The CpG-sites of the CBX3 ubiquitous chromatin opening element are critical structural determinants for the anti-silencing function.

1 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 www.nature.com/scientificreports The CpG-sites of the CBX3 ubiquitous chromatin opening element are critical structural determinants for the anti-silencing function Jessica Kunkiel1,2, Natascha Gödecke3, Mania Ackermann2,4, Dirk Hoffmann2, Axel Schambach2,5, Nico Lachmann2,4, Dagmar Wirth2,3 & Thomas Moritz1,2 Suppression of therapeutic transgene expression from retroviral gene therapy vectors by epigenetic defence mechanisms represents a problem that is particularly encountered in pluripotent stem cells (PSCs) and their differentiated progeny. Transgene expression in these cells, however, can be stabilised by CpG-rich ubiquitous chromatin opening elements (UCOEs). In this context we recently demonstrated profound anti-silencing properties for the small (679 bp) CBX3-UCO element and we now confirmed this observation in the context of the defined murine chromosomal loci ROSA26 and TIGRE. Moreover, since the structural basis for the anti-silencing activity of UCOEs has remained poorly defined, we interrogated various CBX3 subfragments in the context of lentiviral vectors and murine PSCs. We demonstrated marked though distinct anti-silencing activity in the pluripotent state and during PSCdifferentiation for several of the CBX3 subfragments. This activity was significantly correlated with CpG content as well as endogenous transcriptional activity. Interestingly, also a scrambled CBX3 version with preserved CpG-sites retained the anti-silencing activity despite the lack of endogenous promoter activity. Our data therefore highlight the importance of CpG-sites and transcriptional activity for UCOE functionality and suggest contributions from different mechanisms to the overall anti-silencing function of the CBX3 element. Although the design of lentiviral vectors has improved substantially over the last decade, even modern SIN-lentiviral vectors remain subject to positional effects that can lead to silenced or variegated transgene expression1, 2. In particular pluripotent stem cells are problematic in this respect, as these cells harbour strong intrinsic defence mechanisms against foreign and especially viral DNA3–7. Moreover, the differentiation of pluripotent stem cells towards more mature cell types involves genome wide epigenetic remodelling with a potentially detrimental impact on transgene expression. To evade this transcriptional repression so called ubiquitous chromatin opening elements (UCOEs) have been applied8. In previous studies, we and others have demonstrated the ability of UCOEs to establish stable expression of lentiviral transgenes driven by both viral and physiological promoters in murine and human pluripotent as well as multipotent target cells and thereof differentiated progeny9–13, thus demonstrating broad applicability of UCOEs. The most commonly used UCOE is derived from the human HNRPA2B1-CBX3 housekeeping gene locus. This element contains two divergently transcribed promoters, the HNRPA2B1 promoter and the CBX3 promoter which are both spanned by a large stretch of non-methylated CpG-islands (CGI)11, 13. A central 1.5 kb sequence 1Reprogramming and Gene Therapy Group, REBIRTH Cluster of Excellence, Hannover Medical School, 30625, Hannover, Germany. 2Institute of Experimental Hematology, Hannover Medical School, 30625, Hannover, Germany. 3Model Systems for Infection and Immunity Group, Helmholtz Centre for Infection Research, 38124, Braunschweig, Germany. 4Translational Hematology of Congenital Diseases Junior-Group, REBIRTH Cluster of Excellence, Hannover Medical School, 30625, Hannover, Germany. 