The 3D Genome Shapes Up For Pluripotency

The  promise of stem cell  therapies has been profoundly advanced by the seminal  finding that differentiated cells can  be re- programmed to pluripotent stem cells by overexpression of a cocktail of transcrip- tion  factors  (Takahashi  and   Yamanaka,

2006). However this process is very ineffi- cient,  and  understanding  the barriers to reprogramming has  become an  intense area  of study. The 3D organization of the genome is correlated with transcriptional control; for example, chromatin   loops  bring  enhancers  into  physical proximity with  their   gene   targets  (Kagey   et   al., 2010)  and   coregulated genes   occupy shared   nuclear  foci,   enriched  in   key transcription factors (Schoenfelder et al., 2010) specifically  in the cell types where the   genes are   expressed.  Four  recent studies   have   analyzed   the   chromatin interactions with key pluripotency  genes in  mouse  and   human   pluripotent  and differentiated cells, and they have uncov- ered  networks of interactions that are specific to  embryonic stem cells  (ESCs) or induced pluripotent stem cells (iPSCs). One    of   the    recent    studies   showed that   binding    of   overexpressed  OCT4 and  NANOG at  gene loci  was  identical  between  human   iPSCs   and   unreprog- rammed cells from the same experiment, but  that enhancer-promoter loops  within OCT4,   SOX2,   and   NANOG  loci   were specific to iPSCs, concomitant with tran- scription   from   these   endogenous   loci (Zhang  et  al.,  2013).  Looking  more sys- tematically, two  groups  assessed chro-  matin    interactions   genome-wide    with the  pluripotency genes  Nanog   or  Oct4 (Apostolou et al., 2013; Wei et al., 2013), and    another   simultaneously screened the    chromatin   conformations    around several key  loci  (Phillips-Cremins  et  al., 2013)  in  mouse ESCs   and   their  differ- entiated  counterparts.   Again,  a  set   of enhancer-promoter   loops    and   longer-  range  (over several megabases or across different chromosomes) gene coassocia- tions  were  unique  to  pluripotent  cells  or intermediates being successfully reprog- rammed to iPSCs,  despite equal  binding profiles  of the  overexpressed  transcrip- tion factors.

Because gene-centric chromatin inter- actions frequently  correlate with expres- sion of a given gene (Kagey et al., 2010; Schoenfelder et al., 2010), an open  ques- tion is whether such  genomic topologies are a functional  cause or a mere byprod- uct  of transcription.  Time course studies of the gain or loss of pluripotency-linked chromatin interactions  during      iPSC generation  (Wei   et   al.,   2013)  or   ESC differentiation   (Apostolou  et   al.,   2013) found  detectable changes in  the  chro- matin    interactions   days    before    tran- scriptional changes  and   differentiation phenotypes   were   observed.  Although these   data    suggest   a    causal   link between   chromosome contact  estab- lishment    and    gene   activation,   earlier experiments    showed     that     artificially induced enhancer-promoter looping  was able   to  stimulate weak  transcription of the  beta-globin  gene in  erythroid   cells lacking    a    hematopoietic   transcription factor (Deng et al., 2012), indicating that a permissive genome topology is required but not sufficient to induce transcriptional activation. Formation of the transcription- ally   permissive   chromatin   interactions with  pluripotency  gene  loci  appears to distinguish   the   small   number  of  cells able to be  reprogrammed to iPSCs from the nonprogrammable cells (Zhang et al., 2013), suggesting that  this may form the ‘‘epigenetic barrier’’ to pluripotency. Simi- larly, activation of beta-globin by a heter- ologous enhancer inserted on a different chromosome  is   restricted  to   a   small number of ‘‘jackpot’’ cells in the popula- tion (Noordermeer et al., 2011). Although these  data  were   generated  using   two artificial  systems, they  suggest a  model where    chance   chromatin    interactions allow  transient  or   inefficient   transcrip- tional  activation by  bringing  together a hub  of regulatory factors (Figure 1). Pro- gressive stabilization of these interactions and/or additional factors then commit  the gene  to  efficient  transcription, perhaps creating a sufficiently permissive environ- ment to allow transcription to occur  in the absence of the initiating chromatin inter- actions  (Wei  et  al.,  2013).  The  detection of pluripotency-gene-linked    chromatin interactions  specific to  partially  reprog- rammed  intermediate iPSCs   (Apostolou et al., 2013) is consistent with the idea of a   progressive  search   for   a   functional genome  configuration.  It  will be   inter- esting  to see  if such  a model  can explain gene expression control  in normal  differ- entiation  and how it would differ for genes subject  to   more   acute   transcriptional changes, such   as   the  immediate  early genes upon  mitogenic stimulation.

Previous   research   has    implicated several  protein   factors   to   be   respon- sible  for  chromatin interactions.Protein- protein  interactions among  transcription factors  bound  to   promoters  and   en- hancers    stimulate   chromatin   looping (Deng et al., 2012) and longer-range coas- sociations (Schoenfelder et al., 2010), and more  general, non-cell-type-specific fac- tors have also been implicated in genome folding. These include  the insulator-bind- ing  protein  CTCF  (Splinter et  al.,  2006), the cohesin complex that mediates sister chromatid cohesion (Kagey et al., 2010), and  the  transcriptional  coactivator com- plex, Mediator (Kagey et al., 2010). Again, demonstrating   a    causal    relationship between  binding  of  these proteins and formation   of  chromatin  loops has  been difficult (Deng et al., 2012).   Collectively,   the   four recent  studies   shed   further light on  the  interplay between transcription factors and  the ‘‘architectural’’      proteins      in establishing functional   chro- matin    interactions.    Phillips- Cremins  et  al.  (2013),  in  the most    detailed  analysis  of genome folding around several pluripotency gene  loci,  reveal that chromatin loops  are prev- alently formed  around binding sites   of the  architectural  pro-either    be   involved   in   loops at   cell  states  that   have   not yet  been  investigated, or they may function   in             different mechanisms to  maintain  plu- ripotency.      Finally,                       these studies highlighted  gene loop formation    for    transcriptional activation,    but    programmed gene   loops may be equally important in order  to mask dif- ferentiation-specific   genes   in ESCs or during reprogramming to iPSCs, an issue  that remains open for future studies. teins    cohesin,   CTCF,   and/or Mediator,    with    significantly less    contribution from the binding   sites   of  the  pluripo- tency                  transcription        factors Oct4, Sox2, and Nanog. Inter- estingly,   by   comparing  ESC and  neural  precursor interac- tion            profiles,     the     authors were   able   to  distinguish two

Figure 1.  Model for the  Role of Chromatin Interactions in Mediating

Pluripotency

Overexpressed Yamanaka transcription factors (blue ovals) bind to motifs at the  promoters and  enhancers  of pluripotency genes (purple  rectangles) in differentiated tissues (left panel),  but very few are reprogrammed to iPSCs.  Chance chromatin interactions between enhancers and  promoters, and be- tween pluripotency genes,  seem  to  be  required  for  reprogramming (right panel), and are reinforced by further recruitment of the ‘‘architectural proteins’’ Mediator   (yellow squares) and  cohesin  rings  (orange circles),  providing  a permissive  environment  for  transcriptional  activation   of  the  pluripotency genes.

ACKNOWLEDGMENTS

We    thank   Cyril   Sarrauste   for   the artwork.  We apologize to colleagues whose work we could  not cite due  to space constraints.