT by class I bHLH-PAS factors (Semenza, Jiang et al. 1996; Holmes and

Pollenz 1997).

While the bHLH-PAS transcription factors primarily act to modulate gene expression by direct

DNA binding, alternate functions for the bHLH-PAS transcription factors have also been

described. The most notable are the direct control of hypoxic induced translation by HIF2 by

incorporation into the translational machinery (Uniacke, Holterman et al. 2012) and the

regulation of ubiquitin ligase protein degradation mediated by ligand activated AhR (Ohtake,

Baba et al. 2007; Kawajiri, Kobayashi et al. 2009).

19

bHLH-PAS transcription factor function

bHLH/PAS proteins are involved in a diverse array of physiological and pathological processes,

such as the cellular response to hypoxia (Hypoxia Inducible Factors

(HIF1 /HIF2 /HIF3 IPAS)), the maintenance of circadian rhythms (circadian locomotor

output cycles kaput (CLOCK), Period (PER1/PER2/PER3),Neuronal PAS domain protein 2

(NPAS2)), the response to environmental pollutants (Aryl hydrocarbon Receptor (AhR)/ Dioxin

receptor (DR), AhR repressor AhRR), neurogenesis and lung development (NPAS1, and NPAS3)

and appetite control (Single minded 1 (Sim1))(Reppert and Weaver 2002; Kewley, Whitelaw et

al. 2004; Semenza 2012).

In addition to canonical roles in normal physiology, emerging roles in T cell immunity, cancer

progression and metastasis, diabetes, and behaviour for bHLH-PAS transcription factors are

being discovered (Gunton, Kulkarni et al. 2005; Marcheva, Ramsey et al. 2010; Dang, Barbi et

al. 2011; Semenza 2012; Bersten, Sullivan et al. 2013; Hao and Whitelaw 2013). Normal and

cancer related functions of the bHLH-PAS transcription factors was reviewed recently in Paper

I (Bersten, Sullivan et al. 2013).

Neuronal PAS Factor 4

The mammalian NPAS4 gene (also known as Limbic enhanced-PAS (lePAS and NXF), was

discovered from screening a human fetal brain cDNA library with a PAS domain oligonuceotide

probe (Ooe et al, 2004). Human, rat and mouse NPAS4 were cloned and have significant

degree of homology in the bHLH and PAS domains to each other and to drosophila gene

Dysfusion (Dys), Zebrafish (Danio rerio) NPAS4 and the Caenorhabditis Elegans (C.Elegans)

gene C15C8.2 (Cky-1) (Jiang and Crews 2003; Ooe, Saito et al. 2004; Ooe, Saito et al. 2007;

Ooe, Saito et al. 2009; Baxendale, Holdsworth et al. 2012). While NPAS4 is clustered in the

bHLH-PAS transcription factor family phylogenetic analysis of mouse PAS domain containing

20

proteins suggests it has little similarity to any other bHLH-PAS factors (Ooe, Saito et al. 2004;

McIntosh, Hogenesch et al. 2010). This is also observed in C.elegans and drosophila (Jiang and

Crews 2003; Ooe, Saito et al. 2007). Within the bHLH and PAS A domain NPAS4 is most similar

to AhR/SIM2/NPAS1 (43%, 33% and 33% amino acid identity respectively) and within in the

PAS B domain is most similar to HIF1 /PER/BMAL1 (29%/27%/24% amino acid identity).

While the C-terminal region of the mammalian NPAS4 members is highly conserved, there is

little or no sequence homology to either C.elegans (Cky-1), Drosphila (Dys), or any of the

mammalian bHLH-PAS C-termini.

Drosophila Dysfusion

Drosophila SIM and trachealess (Trh) control transcription and development in the CNS

midline cells and tracheal cells, respectively, and both appear to do so through a common DNA

regulatory element known as the Central Midline Element (CME; ACGTG)(Crews 1998). The

NPAS4 related gene Dys was discovered by screening for bHLH containing proteins in

drosophila (Ledent and Vervoort 2001; Peyrefitte, Kahn et al. 2001). Expression of Dys was

subsequently shown in embryonic tissues such as foregut precursors, CNS cells which are

either part of the medial brain or frontal ganglion, tracheal fusion cells, the epidermal leading

edge, the hind gut and the anal pad from stage 11-12 of development (Peyrefitte, Kahn et al.

2001; Jiang and Crews 2003). Dys is expressed prior to tracheal fusion events and RNAi

knockdown or loss of function (LoF) Dys mutants results in aberrant tracheal fusion (Jiang and

Crews 2003; Jiang and Crews 2006). Development of the tracheal system is critical to the

complete formation of the oxygen delivering tubular network (Ghabrial, Luschnig et al. 2003)

and as such, inhibition of tracheal fusion by loss of Dys results in a lethal phenotype (Jiang and

Crews, 2003). The tracheal fusion cells where Dys is expressed extend actin-rich filopodia,

which are guided by cues in a similar fashion to that of axon growth cone guidance (Tanaka-

Matakatsu, Uemura et al. 1996; Englund, Steneberg et al. 2002; Ribeiro, Ebner et al. 2002).

21

Tracheal branches undergo systematic migration from adjacent hemi-segments until they are

in close enough contact for the fusion cells to adhere to each other. Over expression of Dys

throughout the trachea results in inhibition of migration and ectopic branch fusion, implicating

Dys in the migration and fusion of actin-rich filopodia of the Drosophila tracheal system (Jiang

and Crews 2006). Like other members of the Drosophila bHLH-PAS family, Dys appears to

heterodimerise with Tgo to regulate target genes, in this case shotgun, CG13196, and

members only, all of which contribute to cell adhesion or other aspects of tracheal fusion or

migration (Wilk, Weizman et al. 1996; Jiang and Crews 2003; Jiang and Crews 2006). However,

due to the low degree of sequence similarity between NPAS4 and Dys, especially in the C-

terminus, and the absence of any reported expression of NPAS4 in the lung, NPAS4 may have

diverged in function in mammals.

NPAS4 Expression and Regulation

NPAS4 expression is highly restricted to the brain in rodents and humans, with a several

reports of weak expression within the testes (Flood, Moyer et al. 2004; Moser, Knoth et al.