5Division of Hematology/Oncology, Boston Children’s Hospital, Boston, MA, USA. Correspondence and requests for materials should be addressed to T.M. (email: [email protected]) Received: 10 February 2017 Accepted: 10 May 2017 Published: xx xx xxxx OPEN www.nature.com/scientificreports/ 2 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 containing the anti-silencing activity (A2UCOE) has been identified, which is associated with a permissive chromatin structure marked by hyperacetylation of histones H3 and H4, trimethylation of lysine 4 on histone H3 (H3K4me3), and the lack of H114. We have recently shown that a much smaller sequence comprising only the 0.7 kb CBX3 moiety of the A2UCOE (referred to as CBX3) can effectively protect SIN-lentiviral transgene expression in murine and human pluripotent cells as well as their hematopoietic derivatives15. The CBX3 prevented CpG methylation of a spleen focus forming virus (SFFV) promoter and reduced the presence of repressive chromatin marks (e.g. H3K9me3 and H3K27me3) while introducing active marks such as H3K4me3 and PhosPol2. While mechanistically the anti-silencing function of the UCOEs has been attributed to the presence of non-methylated CGIs, a feature shared by all known UCOEs, up to now the anti-silencing properties of the UCOE have not been linked to specific structural features. To potentially identify such features within the CBX3 structure we have generated a number of CBX3-derived subfragments and analysed the functionality of these subfragments in the context of lentiviral vectors and a broad range of integration sites within murine ESCs and iPSCs as well as their differentiated progeny. In addition, we have investigated the relevance of the CpG-islands for the anti-silencing properties by employing two synthetic scrambled-CBX3 versions with either maintained or destroyed CpG-sites. Moreover, to explore the relevance of transcriptional activity for the anti-silencing function we examined the promoter activity of our CBX3 subfragments. So far all investigations assessing UCOE function have been performed in bulk cell populations with varying integration sites while a direct comparison of expression cassettes with and without UCOE in a defined integration site and chromosomal architecture has not yet been performed. Accordingly, it is not known if UCOEs can protect from silencing irrespective of the nature of the chromosomal integration site. Therefore, we now have examined the anti-silencing ability of the CBX3 at two defined chromosomal loci, the ROSA2616 and the TIGRE locus T117, in single-cell derived mESC clones utilising recombinase mediated cassette exchange (RMCE). Results CBX3 subfragments and their activity in pluripotent stem cells. To identify the structural attributes responsible for the anti-silencing function of the CBX3, we generated a total of six partially overlapping subfragments of the CBX3 element (Table1, Fig.1A). To generate cell populations representing a broad range of different integration sites, we employed lentiviral transduction. To this end, the subfragments were introduced into a third-generation self-inactivating (SIN)-lentiviral vector upstream of the viral SFFV promoter driving the expression of the eGFP reporter. Vectors employing the full length CBX3 (C-SG) and the 1.5 kb A2UCOE (A2SG) served as positive controls, whereas the SFFV promoter alone (SG) was used as a negative control (Fig.1B). A murine ESC as well as an induced pluripotent stem cell (miPSC) line were transduced with an MOI of 10 and 30, respectively. By this procedure, similar vector copy numbers (VCN) of approximately 2–3 for mESCs and 1–2 for miPSCs were achieved for all vectors (Supplementary TableS1). Expression of eGFP in pluripotent cells was monitored over a period of 31 days, i.e. 9 passages. Transduction efficiencies were assessed by flow cytometry on day 3 (Supplementary Fig.S1A) and served as a reference point to determine the relative number of eGFP-expressing cells over time. For all vectors and in both the mESC as well as the miPSC line, loss of transgene expression was primarily observed during the first 7 days (passage 2). This drop in transgene expression was more prominent for SGthan C-SGor A2-SG-transduced control cells, indicating the transgene stabilizing function of the UCOEs. Cells transduced with the vectors C(1-508)-SG, C(85-508)-SG, C(170-508)-SG, C(340-508)-SG, or C(503-679)-SG maintained transgene expression at levels comparable to the positive controls ranging from 71% to 80% in mESCs and 61% to 68% in miPSCs (Fig.2A,B). Only the C(1-339) fragment failed to display anti-silencing properties. Following this drop, the transgene expression stabilized and stable eGFP-expression was observed for at least 9 passages i.e. until day 31 (Figs2C,D, S1B,C). Taken together, these results indicate particularly the CpG-rich central and 3′ region of the CBX3 