Research ArticleANTIBODIES

Noncoding RNA transcription alters chromosomal topology to promote isotype-specific class switch recombination

See allHide authors and affiliations

Science Immunology  07 Feb 2020:
Vol. 5, Issue 44, eaay5864
DOI: 10.1126/sciimmunol.aay5864

“Lnc”ing class switch recombination

Besides protein-coding transcripts, a number of transcripts, including microRNAs and ribosomal RNAs, have noncoding functions. One of the less understood classes of noncoding transcripts are long noncoding RNAs (lncRNAs). Rothschild et al. have identified lncRNA-CSRIgA, an lncRNA ~2.6 megabases downstream of immunoglobulin heavy-chain locus, to be a regulator of class switch recombination (CSR). Deletion of lncRNA-CSRIgA impaired class switching in B cells, particularly to the IgA isotype. Deletion of the genomic region encoding lncRNA-CSRIgA not only abrogates transcription but also results in loss of cis-regulatory functions. Lentiviral expression of lncRNA-CSRIgA rescued class switching in lncRNA-CSRIgA–deficient B cells, strongly suggesting that lncRNA does have a functional role in promoting class switch recombination.


B cells undergo two types of genomic alterations to increase antibody diversity: introduction of point mutations into immunoglobulin heavy- and light-chain (IgH and IgL) variable regions by somatic hypermutation (SHM) and alteration of antibody effector functions by changing the expressed IgH constant region exons through IgH class switch recombination (CSR). SHM and CSR require the B cell–specific activation-induced cytidine deaminase (AID) protein, the transcription of germline noncoding RNAs, and the activity of the 3′ regulatory region (3′RR) super-enhancer. Although many transcription regulatory elements (e.g., promoters and enhancers) reside inside the IgH and IgL sequences, the question remains whether clusters of regulatory elements outside IgH control CSR. Using RNA exosome–deficient mouse B cells where long noncoding RNAs (lncRNAs) are easily detected, we identified a cluster of three RNA-expressing elements that includes lncCSRIgA (that expresses lncRNA-CSRIgA). B cells isolated from a mouse model lacking lncRNA-CSRIgA transcription fail to undergo normal levels of CSR to IgA both in B cells of the Peyer’s patches and grown in ex vivo culture conditions. lncRNA-CSRIgA is expressed from an enhancer site (lncCSRIgA) to facilitate the recruitment of regulatory proteins to a nearby CTCF site (CTCFlncCSR) that alters the chromosomal interactions inside the TADlncCSRIgA and long-range interactions with the 3′RR super-enhancer. Humans with IgA deficiency show polymorphisms in the lncCSRIgA locus compared with the normal population. Thus, we provide evidence for an evolutionarily conserved topologically associated domain (TADlncCSRIgA) that coordinates IgA CSR in Peyer’s patch B cells through an lncRNA (lncRNA-CSRIgA) transcription-dependent mechanism.


A remarkable set of developmentally controlled DNA rearrangement and alteration steps in the immunoglobulin (Ig) locus of mammals leads to the generation of antibodies able to interact with the vast number of host-encountered antigens. First, developing progenitor B cells undergo VDJ recombination to accomplish antigen-independent diversity of the B cell receptor and secreted antibodies (13). Thereafter, diversity of the Ig loci is further increased through the somatic hypermutation (SHM) and class switch recombination (CSR) mechanisms, which are both antigen-driven processes (46).