2004; Ooe, Saito et al. 2004; Ramamoorthi, Fropf et al. 2011). Within the CNS, NPAS4

expression appears to be further restricted to neurons within the hippocampus, cortex,

olfactory bulb, and amygdala with very low or no expression seen in other brain areas or the

spinal cord (Moser, Knoth et al. 2004; Lin, Bloodgood et al. 2008; Ramamoorthi, Fropf et al.

2011). Unlike NPAS3 or Sim2, NPAS4 does not appear to be expressed in the developing

mouse embryo. Appreciable levels begin to appear perinatally and increase as SIM2 gradually

decreases up to 4 weeks post natal (Brunskill, Witte et al. 1999; Ooe, Saito et al. 2004).

A common theme among bHLH-PAS transcription factors is signal or stress responsiveness, for

example with HIF activated in hypoxia, AhR responsive to xenobiotics and SIM1 responsive to

appetite signals. The NPAS4 gene is also a signal or stress responsive, as expression has been

shown to be upregulated by seizure (Flood, Moyer et al. 2004; Baxendale, Holdsworth et al.

22

2012) ischemia (Shamloo, Soriano et al. 2006; Leong, Klaric et al. 2013), several

psychostimulants and opioid drugs (Piechota, Korostynski et al. 2010; Guo, Xue et al. 2012;

Martin, Jayanthi et al. 2012). In addition, physiological stimuli such as contextual fear

conditioning (CFC) (Ploski, Monsey et al. 2011; Ramamoorthi, Fropf et al. 2011), light

stimulation (Lin, Bloodgood et al. 2008; Maya-Vetencourt, Tiraboschi et al. 2012) or neuronal

depolarisation (Lin, Bloodgood et al. 2008; Ramamoorthi, Fropf et al. 2011) induce expression

of NPAS4.

At a mechanistic level these stimuli are thought to act primarily through stimulation of voltage

gated calcium channels which initiates, calcium dependent gene expression. NPAS4 induction

can be blocked by calcium chelation, calcium channel blockade, NMDA receptor blockade or

AMPA receptor blockade but does not appear to be upregulated by neurotrophin signals (Lin,

Bloodgood et al. 2008; Ramamoorthi, Fropf et al. 2011). Neuronal depolarisation results in an

extremely rapid (within 5 mins for CFC), robust and transient increase in NPAS4 mRNA and

protein through activation of VDCC channels in a protein synthesis independent manner

(Ramamoorthi, Fropf et al. 2011). NPAS4 induction following depolarisation peaks at 0.5-2 hrs

post depolarisation and is rapidly down regulated to near basal levels by 4-6hrs (Lin,

Bloodgood et al. 2008; Ramamoorthi, Fropf et al. 2011).

Unlike other genes transiently stimulated by neuronal activity, NPAS4 protein is not activated

by neurotrophic factors, growth factors, or forskolin (a protein kinase A activator) making it

uniquely situated to respond only to learning and memory cues (Lin, Bloodgood et al. 2008;

Ramamoorthi, Fropf et al. 2011; Ebert and Greenberg 2013). While the molecular mechanisms

underlying both the restricted brain and activity-dependent expression of NPAS4 remain

unexplored, it is clear that the latter requires calcium dependent signalling events. Activation

of NPAS4 expression, which has been suggested to be mediated by calmodulin dependent

kinase IV (CaMK IV) and cAMP-response element binding protein (CREB), can be blocked by

23

inhibition of nuclear calcium signalling (Greer and Greenberg 2008; Zhang, Zou et al. 2009). In

addition, it has also been suggested that PI3K and ERK/MAPK may also play a role in

upregulation of NPAS4 following NMDA receptor activation (Coba, Valor et al. 2008; Ooe,

Kobayashi et al. 2009). Activity dependent up regulation of NPAS4 is initially protein synthesis

independent and may involve other activity induced transcription factors such as CREB and

MEF2A/D (Flavell, Kim et al. 2008; Ramamoorthi, Fropf et al. 2011; Ebert and Greenberg

2013). Following the initial up regulation of NPAS4 following depolarisation, it’s  proposed  that

NPAS4 may strengthen its own expression through a feed forward pathway. Indeed, NPAS4

has been shown to bind and activate its own promoter, however the ramifications of this are

unknown (Ooe, Saito et al. 2004; Lin, Bloodgood et al. 2008; Kim, Hemberg et al. 2010). While

there is cursory information about stimulus dependent activation, developmental regulation

of NPAS4 expression has not been investigated. Furthermore, it is not clear how NPAS4

protein and mRNA are down regulated following initial induction.

In other contexts an increase of intracellular calcium through endoplasmic reticulum stressors

can also induce NPAS4 expression (Ooe, Motonaga et al. 2009). Recently it has been shown

that pancreatic beta cells are capable of depolarisation induced NPAS4 expression (as

demonstrated by qRT-PCR and immunofluorescence experiments), indicating that NPAS4 may

act more broadly outside the nervous system as a calcium induced transcription factor in a

limited number of tissues (Sabatini, Krentz et al. 2013). However, the pervasiveness or

consequences of NPAS4 upregulation outside the nervous system or in response to ER stress

or calcium induction have yet to be fully elucidated.

Very little is known about the mechanisms which regulate NPAS4 expression, however

recently it has been shown that the NPAS4 promoter is methylated at several CpG sites and

mRNA expression may be repressed by this methylation (Rudenko, Dawlaty et al. 2013). Ten-

eleven translocation 1 (Tet1) is a dioxygenase enzyme which catalyses the addition of a

24

hydroxy group onto methylated CpGs to mediate demethylation (Wu and Zhang 2011). It has

been proposed that demethylation may be important for regulation of activity inducible genes

and indeed knockout of Tet1 in mice leads to hypermethylated and suppressed NPAS4

expression in the hippocampus and cortex (Kaas, Zhong et al. 2013; Rudenko, Dawlaty et al.

2013). Interestingly, all bHLH-PAS transcription factor binding sites contain a CpG dinucleotide

between the Class I and Class II half sites which could potentially be methylated, however the

effect of non-methyl versus methyl-CpG bHLH-PAS response elements on binding of

heterodimers has not been explored.