region to be involved in UCOE-mediated transgene stabilization in pluripotent cells. Biologic activity of the various fragments was also reflected by the median fluorescence intensity (MFI) of eGFP expression in transduced cell populations (Table2), indicating the contribution of the CBX3-derived elements to the overall promoter activity. While the lowest MFI was observed for SG-transduced cells, at least in mESCs, C(1-339)-SG transduced cells displayed similar expression intensities as C(1-508)-SG, C(85-508)-SG, and C(170-508)-SG transduced cells. Lower expression intensities were observed particularly for C(340-508)-SG transduced cells in both mESCs and miPSCs. UCO elements Length (bp) CpG-sites (#) G+C (%) CpG density (CpG/100 bp) A2UCOE 1550 142 64.0 9.2 CBX3 679 63 67.9 9.3 CBX3(1-339) 339 18 57.6 5.3 CBX3(1-508) 508 35 62.2 6.9 CBX3(85-508) 424 31 63.9 7.3 CBX3(170-508) 339 26 66.7 7.7 CBX3(340-508) 168 17 71.4 10.1 CBX3(503-679) 177 28 83.1 15.8 Table 1. Structural attributes of the A2UCOE, CBX3 and CBX3 subfragments. bp, base pairs; #, number of www.nature.com/scientificreports/ 3 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 Anti-silencing function of CBX3 subfragments in differentiated mESCs/iPSCs: correlation with CpG-sites. During differentiation stem cells undergo extensive epigenetic remodelling and euchromatic regions that served as favourable integration sites for lentiviral vectors may become heterochromatic, thereby posing a hindrance to lentiviral transgene expression1, 2. The in vitro differentiation of mESCs and miPSCs therefore represents a much more stringent system to analyse the potency of anti-silencing elements such as our CBX3 subfragments. To this end, the cells were subjected to undirected differentiation and the expression of eGFP in differentiated, SSEA-1 negative cells was assessed on day 8 (Fig.3A,B). Whereas SG-transduced control mESCs (7%) as well as miPSCs (4%) showed a nearly complete loss of transgene expression upon differentiation, incorporation of CBX3 or A2UCOE into the vector markedly increased the population of transgene positive cells (CBX3: 63% (mESC), 81% (miPSC); A2UCOE: 64% (mESC), 72% (miPSC)). Also, all constructs with CBX3 subfragments stabilized transgene expression to some extent in differentiated mESCand miPSC-derived cells when compared to SG controls, although these differences were significant only for CBX3(1-508) and CBX3(85-508). With eGFP-expressing cells ranging from 34% to 41%, however, neither of these subfragments showed full CBX3 activity. On the other hand, CBX3(1-339) and CBX3(340-508) clearly showed the lowest anti-silencing activity in both cell types. No major differences were observed between C(1-508)-SG, C(85-508)-SG, C(170-508)-SG and to some degree also C(503-679)-SG transduced cells. These data indicate that in addition to a central region common to the three subfragments with higher activity, also the 3′-end region of the CBX3 (bp 503 to bp 679) is involved in anti-silencing. Interestingly, these four CBX3 subfragments all contain a high number of CpG-sites and the amount of CpG-sites per fragment was found to positively correlate with its anti-silencing capacity (measured as the percentage of eGFP-positive cells following differentiation) based on data for mESCs (R2 = 0.6811, p < 0.0001) as well as miPSCs (R2 = 0.6248, p < 0.0001) (Fig.3C,D). While a similar correlation was observed for UCOE length (R2 = 0.6299, p < 0.0001 in mESCs; R2 = 0.5848, p < 0.0001 in miPSCs), no significant correlation was detected for CpG-site density (CpGs/100 bp) or GC-content (data not shown). Functionality of a scrambled CBX3-UCOE in pluripotent and differentiated mESCs/miPSCs. To further assess the relevance of the CpG-sites within the CBX3 element, two scrambled versions of the CBX3 were designed. The first element was generated by random shuffling of the DNA sequence while retaining the CpG-sites (roughly 20% of total sequence) at their original position without changing the overall nucleotide composition (CscrCpG). The second element was designed by randomly shuffling the complete DNA sequence thereby destroying the CpG-sites (Cscr; alignment of sequences given in Supplementary Fig.S2). Again, the elements were inserted into the SG lentiviral vector thus giving rise to the constructs CscrCpG-SG and