-OH No! hiPSCs Misplace Their 5hmCs

The  ability to  reprogram cell fate  by  the overexpression  of  a   handful   of  genes has   opened  the  door   to  limitless  pos- sibilities   for  modeling   development  or diseases in  vitro  and   eventual  patient- specific  clinical  applications. Yamanaka and  colleagues first described this para- digm by overexpressing four transcription factors in fibroblasts, leading  to their con- version  to a pluripotent state. Since  then, pluripotent  cells   derived    by   induction (induced pluripotent stem cells, or iPSCS) have  been extensively compared to their embryo-derived counterparts (embryonic stem cells,  or ESCs). Many studies have identified  epigenetic and   transcriptional differences between these types of plurip- otent  stem cells (Chin et al., 2009;  Lister et  al., 2011;  Ohi et  al., 2011;  Ruiz et  al.,

2012),  while others have  suggested  that these differences  are  no  more  dramatic than    those found when comparing different   ESC   lines   (or  iPSC   lines)  to each other  (Bock et al., 2011). Ultimately, high-quality   human    iPSCs    and    ESCs could   be functionally  equivalent, while potentially    still   distinguishable   at   the molecular level depending on the resolu- tion  of  the  analysis  and  the  number of lines analyzed.

In the present study,  Wang et al. (2013) exploit  high-resolution  analyses of  DNA methylation     to    confirm    and     extend
previous  findings that hiPSCs are distin- guished  from  hESCs by  altered  methyl- ation  status of  subtelomeric DNA. Previ- ously,  three  groups provided compelling evidence  that   hiPSCs   appear  to   have aberrant  patterns  of   5-methyl-Cytosine (5mC)   modification  within subtelomeric DNA regions  (Lister et al., 2011; Ohi et al., 2011;  Ruiz et  al.,  2012).  This  patterning was   thought   to   result   from   inefficient erasing  and/or rewriting of the methylome during    reprogramming, reflecting an epigenetic memory  of the  somatic  state from which  they  were derived. However, since  these  previous studies relied  upon traditional  bisulfite  sequencing,  they did not   distinguish    5mC   from  5-hydroxy- methyl-Cytosine (5hmC), a mark that  has  recently  been shown  to  be  a  signature aspect of the methyl-DNA repertoire and important for gene regulation.  Wang et al. first found that TET1, a key enzyme  that converts   5mC   to   5hmC,    is   strongly induced during reprogramming, and  that blocking   its   expression abrogated  the reprogramming process.   This  suggests that  the  selective  conversion of 5mC  to 5hmC  is important for acquisition  of the pluripotent    state  (Wang  et   al.,   2013). Similar  results  were  obtained previously for Tet1 and Tet2 in murine reprogramming (Koh et al., 2011). Interestingly,  knocking down   TET1  in  human   pluripotent  cells had  little effect  on  the  pluripotent  state,  suggesting  that  the   key  role  for  TET1 occurs  during   the   initial  conversion  of 5hmC   during  reprogramming  and  does not involve the maintenance of this mark (Wang et al., 2013).

Extensive  profiling  of  5hmC   in  fibro- blasts, hiPSCs, and hESCs demonstrated that, out of 372,423 regions that were en- riched   for  5hmC  marks   throughout the genome, just 113 (0.03%) could be classi- fied as  differentially  hydroxyl-methylated regions (DhMRs) when  drawn  from com- parisons between hiPSCs  and  hESCs.  The  finding  that  such   a  tiny  fraction  of the genome appeared to be differentially methylated  is  consistent   with  previous reports  and   is  further  evidence  of  the remarkably    faithful    process   that     is induced upon  introduction of the  Yama- naka  factors. On the other hand, 93%  of these DhMRs  were  hypohydroxymethy- lated,    which   would indicate that the reprogramming process appears to spe- cifically fail to  convert  5mC  to  5hmC  in certain   portions  of the  genome, leaving these  regions in a  state more  typical  of the somatic cells from which  they came. Furthermore, the majority of these DhMRs tended to  be  localized   to  subtelomeric regions of the genome.