NPAS4–dependent Transcriptional Regulation

NPAS4 is able to heterodimerise with the partner proteins ARNT and ARNT2 to bind to many

bHLH-PAS regulatory sequences by in vitro DNA binding assays and luciferase reporter

experiments (Ooe, Saito et al. 2004). Due to overlapping expression between NPAS4 and

ARNT2 within the CNS, and the suggestion that NPAS4 may interact more strongly with ARNT2

than ARNT, NPAS4 is anticipated to function in vivo with ARNT2 (Hirose, Morita et al. 1996;

Aitola and Pelto-Huikko 2003; Ooe, Saito et al. 2004; Ooe, Saito et al. 2009). bHLH-PAS

regulatory elements which NPAS4/ARNT2 have been shown to bind contain the core CME /

HRE (ACGTG) or the XRE (TNGCGTG) (Moser, Knoth et al. 2004; Ooe, Saito et al. 2004),

however the preferred DNA binding sequence of NPAS4/ARNT2 heterodimers are unknown.

Comparison of activities of NPAS4 on bHLH-PAS responsive reporters and in vitro DNA binding

suggests that NPAS4 binds most strongly to TCGTG and GCGTG containing sites (Ooe, Saito et

al. 2004). Analysis of NPAS4 Chromatin immunoprecipitation -sequencing (ChIP-Seq)

experiments in depolarised mouse cortical neurons, together with in vitro site selection

experiments to identify the preferred NPAS4/ARNT DNA binding motif, also supports this

notion (Kim, Hemberg et al. 2010)( D.C. Bersten, V. Bhakti, and M.L. Whitelaw unpublished

results).

25

Initial functional experiments suggested that NPAS4 may control expression of the cytoskeletal

binding protein Drebrin and the apoptosis inducing gene Bax (Ooe, Saito et al. 2004; Hester,

McKee et al. 2007). However, examination of NPAS4 dependent target genes during neuronal

depolarisation suggest that these maybe either context dependent or artefacts of

overexpression in cell lines (Lin, Bloodgood et al. 2008; Kim, Hemberg et al. 2010; Pruunsild,

Sepp et al. 2011; Ramamoorthi, Fropf et al. 2011). ChIP-seq and microarray analysis of NPAS4

bound and differentially regulated target genes following KCl-depolarisation has been

performed and reveals extensive occupancy of NPAS4 at promoters and enhancers to regulate

many activity dependent genes involved in inhibitory synapse development (Lin, Bloodgood et

al. 2008; Kim, Hemberg et al. 2010; Ramamoorthi, Fropf et al. 2011; Bloodgood, Sharma et al.

2013). NPAS4 occupancy at enhancer sites correlates well with CBP occupancy and supports

existing data suggesting CBP/p300 acts as an NPAS4 transactivator (Ooe, Saito et al. 2004; Kim,

Hemberg et al. 2010). A number of confirmed NPAS4 target genes have been examined

including BDNF, c-Fos, Egr1 and Arc (Lin, Bloodgood et al. 2008; Kim, Hemberg et al. 2010;

Pruunsild, Sepp et al. 2011; Ramamoorthi, Fropf et al. 2011). It is proposed that NPAS4

predominantly acts through enhancer recruitment of RNA Pol II to activate transcription of its

target genes (Kim, Hemberg et al. 2010; Ramamoorthi, Fropf et al. 2011). NPAS4 binds directly

to BDNF promoter I, promoter III and promoter IV to act as one of the main drivers of activity

induced expression of BDNF (Lin, Bloodgood et al. 2008; Pruunsild, Sepp et al. 2011;

Bloodgood, Sharma et al. 2013). Deletion of promoter IV NPAS4 response element

significantly weakens activity induced expression of BDNF although it does not alter the initial

responsiveness of BDNF to depolarisation (Pruunsild, Sepp et al. 2011). It has therefore been

proposed that NPAS4 acts to enhance or strengthen activity induced gene expression

(Pruunsild, Sepp et al. 2011; Ramamoorthi, Fropf et al. 2011). In addition, it has been

suggested by several groups that NPAS4 may act directly on its own promoter to strengthen its

activation (Ooe, Saito et al. 2004; Lin, Bloodgood et al. 2008). Importantly, ARNT2 has also

26

been shown to bind to BDNF promoters and regulate BDNF expression, although the extent of

its contribution to activity dependent or NPAS4 dependent gene expression remains

unexplored (Pruunsild, Sepp et al. 2011).

NPAS4 Function Activity-Dependent Transcription in the CNS

Neuronal activity plays key roles in neuron survival, proliferation, migration, axon/dendrite

guidance and growth, synapse formation, maintenance and elimination, and neurotransmitter

specification (Spitzer 2006; Greer and Greenberg 2008; Spitzer 2012). These processes guide

neuronal activity profiles in the prospective central nervous system and underlie the cellular

and molecular basis of memory and learning. In addition, genetic aberrations in neuronal

activity are thought to be critical to many neuropsychiatric disease states and may be the

underlying cause of autism and schizophrenia (Ramocki and Zoghbi 2008; Ebert and Greenberg

2013). Importantly, over excitation due to disruption of neuron homeostasis may also

underlie the pathogenesis of many neurodegenerative disorders such as dementia,

Alzheimer’s disease, stroke and seizure(Saxena and Caroni 2011).

Sensory experience drives synaptic plasticity and learning in animals. One seminal experiment

outlining this phenomena showed that ocular occlusion during critical periods around birth,

when neuron development and synapse formation occur, leads to cortical circuit

rearrangements to favour the non-occluded eye (Wiesel 1982). It is now well established that

sensory experience and neuronal activity drive synaptic plasticity during critical periods and

throughout adulthood. Electrical activity in neurons can take many forms, of which the best

established being activation of voltage-gated calcium channels leading to action potentials

(Greer and Greenberg 2008). During early stages of neuron development GABA and glycine

are able to generate depolarising chloride currents due to efflux of intracellular chloride ions

(Spitzer 2006). Later in neuron development this phenomena is reversed and leads to chloride

27

influx and inhibition of neuronal activity. In addition, glutamate generates depolarisation via

activation of voltage gated calcium and sodium channels. Calcium influx through voltage

gated calcium channel triggers rapid activation of gene expression, which in turn acts to

modulate synapse formation and plasticity. Calcium activated gene expression in neurons is

mediated by a number of activity-dependent transcription factors including CREB, MEF2A/D,

USF1/2, NF B, MeCP2, SRF and NPAS4, many of which are regulators of synapse formation

and function (Greer and Greenberg 2008).