Cscr-SG (Fig.4A). Figure 1. CBX3 subfragments and lentiviral vectors. (A) Schematic representation of the CBX3 element with its CpG-sites (black bars) and deleted splice sites (red bars) as well as the CBX3 subfragments obtained by PCR (CBX3(1-339) - CBX3(340-508)) and the CBX3(503-679) subfragment generated by restriction digestion with XhoI and ApaI. (B) Third-generation self-inactivating (SIN) lentiviral constructs expressing an eGFP cDNA from the spleen focus forming virus (SFFV) promoter in the absence or presence of either the 1.5 kb A2UCOE, the 679 bp CBX3 or any of the CBX3 subfragments. Abbreviations: ΔLTR: Long terminal repeat harboring the SIN mutation in the U3 region of the LTR; ψ: extended encapsidation signal; RRE: Rev-response element; cPPT: central polypurine tract; GFP: (enhanced) green fluorescent protein; wPRE: woodchuck hepatitis virus posttranscriptional element. www.nature.com/scientificreports/ 4 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 Again, mESCs and miPSCs were transduced while aiming for comparable transduction efficiencies and VCNs (Supplementary TableS3). The eGFP expression in pluripotent cells was monitored by flow cytometry for a period of 24 days (Fig.4B,C). In agreement with our previous data, loss of eGFP expression was primarily observed up to day 7 (passage 2) with relatively constant numbers of eGFP-positive cells thereafter. After 24 Figure 2. Activity of CBX3 subfragments in pluripotent stem cells. Murine ESCs and iPSCs were transduced with lentiviral vectors with or without the A2UCOE, the CBX3 or the CBX3 subfragments. The eGFP expression was monitored for 31 days and the relative expression (d3 = 100%) is shown for day 7 in (A) mESCs and (B) miPSCs. The time-course of relative eGFP expression for 31 days is depicted for (C) mESCs and (D) miPSCs. Activity of the subfragments was compared to the SG negative control. Data represent independent experiments n = 3; mean ± SD; *p < 0.05, **p < 0.01, ****p < 0.0001 as determined by one-way ANOVA. Median fluorescence intensity (MFI) Vector mESC miPSC d3 d31 d3 d31 SG 54 ± 6 28 ± 4 26 ± 9 33 ± 1 A2-SG 201 ± 51*** 142 ± 21**** - - C-SG 223 ± 25**** 121 ± 17**** 122 ± 61** 106 ± 38* C(1-339)-SG 119 ± 27 104 ± 22*** 37 ± 17 57 ± 6 C(1-508)-SG 148 ± 10*128 ± 13**** 75 ± 22 89 ± 12 C(85-508)-SG 142 ± 37*135 ± 23**** 89 ± 24 97 ± 39* C(170-508)-SG 125 ± 37 113 ± 13*** 75 ± 14 97 ± 16* C(340-508)-SG 53 ± 17 43 ± 11 48 ± 11 48 ± 10 C(503-679)-SG 87 ± 4 56 ± 7 67 ± 6 81 ± 22 Table 2. Expression intensity of CBX3 subfragments in pluripotent stem cells. eGFP fluorescence mediated by the UCO elements was compared to the SG negative control. Data represent independent experiments n = 3; mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by one-way ANOVA. www.nature.com/scientificreports/ 5 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 days, only the CscrCpG-SG transduced mESCs and miPSCs displayed significantly more eGFP-positive cells (72% and 44%) than SG-transduced negative controls (43% and 26%). In contrast, no significant differences to the negative control were identified in both Cscr-SG transduced mESCs (51%) and miPSCs (29%) after 24 days (Supplementary Fig.S3A,B). Of note, only C-SG transduced cells demonstrated an increase in MFI above SG levels (Supplementary TableS3), whereas neither the incorporation of CscrCpG nor Cscr could achieve this. To assess the anti-silencing activity of the scrambled CBX3 elements during differentiation, we determined the relative eGFP expression in SSEA-1 negative cells following eight days of undirected differentiation using day 0 as reference (Fig.4D–F). Confirming our previous data, SG transduced mESCs and miPSCs lost most of their expression upon differentiation (down to 8% and 3%, respectively), whereas C-SG transduced mESCs and miPSCs demonstrated significantly higher levels of eGFP expressing cells (48% and 68%, respectively). Interestingly, also CscrCpG-SG cells showed significantly elevated levels of transgene expression upon differentiation of both mESCs (33%) and miPSCs (47%). Cscr-SG transduced cells, on the other hand, failed to show significant anti-silencing activity in this assay with transgene expression in mESCor miPSC-derived cells of only 17% or 4%, respectively. These results further highlighted the importance