These findings are interesting in light of previous studies showing similar patterns of 5mC in subtelomeric regions of hiPSCs (Lister  et  al.,  2011;   Ruiz  et  al.,  2012). Remarkably, a list of nine genes observed to   be   differently   expressed  in  hiPSCs  versus   hESCs    by   Ruiz   et   al.   (2012) highly  overlaps  with   the   list  of   sum- marized  hypomethylation  hotspots pre- sented in Wang et al. The fact that multiple labs   using   distinct   cell  lines   came  to similar conclusions is strong   evidence that    these    subtelomeric   regions   are indeed  hotspots of  reprogramming and warrant further consideration as possible proxies  for defining  the  quality  of PSCs at  the  molecular  level. This metric  could prove  to  be  useful   because no  assay currently exists to quantitatively  assess the   quality  of  human  pluripotent stem cells.  Regardless, the  methylation status of these hotspots across human  pluripo- tent   stem  cell  lines   is  clear   evidence that, at a minimum, hiPSCs exhibit signif- icantly   more   epigenetic  variation   than existing hESCs.

The  data in Wang  et  al. might  appear to  be  confounding to  previous studies that    did   not   find   consistent    DhMRs between     hESCs     and     hiPSCs.    The simplest explanation is that  the  molecu- lar  differences between  these types of pluripotent stem  cells  are   quite   subtle (just  0.03%  of  5hmC-enriched  regions in Wang et  al., 2013).  Therefore,  studies in  which   many   lines   have  been  com- pared at  low  resolution  (with  Reduced Resolution   Bisulphite    Sequencing    or DNA Methylation  Arrays) did  not  identify consistent    differences   (Bock    et    al., 2011;   Nazor  et  al.,  2012),  while  those that  used high  resolution  (single  nucleo- tide)  analyses  on  fewer   lines  have   re- ported  differences  (Lister  et  al.,  2011; Ohi et al., 2011;  Ruiz et al., 2012; Wang et    al.,   2013).   Regardless,    the    more compelling issue   is  why  these subtelo- meric   domains   are   apparently  difficult to appropriately methylate during reprog- ramming.  Furthermore,   can  one   take advantage of  this  observation  to  learn something  about the  basic mechanisms of  how  the  Yamanaka factors  drive  this transformation?   Numerous    epigenetic barriers to reprogramming have  recently been  identified,   partially  explaining  the very   low   inefficiency   of   the   process (Watanabe et al., 2013). Wang et al. sug- gest that  activity  of TET1  represents yet another barrier  to proper reprogramming to the pluripotent state.

The replication and organization of telo- meres  presents  a   serious   engineering problem for cells.  The  ends of chromo- somes require their own unique  machin- ery  to  preserve the  length  and  integrity of telomeres, while isolating  the subtelo- mere  domain   from  such  machinations. Among   the   panoply   of   reorganization events that must occur during reprogram- ming  is the  reestablishment  of  telomere length   by   telomerase.   Another   recent paper demonstrated that TRF1, a compo- nent  of the  shelterin complex that  main- tains  telomere integrity,  was required for reprogramming  (Schneider et  al.,  2013). Together with those of Wang et al., these findings  suggest that  reorganizing  telo- meres during  reprogramming  is not  just a  matter  of  restoring  their  length,  but also requires the  activity  of the  shelterin complex  and   epigenetic  remodeling  of the subtelomeric domain. This latter reor- ganization appears to fall somewhat short  in  hiPSCs,  affecting   the   expression  of several genes in these regions (TCERG1l, TMEM132D, etc.) (Chin et al., 2009; Ruiz et al., 2012; Wang et al., 2013). The key unresolved  issue   is  whether the  small degree of  hypomethylation  observed in hiPSCs    has    a   functional    significance or  is  inconsequential. Regardless,  one should   take  into  account the  increased degree  of  epigenetic   variability  across hiPSC  lines  when modeling disease  or development in vitro.