NPAS4 and synapse formation

Initial experiments in cultured hippocampal neurons showed that NPAS4 was able to regulate

the number and strength of inhibitory inputs on excitatory pyramidal neurons both

perisomatically and dendritically (Lin, Bloodgood et al. 2008). In contrast, NPAS4 seemed not

to have any effect on excitatory synapse number or dendritic patterning of the neurons (Lin,

Bloodgood et al. 2008). More recently, using conditional knockout NPAS4 mice where NPAS

has been sparsely ablated in CA1 hippocampal neurons, recording of inhibitory post synaptic

potentials in layers containing either dendrites or soma has revealed that NPAS4 has opposing

actions on inhibitory synapse formation in the two compartments, increasing inhibition within

the soma and decreasing inhibition in dendrites (Bloodgood, Sharma et al. 2013). The function

of compartment specific regulation of inhibitory synapses is as yet unclear, however, somatic

inhibition may be more effective at limiting action potential spikes than dendritic inhibition,

while increased activity in dendrites may aid more effective dendritic synapse plasticity

(Sylwestrak and Scheiffele 2013). The downstream mechanism(s) by which NPAS4 imparts

compartment specific neuron inhibition is also unclear. It has been shown that specific

transcripts of BDNF are regulated by NPAS4 binding and activation at promoter I, III and IV

(Lin, Bloodgood et al. 2008; Pruunsild, Sepp et al. 2011; Ramamoorthi, Fropf et al. 2011;

Bloodgood, Sharma et al. 2013). While BDNF has been shown to be involved in inhibitory

28

synapse formation (Marty, Wehrle et al. 2000; Lin, Bloodgood et al. 2008) and some

transcripts can be can be targeted to different cellular compartments (Pattabiraman, Tropea

et al. 2005; An, Gharami et al. 2008; Dean, Liu et al. 2012), the contribution of NPAS4 to

regulating subcellular levels of BDNF through transcript specific regulation is yet to be

addressed.

NPAS4 in memory, learning and behaviour.

The selective and transient induction of NPAS4 following sensory experience and neuron

depolarisation prompted several groups to investigate the function NPAS4 in memory and

learning in mice (Ploski, Monsey et al. 2011; Ramamoorthi, Fropf et al. 2011; Coutellier, Beraki

et al. 2012; Maya-Vetencourt, Tiraboschi et al. 2012). In independent mouse knockout models

NPAS4 has been shown to be involved in contextual memory formation in adult mice. Either

global or conditional knockout of NPAS4 in the CA3 region of the hippocampus attenuated

long-term memory formation in mice, which is dependent on NPAS4 transcriptional activity

(Ramamoorthi, Fropf et al. 2011; Coutellier, Beraki et al. 2012). However, short-term memory

was affected in only in the global NPAS4 knockout mice (Ramamoorthi, Fropf et al. 2011).

Additional behavioural deficits such as increased aggression and hyperactivity are observed in

the NPAS4 null mice and decreased contact with unfamiliar mice are observed in animals

heterozygous null for NPAS4 (Coutellier, Beraki et al. 2012). Both heterozygous and

homozygous null NPAS4 mice also have defects in sensorimotor gating which may reflect

impaired cognitive processing of sensory information (Coutellier, Beraki et al. 2012). Taken

together this outlines the importance of NPAS4 in sensory induced memory, learning and

cognitive abilities which results in behavioural and social defects in NPAS4 null mice.

29

NPAS4 in Neuroprotection

NPAS4 null mice have been shown to have much reduced life span, with only 20-30% of mice

surviving to 16 months (Ooe, Motonaga et al. 2009). This reduction in life span was associated

with a marked increase in age related neurodegeneration in knockout animals in addition to

increased susceptibility of NPAS4 null mice to glutamate induced neurotoxicity (Ooe,

Motonaga et al. 2009). It is therefore hypothesised that induction of NPAS4 expression may

confer neuroprotection during hyperexcitation or ischemia. This is supported by experiments

where overexpression of NPAS4 (Hester, McKee et al. 2007) increased cell survival or

knockdown decreased cell survival in post-mitotic neurons induced to undergo apoptosis

(Zhang, Zou et al. 2009). While priming post-mitotic neurons with elevated NPAS4 by

overexpression or increased neuron depolarisation may provide neuroprotection, sustained

overexpression in cultured cell lines appears to induce apoptosis (Hester, McKee et al. 2007).

Taken together, this suggests that moderate transient upregulation of NPAS4 maybe

protective while sustained high level NPAS4 expression maybe detrimental to cell survival.

30

Project Rationale

NPAS4 expression is mostly restricted to the brain and exhibits transient increases to dampen

neuronal activity by upregulating inhibitory synapse number and strength on excitatory

neurons. This homeostatic regulation of inhibitory synapses implicates NPAS4 in various

neuropsychiatric diseases including autism as well as neurodegenerative diseases in which

excitotoxicity plays a role in the aetiology of the disease. While these are of key interest in the

field, the mechanisms of regulation of NPAS4 gene expression, the potential contribution to

human disease and effective modelling of behaviour in mice are yet to be explored. These

aspects were therefore the focus of the research in this thesis, with the aims of;

1. Exploring the molecular mechanisms which control and restrict NPAS4 expression to

the brain

2. Investigating the contribution of human single nucleotide variants to the function of

the NPAS4/ARNT2 transcriptional dimer

3. Generating technological platforms to effectively model NPAS4 function in vitro and in

mice

31

Chapter 2: Results

2.1 Elucidation of molecular mechanisms, which restrict NPAS4 expression to the brain.

NPAS4 is an activity dependent bHLH-PAS transcription factor that controls the expression of

genes involved in inhibitory synapse formation and maintenance. NPAS4 expression is

predominantly restricted to cortical and hippocampal regions of the brain and is highly

induced following neuronal depolarisation, paradigms of learning, seizure or ischemia. The

molecular mechanisms underlying restricted expression pattern were addressed in paper II,

entitled  “  Regulation of the neuronal transcription factor NPAS4 by REST and microRNAs”.

For the study in paper II we generated antibodies against NPAS4 protein and confirmed both

activity-induced expression in primary cultured neurons and restriction of mRNA expression to

predominantly cortical/hippocampal brain regions. Paper II identified conserved binding sites

for the transcriptional repressor RE-1 silencing transcription factor (REST) within the promoter

and Intron I of the NPAS4 gene. We found REST to be strongly bound to these sites to repress

NPAS4 expression in non-neuronal and undifferentiated cells. Deletion of these REST binding

sites derepressed NPAS4 expression in embryonic stem cells, and reporter gene experiments

in HEK293T cells showed that both intron I and promoter RE-1 sites are important for

repression. We also showed that REST binding correlates with CTCF occupancy, known to

induce DNA looping and mediate insulator function, this suggesting that CTCF may aid

silencing of NPAS4.