of the CpG-sites for the anti-silencing function of the CBX3. Promoter activity of CBX3-UCOE fragments in pluripotent and differentiated mESCs/miPSCs. The data given above on MFI for eGFP expression suggest that both the A2UCOE and the CBX3 still harbour intrinsic transcriptional activity. The dual promoter activity of the A2UCOE, however, seems to be dispensable for its anti-silencing activity, since the CBX3 with its single promoter is still highly efficient in this respect. On the other hand, another 0.7 kb UCOE derived from the first intron of CBX3 but lacking intrinsic transcriptional activity was shown to mediate only partial protection from silencing of the SFFV promoter in vitro18. Taken together with the observation that the A2UCOE showed an orientation-dependent anti-silencing Figure 3. Activity of the CBX3 subfragments in differentiated cells. The transduced cells were subjected to an undirected EB-based differentiation and the eGFP expression was determined on day 0 as well as in SSEA-1 negative cells on day 8. The relative expression of eGFP was determined (d0 = 100%) for (A) mESCs and (B) miPSCs. (C,D) The correlation of relative eGFP expression in differentiated cells and the number of CpG-sites contained within each element is shown for (C) mESCs and (D) miPSCs (p < 0.0001). Activity of the subfragments was compared to the SG negative control. Data represent independent experiments n = 3; mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by one-way ANOVA. www.nature.com/scientificreports/ 6 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 effect with improved function only when the CBX3 promoter was facing the juxtaposed heterologous promoter11, this poses the question whether intrinsic promoter activity in the direction of the transgene has a role in the anti-silencing ability of the UCOEs. Since the CBX3 is a CpG-island promoter and thus does not contain known defined motifs, such as a TATA box, that could be mutated to erase the promoter function, we investigated the promoter activity of the CBX3 subfragments instead and compared this to their anti-silencing ability. For these studies, the SFFV promoter was excised from the SG lentiviral vectors, giving rise to the vector constructs depicted in Fig.5A. Three days following the transduction of mESCs and miPSCs, the eGFP expression was assessed by flow cytometry (Fig.5B). Since for some constructs it was difficult to determine the titre of the lentiviral supernatants due to low or no eGFP expression, some variability of VCN could not be avoided (Supplementary TableS4). As expected, right after transduction (day 3) the eGFP control construct (G) showed only a minimal background MFI for mESCs as well as miPSCs. Minimal to no promoter activity was also observed for the subfragments C(1339), C(340-508) as well as the synthetic element CscrCpG. In contrast, C(1-508), C(85-508), C(170-508), and Figure 4. Anti-silencing activity of synthetic scrambled CBX3 elements. The DNA sequence of the CBX3 was shuffled randomly to create two synthetic elements that harbor the same nucleotide composition in a random order with one sequence maintaining the CpG-sites at their original position (CscrCpG) and the other completely destroying these sites (Cscr). (A) The elements were introduced into the SFFV.eGFP (SG) third generation SIN-lentiviral vector upstream of the promoter. (B,C) Following transduction, the eGFP expression was monitored by flow cytometry for 24 days in SSEA-1 positive (B) mESCs and (C) miPSCs. (D–F) The transduced cells were subjected to an EB-based undirected differentiation with transgene expression measured on day 0 and in SSEA-1 negative cells on day 8. (D) Representative flow cytometric analysis of mESCs and miPSCs prior to (d0) and after differentiation (d8). Relative eGFP expression was determined in differentiated (SSEA-1 negative) (E) mESCs and (F) miPSCs. Activity of the synthetic elements was compared to the SG negative control. Data represent independent experiments n = 3; mean ± SD; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 as determined by one-way ANOVA. www.nature.com/scientificreports/ 7 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 C(503-679) clearly demonstrated endogenous promoter activity, which for some constructs was similar to the activity of CBX3 that was used as the positive control. To facilitate the read-out, for further analysis the background MFI of G-transduced cells was deducted from all MFIs measured in the same experimental setting yielding MFIcorr. Moreover, to account for the differences in VCN, the MFIcorr per VCN (MFIcorr/VCN) was calculated and used for further analysis. When MFIcorr/VCN was monitored over a 28-day period, relatively stable eGFP expression levels were observed for the individual constructs (Supplementary Fig.S4A,B) and even on day 28 significant promoter activity was detected for the C(1-508) subfragment in both mESCs and miPSCs as well as for the C(85-508) in mESCs and the C(503-679) in miPSCs. Although not significant, C(170-508) also showed promoter activity in both mESCs and miPSCs (Fig.5C,D). Of note, despite its anti-silencing activity in the context of SG lentiviral vectors, the CscrCpG showed Figure 5. Promoter activity of CBX3 derived fragments. (A) Third generation SIN-lentiviral vectors harboring an eGFP cDNA preceded by CBX3-derived elements. (B) Representative flow cytometric analysis of eGFP expression in mESCs and miPSCs three days after transduction. (C-D) Median fluorescence intensity (MFI) of eGFP-expressing cells was measured by flow cytometry, corrected for background eGFP expression (MFIcorr; G = 0) and determined per vector copy number (VCN) in pluripotent (SSEA-1 positive) (C) mESCs and (D) miPSCs 28 days after transduction. Activity of the subfragments was compared to the eGFP negative control. Data represent independent experiments n = 3; mean ± SD; *p < 0.05, ***p < 0.001, ****p < 0.0001 as determined by one-way ANOVA. www.nature.com/scientificreports/ 8 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 almost no eGFP expression in these experiments, indicating that the mediated anti-silencing effect was independent of transcriptional activity. In contrast, the other fragments that previously demonstrated good anti-silencing ability in the context of SG-based lentiviral vectors, i.e. C(1-508), C(85-508), C(170-508), and C(503-679), exhibited notable promoter activity in mESCs as well as miPSCs. According to these data, for the CBX3 subfragments higher endogenous promoter activity tended to correlate positively with their ability to protect SG-based vectors from silencing in mESCs (p < 0.0001) or miPSCs (p = 0.0007) (Supplementary Fig.S4C,D). Little endogenous promoter activity was, however, observed for all constructs except for the complete CBX3 element following eight days of directed differentiation (Supplementary Fig.S4E,F). CBX3 function at defined chromosomal loci. Lentiviral transduction of stem and progenitor cells is accompanied with genome-wide integration of vectors. This was described to result in heterogeneous transgene expression in individual cell clones of cell populations9–12, 19. It was shown that transduction with UCOE-containing vectors resulted in significantly less heterogeneity than populations of cells with UCOE-free vectors9, 10, 15. However, it was not clear whether the UCOE-mediated protection from epigenetic modulations was exhibited equally at all integration sites or only as a consequence of integration into chromosomal sites with favourable epigenetic features. To address this notion, we used Flp recombinase mediated cassette exchange (RMCE) to investigate the impact of a UCOE at two defined chromosomal sites. In particular, we integrated a cassette comprising the silencing prone viral SFFV promoter driving eGFP with or without the CBX3 UCOE at two defined chromosomal loci in mouse embryonic stem cells (mESCs); the ROSA26 and the TIGRE locus (Fig.6A). The incorporation of the eGFP reporter allowed single cell analyses to visualise the heterogeneous effect of transgene silencing. In the ROSA26 locus of pluripotent mESCs the percentage of transgene expressing cells gradually decreased to 60-80% for all isogenic SFFV.eGFP (SG) as well as all CBX3.SFFV.eGFP (C-SG) clones (Supplementary Fig.S5A), whilst the expression of eGFP remained fairly stable over 6–10 consecutive passages at the TIGRE locus for both SG and C-SG clones (Supplementary Fig.S5B). The comparable loss of eGFP expression at ROSA26-targeted isogenic subclones indicates similar epigenetic silencing kinetics upon RMCE, which suggests the presence of a highly