In addition, we also showed that the highly conserved 3’UTR of NPAS4 might also play a role in

regulating NPAS4 expression. We established that NPAS4 can be repressed by miR-224 and

32

miR-203 and went on to show that miR-224 is enriched with the hypothalamic/midbrain

regions; being expressed from an intron of the GABA A receptor epsilon gene along with miR-

452. We propose that miR-224 may fine tune NPAS4 expression to restrict expression to

cortical regions of the brain. The work describes the first mechanistic explanation of the brain-

restricted expression of NPAS4.

2.2 Human variants in the NPAS4/ARNT2 transcriptional complex can disrupt function

Neuropsychiatric disease has been suggested to derive, in part, from defects which disrupt

neuronal excitation/inhibition balance (Rubenstein and Merzenich 2003; Kehrer, Maziashvili et

al. 2008; Rubenstein 2010; Yizhar, Fenno et al. 2011). Loss of function mutations in synapse

effectors also suggests synapse dysfunction may play a key role in disrupting this homeostatic

balance in these diseases (Ramocki and Zoghbi 2008; Sudhof 2008; Ebert and Greenberg

2013). In addition, glutamatergic hyperexcitation, which may also arise from disruption of the

excitatory/inhibitory balance or other neuronal stressors, may promote the progression of

neurodegenerative disorders (Bezprozvanny and Mattson 2008; Saxena and Caroni 2011).

Given the links between NPAS4 function and neurological disease like phenotypes in mice, we

sought to investigate whether non-synonymous single nucleotide variants found within human

NPAS4, or its heterodimeric partner protein ARNT2, may disrupt function of the dimer. If so,

we reasoned that a rationale for exome sequencing of NPAS4 in neurological disease affected

patient cohorts might emerge.

In paper III,  “Human  Variants  in  the  Neuronal  Basic  Helix-Loop-Helix/Per-Arnt-Sim (bHLH/PAS)

Transcription  Factor  Complex  NPAS4/ARNT2  Disrupt  Function” we found that several variants

in NPAS4 or ARNT2 were able to reduce or ablate a bHLH-PAS responsive luciferase reporter

gene. We also showed that one variant, NPAS4.F147S, failed to upregulate endogenous BDNF

expression due to disruption of dimerisation with ARNT2. We also showed that an ARNT2

33

variant, ARNT2.R46W, had reduced activity due to disrupted nuclear localisation. Structural

modelling of the bHLH-PAS regions of the NPAS4/ARNT2 heterodimer, based on the related

CLOCK/BMAL structure (Huang, Chelliah et al. 2012), indicated that the conserved F147

residue lies at the PAS A dimer interface and may directly participate in heterodimer

formation. This is also supported by the finding that mutations which weaken AhR-ARNT

dimerisation lie within this region (Hao, Whitelaw et al. 2011). Paper III shows the potential for

de novo discovery of disease relevant variants by selecting polymorphisms from databases for

screening in function based assays.

2.3 Generation and validation of Inducible and reversible systems for lentiviral and ES cell (Col1a1-RMCE) mediated manipulation of NPAS4 levels.

Inducible and reversible systems for modulating gene expression are critical for investigating

biological pathways. These systems are of particular importance when studying genes critical

for essential cell processes. For example, removal or knockdown of genes critical for cell

survival or proliferation results in cell death, making generation of animal models and cell lines

redundant. In addition, inducible and reversible manipulation of gene expression is of

particular interest in neuroscience, immunology, developmental and cancer biology disciplines

where induction and reversal of genes proposed to underpin a phenotype would be

advantageous for studying disease progression or biological processes.

Several tools for inducible and reversible control of gene expression have been developed in

the past and have been used with limited success. The best studied example being the

tetracycline inducible system combined with either shRNA-mediated knockdown or cDNA

mediated ectopic or over expression. The use of the tetracycline inducible system has been

hampered for several technical reasons. Initial systems employed tetracycline repressor TetR

variants, which repressed transcription from a constitutive promoter containing tet operator

34

elements, but these systems were generally very leaky in cell lines and animals. While the use

of Tet activator variants and modified tet operator elements reduced background expression

in the absence of doxycycline, most of the variants still have poor sensitivity to Dox, which is a

limitation for in vivo experiments, eg in tissues where DOX transport is poor (i.e. brain). Most

currently used systems still contain significant background expression and generally the tet

responsive promoter and tet activator are delivered on separate plasmids, leading to mosaic

inducible expression in cell lines.

Reliable and reproducible generation of inducible cell lines and mice has been inadequate due

to technical limitations such as cell lines being refractory to transfection and, particularly in

mice, random integration effects. Furthermore, the use of shRNA systems for use in

generating inducible and reversible knockdown in cells and mice relies heavily on efficient

selection of potent shRNA sequences, which are able knockdown gene expression in single

copy.

In manuscript IV,  “Inducible and reversible lentiviral and recombination mediated cassette

exchange (RMCE) systems for controlling gene expression”,  we attempt to overcome many of

the limitations of previous inducible and reversible systems for ectopic expression and

knockdown of gene expression. We design and test single inducible and reversible lentiviral

constructs for reliable generation of cell lines, which utilise Tet activator variants with

dramatically improved sensitivity to doxcycline and no detectable background expression.

These constructs allow for uniform inducible and reversible expression in cell lines. In

addition, we overcome limitations of reliable targeting into embryonic stem cells for the

generation of mice by adapting a Recombination Mediated Cassette Exchange (RMCE) strategy

which results in ~100% efficient targeting of single inducible and reversible constructs into the

Col1a1 locus. We also design and test constructs which allow for tissue specific Dox inducible

35

expression either by Cre mediated gain of Tet activator expression or using tissue specific

promoters.

Using published high throughput data which tested the efficiency of ~60,000 shRNA

sequences, we built an algorithm to predict shRNA sequences with high knockdown potency.

We demonstrate that this is a potentially useful tool for predicting shRNA targets by selecting

and testing sequences for knockdown of ARNT and ARNT2, using the single integration RMCE

embryonic stem cells. We find that the majority of predicted shRNA sequences against either

ARNT or ARNT2 are able to efficiently knockdown endogenous protein expression in

embryonic stem cells.