reproducible modification. To assess the CBX3 function during differentiation, the transgene expression was determined for both ROSA26 and TIGRE targeted cells (Fig.6B,C) on day 0 and day 8 of an undirected EB-based differentiation. Here, day 0 served as reference point to determine the relative number of eGFP-expressing cells after eight days of differentiation. The SG clones of ROSA26 targeted cells showed nearly a complete loss of transgene expression with only 1–5% eGFP-positive cells remaining. Incorporation of the CBX3 increased eGFP-positive cells to 7–15% (Fig.6B). At the TIGRE locus, undirected differentiation resulted in 31–42% eGFP-expressing cells in SG Figure 6. Anti-silencing function of the CBX3 at two defined chromosomal loci. (A) Murine ESCs were targeted at either the ROSA26 or the TIGRE chromosomal locus by using FLP-mediated cassette exchange to introduce either an SFFV.eGFP (SG) or CBX3.SFFV.eGFP (CSG) cassette. Subclones of (B) ROSA26- (SG: 3 clones; CSG: 4 clones) and (C) TIGRE- (SG: 3 clones; CSG: 3 clones) targeted cells were subjected to EBbased undirected differentiation and the relative percentage of eGFP-expressing cells was determined by flow cytometry on day 8 in SSEA-1 negative cells (d0 = 100%). (D) Representative flow cytometric analysis of targeted cells on day 0 and day 8. Bar charts represent mean ± SD; n = 3 per clone. Student’s t-test was performed on pooled SG-clones (ROSA26, n = 9; TIGRE, n = 9) and C-SG clones (ROSA26, n = 12; TIGRE, n = 9); **p < 0.01, ****p < 0.0001. www.nature.com/scientificreports/ 9 SCIEnTIfIC RepoRTS | 7: 7919 | DOI:10.1038/s41598-017-04212-8 clones, which was increased to 71–89% eGFP-positive cells in C-SG clones (Fig.6C). Representative flow cytometric analyses are shown in Fig.6d. Taken together, while different levels of epigenetic repression were observed in these two defined chromosomal sites, the CBX3 was able to significantly stabilise transgene expression at both loci. This proves that UCOEs can overcome epigenetic restrictions in defined chromosomal sites, thereby justifying the use of bulk populations to investigate UCOE functions. Discussion UCOEs such as the A2UCOE have been demonstrated to efficiently negate epigenetic repression of transgene expression from retroviral vectors in a variety of cell lines and primary cells including hematopoietic stem cells as well as murine and human pluripotent ESCs and iPSCs and their differentiated progeny9–11, 20. This protective activity has been linked with reduced promoter CpG methylation as well as decreased levels of repressive and increased levels of active histone marks in the chromosomal neighbourhood of UCOEs. While traditionally bidirectional promoter activity has been deemed a necessary trait for the anti-silencing function of UCOEs, we recently described the 0.7 kb CBX3 element that lacks the HNRPA2B1 promoter moiety of the A2UCOE but still potently stabilizes lentiviral transgene expression in multipotent and pluripotent stem cells15. Thus, the CBX3 represents a unique single-promoter UCOE that still shows most if not all of the anti-silencing activity of the A2UCOE and due to its small size it does not adversely affect viral titres15. Although the CBX3 has been shown to protect transgene expression from viral, housekeeping and tissue-specific promoter elements, a link of this activity to distinct structural features of the CBX3 is still missing. To address this point, we investigated the anti-silencing ability of various CBX3 subfragments. While all subfragments except for the 5′ CBX3(1-339) fragment showed substantial anti-silencing activity in the pluripotent state, in differentiated cells two subfragments, CBX3(1-339) and CBX3(340-508), failed to show anti-silencing activity. The remaining subfragments CBX3(1-508), CBX3(85-508), CBX3(170-508) and CBX3(503-679) all showed activity in pluripotent as well as differentiated cells. These data indicate different mechanisms to be responsible for transgene silencing in the pluripotent state or during cellular differentiation. While transgene silencing in the pluripotent state is well documented1, 9, 10, 21, also the extensive chromatin remodelling associated with the exit from pluripotency has been advocated as a factor that impairs retroviral transgene expression. In this context, loss of