36

Chapter 3: Discussion

NPAS4 has recently been revealed as a critical gene in the regulation and maintenance of

neuronal activity in excitatory neurons. NPAS4 protein and mRNA are transiently upregulated

by neuronal activity to homeostatically regulate inhibitory synapse inputs, preferentially in the

soma of neurons, to dampen neuronal activity. The signal regulated nature of NPAS4 has many

similarities with other bHLH-PAS domain containing proteins such as the AhR, which is signal

regulated by binding to polyaromatic hydrocarbons, resulting in up regulation of genes

involved in cellular detoxification. In response to hypoxic signals, the HIF proteins are

stabilised and become transcriptionally active to regulate genes, which respond to low oxygen

tension by altering metabolism and promoting increased angiogenesis. Responding to

physiological or stress signals to induce genetic programs for protection, adaption or

maintaining homeostasis seems to be unifying theme for bHLH-PAS transcription factors.

Tight temporal and spatial control of NPAS4 expression is likely to be critical for the correct

specification of inhibitory synapses and the homeostatic activity balance of neurons within the

CNS. Sustained overexpression in cell lines and by pathological inducers such as seizure and

ischemia may indeed be detrimental to neuronal cellular function. Therefore the precise

mechanisms which control NPAS4 expression are likely to be highly regulated and may have

importance in disease states. In paper II we describe a mechanism that restricts NPAS4

expression to the brain by finding the repressor REST silences expression in non-neuronal and

undifferentiated tissues (Figure 2.). REST binds strongly to multiple sites within the NPAS4

gene to mediate silencing of NPAS4 expression outside the CNS, implying that repression of

NPAS4 expression may be functionally important during development. Silencing the NPAS4

gene outside the CNS may also be aided by CTCF, which flanks the NPAS4 gene and overlaps

37

with REST binding sites, perhaps implying that looping may enhance silencing. This possibility

remains to be explored. Interestingly, CTCF binding at sites flanking the NPAS4 gene has been

reported for brain tissue and CTCF may act to insulate activity inducible expression of NPAS4

from adjacent genes, which are not depolarisation regulated.

We also show that miR-224 may play role in restricting NPAS4 expression to cortical brain

regions by targeting the NPAS4  3’UTR (Paper II). Given that NPAS4 can influence the number

of inhibitory inputs, miR-224 may allow for an altered activity profile in certain neuronal

subtypes and it would therefore be of interest to investigate the in vivo function of loss of miR-

224 in mice. Interestingly, we found that inhibition of both methyltransferases and histone

deacetylases in embryonic stem cells leads to an upregulation of NPAS4 expression, which is

supported by recent evidence that the NPAS4 promoter is methylated and regulated by the

demethylase promoting enzyme Tet1 (Rudenko, Dawlaty et al. 2013). While the mechanism of

how Tet1 is regulated in response to neuronal activation in unknown, it has been shown that

the methyl-CpG-Binding protein 2 (MeCP2) can be phosphorylated in response to

depolarisation, leading to dissociation of MeCP2 from methyl-CpGs and corepressors. This may

allow Tet1 access to catalyse demethylation and depression of transcription (Chen, Chang et

al. 2003; Ebert, Gabel et al. 2013).

38

Figure 2. Schematic diagram of the regulation of NPAS4 expression by REST and miR-224. RE-

1 Silencing Transcription factor (REST) represses NPAS4 expression in embryonic stem cells

and non-neuronal cells by binding to multiple RE-1 sites in the promoter and intron I of the

NPAS4 gene. In unstimulated primary neurons, REST expression is down regulated

derepressing NPAS4 expression. microRNA 224 (miR-224) is expressed in midbrain and

hypothalamic brain regions and is able to repress NPAS4 expression, which may restrict NPAS4

expression to cortical and hippocampal neurons. Upon depolarisation of neurons, NPAS4

expression rapidly increases and binds its own promoter, which may further increase its own

expression. Reproduced from paper II Bersten et al, Biochim Biophys Acta. 2014.

Significant questions still remain in the regulation of NPAS4 expression. For example, while

strong binding of NPAS4 to its own promoter is present following depolarisation and NPAS4

has been shown to activate its own promoter in luciferase reporter gene assays, it is still

unclear what involvement NPAS4 has in stimulating its own expression ((Ooe, Saito et al.

2004)Paper II). For example, initial upregulation of NPAS4 mRNA following depolarisation is

unhindered by cyclohexamide treatment, implying that other proteins may be responsible for

the initial wave of NPAS4 expression (Ramamoorthi, Fropf et al. 2011). In addition, Cre

mediated knockout of NPAS4 in neurons does not affect depolarisation mediated NPAS4

promoter driven luciferase induction (Ramamoorthi, Fropf et al. 2011), but induction kinetics

have not been investigated in NPAS4 knockout GFP reporter mice . As such there is a critical

need to identify proteins, which initiate the transient induction of NPAS4 following neuronal

depolarisation. As with the BDNF promoter, NPAS4 may not be responsible for the initial

induction of its own expression, but rather strengthening the response after gene induction

(Pruunsild, Sepp et al. 2011). In addition, NPAS4 protein is rapidly down regulated after being

39

initially induced by depolarisation, peaking in expression at ~2 hrs and almost absent after

6hrs. The mechanism by which this down regulation occurs remains unexplored.

Elucidation of NPAS4 function has revealed a critical homeostatic role in regulating neuronal

activity by controlling the balance between excitatory and inhibitory synaptic inputs. Mice

lacking NPAS4 have many phenotypic characteristics similar to those seen in both

neuropsychiatric diseases such as autism and schizophrenia and in neurodegenerative disease.

Hypotheses for the aetiology of both of these disease states include dysfunction in

excitatory/inhibitory balance within the brain, leading to cognitive dysfunction. At least in the

case of autism like behaviours, this has been elegantly shown in mice using optogenetic

techniques to disrupt excitatory/inhibitory balance in cortical neurons (Yizhar, Fenno et al.

2011). For aforementioned reasons NPAS4 and the regulation of NPAS4 expression may have a

role either directly or indirectly in these diseases.

In Paper III we investigate the potential contribution of human non-synonymous single

nucleotide variants on the function of the NPAS4/ARNT2 transcriptional complex. We find that

several naturally occurring variants have the ability to disrupt NPAS4/ARNT2 transcriptional

function and find biochemical mechanisms for the defective complex include either

attenuating heterodimer formation or nuclear localisation. The location of variants showing

loss of heterodimer formation was consistent with previous loss of dimerisation mutations

discovered by bacterial two hybrid screening (Hao, Whitelaw et al. 2011). We also recently

published a collaborative work describing loss of function for single nucleotide variants in

SIM1, which were associated with severe obesity (Bonnefond, Raimondo et al. 2013;

Ramachandrappa, Raimondo et al. 2013). The location of loss of function variants described in

these manuscripts is consistent with our data in Paper II in that they are commonly clustered

N-terminally in domains critical for heterodimerisation, DNA binding and nuclear localisation.

40

Many of the variants described in paper III were discovered from high throughput genome

sequencing efforts, the implication of which is that many have not been independently

validated. In addition, while many of the variants studied in paper III are common variants

within the human population, those which display altered activity appear to be rare or de novo

variants. Anonymous donation of samples in these sequencing efforts makes linking genotype

to phenotype difficult. As such we cannot draw conclusions about the relevance of these

variants to disease, suffice to say that common variations do not appear to significantly alter

NPAS4 function in reporter experiments.

While links with disease phenotype was beyond the scope of Paper III, it suggests that the

identified loss of function variants may have implications for neuropsychiatric or

neurodegenerative disease. Furthermore, it also outlines the potential use of de novo

functional screening of human variants in genes with already defined roles in disease.

Following our observations regarding the ability of naturally occurring variants to disrupt

NPAS4/ARNT2 function outlined in paper III and the tightly regulated mechanisms to control

NPAS4 expression outlined in paper II, we propose that screening for disease associated single

nucleotide and copy number variants within coding and regulatory regions of NPAS4 is

warranted, within neuropsychiatric or neurodegenerative disease populations.

As discussed above, bHLH-PAS transcription factors often act transiently to respond to signals

and perform protective or homeostasis related functions. Homeostatic regulation of key

signal response pathways is consistent with bHLH-PAS family members having important

developmental and disease related roles. Targeted disruptions of many of the bHLH-PAS genes

result in embryonic lethality and many have distinct roles in different tissues. It would

therefore be advantageous to study function following gene manipulation in an inducible and

reversible manner. Specifically, development of inducible, reversible and tissue specific gain or

41

loss of HIF1 / HIF2 mouse models would significantly aid the elucidation and distinction for

roles of the HIFs in cancer initiation, progression and metastasis. In addition, An Inducible and

reversible loss of NPAS4 mouse model would be invaluable to the advancement of the study of

the basic mechanisms underlying synapse formation, memory and learning formation and the

investigation and modelling of neuropsychiatric disease.

We therefore embarked on designing and testing generic platforms in which bHLH-PAS

transcription factors could be manipulated in an inducible, reversible and tissue specific

fashion. Utilising optimised Tet-On variants with improved sensitivity to Dox and reduced

background we constructed lentiviral and recombination mediated cassette exchange

plasmids, which contained all the components necessary for Dox-inducible expression (see

paper IV and Figure 3.).

We demonstrated that lentiviral generation of cell lines affords homogeneous inducible and

reversible manipulation of gene expression with no detectable background. We also

generated Col1a1 targeted RMCE FLP-In embryonic stem cells and mice for efficient FLP

mediated targeting of FLP-INDUCER Dox regulated constructs into the Col1a1 locus. We

demonstrate tight, inducible and reversible expression from mES-FLP-INDUCER cells able to

knockdown gene expression or ectopically express cDNA similar to endogenous levels.

Modular design of FLP-INDUCER plasmids allow for promoter exchange to enable tissue

specific promoters to be used to gain tissue/cell specific expression. We also show that LoxP-

STOP-LoxP mES FLP-INDUCER cells may, in principle, be used to limit inducible expression to

certain tissues or cells (Figure 3.).

While there are other established (i.e. Cre/loxP) and emerging (CRISPR/TALEN) techniques to

allow temporal and/or spatial gene ablation, some of which can employ inducible knockout

strategies such as tamoxifen inducible nuclear translocation of oestrogen receptor (ER)-CRE

recombinase  to  remove  flox’d  genes. These approaches irreversibly knockout the gene and

require  time  consuming  generation  of  flox’d  genes  and  crossing  into  CRE  driver  lines.  

42

Furthermore, in many disease models it may be advantageous to reversibly modulate gene

expression to monitor the progression/reversion of disease. We note that the FLP-INDUCER

system may be well suited to these specific disease model applications while also being

compatible with established CRE-mediated tissue specific and/or temporal control of gene

expression.

Alternative RMCE targeting loci have also been successfully used in the past for transgenic

mouse experiments including inducible Tet-On expression systems (Tchorz, Suply et al. 2012;

Haenebalcke, Goossens et al. 2013). Several of these systems would be compatible with the

FLP-INDUCER system with minimal modification, presenting an opportunity to generate multi

allelic inducible mouse models (Tasic, Hippenmeyer et al. 2011; Tchorz, Suply et al. 2012).

While the Rosa26 locus has been used extensively for generating site-specific transgenesis,

expression from this locus appears to result in significant mosaicsim and lower inducible

transgene induction as compared to the Col1a1 locus (Haenebalcke, Goossens et al. 2013). In

addition to the Col1a1 and rosa26 sites of integration the Hipp11 and hprt loci have

successfully been used to drive ubiquitous expression in mouse models and may represent

useful sites for RMCE mediated FLP-INDUCER insertion (Bronson, Plaehn et al. 1996; Tasic,

Hippenmeyer et al. 2011).

43

Figure 3. Lentiviral and Collagen 1a1 Recombination Mediated Cassette Exchange (RMCE)

systems for inducible and reversible knockdown or overexpression. A schematic showing the

single construct modular design of lentiviral and FLP-INDUCER vector systems used to

44

generate stable cell lines or site-specific integration into germline competent mouse

embryonic stem cells.

In an attempt to more accurately predict high potency miR30 shRNAs able to knockdown gene

expression using the FLP-INDUCER system, we also designed and tested an shRNA prediction

tool. This tool incorporated experimental data assessing the potency of ~60,000 shRNAs from

high throughput experiments as well as mRNA secondary structure to score shRNA sequences

based on their putative targeting efficiency. We show by cloning and testing 10 shRNA targets

against ARNT or ARNT2 using the mES-FLP-INDUCER system that the prediction tool can

effectively identify high potency shRNA sequences.