transgene expression in differentiated cells has been described by a number of groups6, 22, and this notion is further supported by previous data from our group describing profound transgene silencing as well as anti-silencing efficacy of the A2UCOE particularly during the early differentiation state of ESCs and iPSCs when cells leave pluripotency9, 10, 15. Of note, our data imply that the anti-silencing function of the CBX3 element cannot be mapped to a single region and thus emphasize the CBX3 element as a minimal UCOE version with almost complete functionality of the original 1.5 kb A2UCOE. While all subfragments displayed at least some level of activity, none of the fragments could completely substitute for the CBX3 in our experimental models. Nevertheless, the considerable activity of the CBX3(503-679) and the three subfragments containing the 170-508 bp sequence indicate the relevance of the central as well as the 3′ portion of the CBX3 in this context. In contrast, the rather low activity of the CBX3(1-339) subfragment and the similar properties of the CBX3(1-508), CBX3(85-508), and CBX3(170-508) subfragments suggest only moderate contribution, if any at all, from the 5′ end of the CBX3. We have not been able, however, to generate CBX3 subfragments lacking only the 5′ 170 or 339 bp to more formally test this hypothesis. Importantly, we demonstrated a significant correlation of fragment length and, even more important, overall CpG content with anti-silencing activity, further supporting the notion that the highly CpG-dense central and 3′ regions may be crucial determinants of CBX3 function. The CpG-islands within the A2UCOE have long been considered a central component of the anti-silencing function. Indeed, we now show that a scrambled CBX3 element (CBX3scrCpG) that bears no resemblance to the CBX3 (or any other UCOE) besides maintained CpG-sites, has the ability to protect lentiviral transgene expression in pluripotent cells and thereof differentiated progeny. A completely scrambled CBX3 variant (CBX3scr) lacking the CpG-sites displayed minimal to no functionality. Although not completely equivalent in functionality to the CBX3 element, this ability of the CBX3scrCpG is quite remarkable, and strongly indicates the CpG-sites of the CBX3 as important structural determinants of its anti-silencing function. CpG-islands are detected in the proximity of 60-70% of gene promoters23, 24, and their non-methylated CpG-sites have been identified as specific targets of various histone modifying proteins as well as DNA demethylating enzymes (reviewed in ref. 25). All of these proteins including amongst others Cfp1 (CxxC finger protein 1), Mll (mixed lineage leukemia protein), or Kdm2a (H3K36 demethylase) share a common feature, a zinc finger CxxC binding domain that specifically recognizes and binds non-methylated CpG-sites26. Specifically Cfp1 has been shown to associate with non-methylated CpGs both in vitro and in vivo27 and to interact with the H3K4 methyltransferase Setd128, 29, thereby leading to the deposition of the active chromatin mark H3K4me3. Interestingly, artificial, promoter-less, non-methylated CpG-stretches were able to recruit Cfp1 and create new H3K4me3 marks when integrated into the genome of mouse ESCs27. Thus, future studies that investigate the physical interaction of CBX3 with such CxxC-domain containing proteins may help to shed light on the molecular mechanism by which the CpG-islands of the CBX3 mediate its anti-silencing function. Moreover, regions highly enriched in non-methylated CpG-dinucleotides (CpG islands) are associated with an open chromatin structure that particularly in pluripotent cells is marked by a bivalent chromatin containing both active as well as repressive chromatin marks30. Synthetic CpG-islands have been shown to induce this bivalent state in ESCs if the synthetic sequence in addition to the CpG-dinucleotides is also enriched for G and C nucleotides31. This may imply that in addition to the CpG-sites also the overall GC-density of the CBX3 and thereof derived subfragments supports the overall function of these elements. Considering the impact of transcriptional activity on the chromatin status and the recruitment of binding factors, such as Cfp1, in the vicinity of CGIs, it becomes evident that the promoter activity of CBX3 may