We are now well situated to rapidly generate inducible and reversible knockdown mouse

models for studying the function of the bHLH-PAS transcription factors. As discussed above

this will be invaluable to studying the more complex adult roles for the bHLH-PAS transcription

factors in normal physiology and disease.

45

References

Aitola, M. H. and M. T. Pelto-Huikko (2003). "Expression of Arnt and Arnt2 mRNA in developing murine tissues." The journal of histochemistry and cytochemistry : official journal of the Histochemistry Society 51(1): 41-54.

An, J. J., K. Gharami, et al. (2008). "Distinct role of long 3' UTR BDNF mRNA in spine morphology and synaptic plasticity in hippocampal neurons." Cell 134(1): 175-187.

Antonsson, C., M. L. Whitelaw, et al. (1995). "Distinct roles of the molecular chaperone hsp90 in modulating dioxin receptor function via the basic helix-loop-helix and PAS domains." Molecular and cellular biology 15(2): 756-765.

Atchley, W. R. and W. M. Fitch (1997). "A natural classification of the basic helix-loop-helix class of transcription factors." Proceedings of the National Academy of Sciences of the United States of America 94(10): 5172-5176.

Baxendale, S., C. J. Holdsworth, et al. (2012). "Identification of compounds with anti-convulsant properties in a zebrafish model of epileptic seizures." Disease models & mechanisms 5(6): 773-784.

Baxevanis, A. D. and C. R. Vinson (1993). "Interactions of coiled coils in transcription factors: where is the specificity?" Current opinion in genetics & development 3(2): 278-285.

Bersten, D. C., A. E. Sullivan, et al. (2013). "bHLH-PAS proteins in cancer." Nature reviews. Cancer 13(12): 827-841.

Bezprozvanny, I. and M. P. Mattson (2008). "Neuronal calcium mishandling and the pathogenesis of Alzheimer's disease." Trends in neurosciences 31(9): 454-463.

Blackwell, T. K., L. Kretzner, et al. (1990). "Sequence-specific DNA binding by the c-Myc protein." Science 250(4984): 1149-1151.

Blackwell, T. K. and H. Weintraub (1990). "Differences and similarities in DNA-binding preferences of MyoD and E2A protein complexes revealed by binding site selection." Science 250(4984): 1104-1110.

Bloodgood, B. L., N. Sharma, et al. (2013). "The activity-dependent transcription factor NPAS4 regulates domain-specific inhibition." Nature 503(7474): 121-125.

Bonnefond, A., A. Raimondo, et al. (2013). "Loss-of-function mutations in SIM1 contribute to obesity and Prader-Willi-like features." The Journal of clinical investigation 123(7): 3037-3041.

Brunskill, E. W., D. P. Witte, et al. (1999). "Characterization of npas3, a novel basic helix-loop-helix PAS gene expressed in the developing mouse nervous system." Mechanisms of development 88(2): 237-241.

Card, P. B., P. J. Erbel, et al. (2005). "Structural basis of ARNT PAS-B dimerization: use of a common beta-sheet interface for hetero- and homodimerization." Journal of molecular biology 353(3): 664-677.

Chan, W. K., G. Yao, et al. (1999). "Cross-talk between the aryl hydrocarbon receptor and hypoxia inducible factor signaling pathways. Demonstration of competition and compensation." The Journal of biological chemistry 274(17): 12115-12123.

Chapman-Smith, A., J. K. Lutwyche, et al. (2004). "Contribution of the Per/Arnt/Sim (PAS) domains to DNA binding by the basic helix-loop-helix PAS transcriptional regulators." The Journal of biological chemistry 279(7): 5353-5362.

46

Chapman-Smith, A. and M. L. Whitelaw (2006). "Novel DNA binding by a basic helix-loop-helix protein. The role of the dioxin receptor PAS domain." The Journal of biological chemistry 281(18): 12535-12545.

Chen, W. G., Q. Chang, et al. (2003). "Derepression of BDNF transcription involves calcium-dependent phosphorylation of MeCP2." Science 302(5646): 885-889.

Coba, M. P., L. M. Valor, et al. (2008). "Kinase networks integrate profiles of N-methyl-D-aspartate receptor-mediated gene expression in hippocampus." The Journal of biological chemistry 283(49): 34101-34107.

Coumailleau, P., L. Poellinger, et al. (1995). "Definition of a minimal domain of the dioxin receptor that is associated with Hsp90 and maintains wild type ligand binding affinity and specificity." The Journal of biological chemistry 270(42): 25291-25300.

Coutellier, L., S. Beraki, et al. (2012). "Npas4: a neuronal transcription factor with a key role in social and cognitive functions relevant to developmental disorders." PloS one 7(9): e46604.

Crews, S. T. (1998). "Control of cell lineage-specific development and transcription by bHLH-PAS proteins." Genes & development 12(5): 607-620.

Crews, S. T. and C. M. Fan (1999). "Remembrance of things PAS: regulation of development by bHLH-PAS proteins." Current opinion in genetics & development 9(5): 580-587.

Crews, S. T., J. B. Thomas, et al. (1988). "The Drosophila single-minded gene encodes a nuclear protein with sequence similarity to the per gene product." Cell 52(1): 143-151.

Dang, E. V., J. Barbi, et al. (2011). "Control of T(H)17/T(reg) balance by hypoxia-inducible factor 1." Cell 146(5): 772-784.

Davis, R. L. and H. Weintraub (1992). "Acquisition of myogenic specificity by replacement of three amino acid residues from MyoD into E12." Science 256(5059): 1027-1030.

Dean, C., H. Liu, et al. (2012). "Distinct subsets of Syt-IV/BDNF vesicles are sorted to axons versus dendrites and recruited to synapses by activity." The Journal of neuroscience : the official journal of the Society for Neuroscience 32(16): 5398-5413.

Dioum, E. M., J. Rutter, et al. (2002). "NPAS2: a gas-responsive transcription factor." Science 298(5602): 2385-2387.

Ebert, D. H., H. W. Gabel, et al. (2013). "Activity-dependent phosphorylation of MeCP2 threonine 308 regulates interaction with NCoR." Nature 499(7458): 341-345.

Ebert, D. H. and M. E. Greenberg (2013). "Activity-dependent neuronal signalling and autism spectrum disorder." Nature 493(7432): 327-337.

Englund, C., P. Steneberg, et al. (2002). "Attractive and repulsive functions of Slit are mediated by different receptors in the Dr