Rapid progress has been made in understanding the spatial organization of the mammalian genome, including the identification of chromosomal subcompartments known as topologically associating domains (TADs) (79). A TAD is a genomic region enriched with regulatory DNA sequences that physically interact with each other more frequently than with sequences outside the TAD. DNA loops that encompass a TAD are often defined by CTCF (CCCTC-binding factor) sites at loop anchors that promote DNA element interactions during TAD formation by a process known as loop extrusion (7). The IgH locus requires several key interactions to accomplish productive DNA alterations that facilitate VDJ rearrangement, SHM, and CSR including interactions between and among (i) the two enhancer elements iEμ and 3′ regulatory region (3′RR); (ii) the promoters placed upstream to isotype switch sequences that drive germline transcription, a prerequisite for CSR; and (iii) CTCF/cohesin binding sites, some of which are located surrounding the 3′RR. The Ig heavy-chain locus (IgH) is restricted to a single TAD that spans from the VDJ exons to the 3′RR, with the 3′RR forming a sub-TAD within the IgH locus TAD (TADIgH). These various elements are involved in controlling DNA loops during VDJ recombination and promoting accessibility of recombination substrates via mechanisms involving chromatin scanning and formation of DNA double-strand breaks at CTCF/cohesin binding sites during TAD formation (1013). The complex regulation of DNA rearrangements and transcription control occurring inside the IgH locus provides a noteworthy opportunity to study various mechanisms that drive TAD regulation (9, 14). Whether regions outside the IgH locus play roles in controlling genomic rearrangements and whether these sequences control tissue context-specific antibody diversification processes remain unclear. Flanking the IgH locus, a series of TADs exists whose function in regulating IgH function is not understood and that have not been investigated previously.

In the literature, various protein factors and transregulatory elements have been shown to control CSR and SHM (15, 16), but there is more to learn about isotype-specific and tissue/microenvironment-specific regulation of antibody diversification, and although some evidence exists regarding inter-TAD and intra-TAD interactions, the physiological meaning of such interactions is only starting to emerge (1720). In this regard, in the Peyer’s patches (PPs) present in the small intestine of mice, there is robust CSR to the IgA-specific isotype, and thus, PPs provide a defined environment for identifying and characterizing tissue-specific antibody gene diversification elements. IgA is the most abundant antibody secreted in the intestinal tract. IgA interacts with a small plasma cell-derived polypeptide termed joining (J) chain to form IgA dimers that recognize polymeric Ig receptor (pIgR) on the basolateral surface of intestinal epithelial cells (IECs). pIgR facilitates the release of secretory IgA onto the surface of the gut that increases the stability of secretory IgA in the intestinal lumen and anchors secreted IgA (sIgA) to mucus. Secretory IgA favors both maintenance of noninvasive commensal bacteria and neutralization of invasive pathogens through multiple mechanisms. By using the V region of IgA, sIgA blocks certain bacterial epitopes from interacting with the apical surface of IECs. In addition, sIgA limits microbial motility by nonspecifically binding bacteria through glycans associated with the secretory component and constant region α (Cα) of IgA. Besides neutralizing pathogens in the intestinal lumen, IgA can intercept microbes and toxins inside IECs. sIgA delivers these protective functions without activating the complement cascade, and it also impedes inflammatory damage to the epithelial barrier (21). Accordingly, it would be both interesting and beneficial to identify new mechanisms that control IgA class switching in PPs.

Here, we observe that an evolutionarily conserved TAD that is present about 2.6 Mb away from the IgH locus is important for controlling specific IgA CSR in mice. This newly identified conserved TAD (named TADlncCSRIgA) regulates IgA CSR by promoting specific intra-TAD interactions as well as interactions with the hypersensitivity sites of the 3′RR as previously shown (22). Our study is a rare example of molecular dissection of intra-TAD interactions during B cell development and function and identifies a role for TAD regulation in tissue-specific regulation of antibody diversification.


High-throughput chromosome capture–mediated identification of TADlncCSRIgA

We and others have reported previously that an 11-subunit complex, known as RNA exosome, regulates the levels of noncoding RNAs in mammalian cells (2227). In the absence of RNA exosome activity, accumulation of noncoding RNAs eventually leads to formation of DNA/RNA hybrids at various transcription regulatory elements in the genome, leading to the onset of genomic instability (22, 28, 29). Exosome activity deletion in cells substantially reduces the decay of long noncoding RNAs (lncRNAs) and enhancer RNAs (22, 23, 30). From the transcriptome of B cells that lack RNA exosome activity, we identified three bidirectionally transcribed regions that express RNA exosome–sensitive ncRNAs [schematized in Fig. 1A; one of the three loci, noncoding lncRNA-CSRIgA expressing lncCSRIgA, was previously identified in (22)]. Of all the ncRNAs that are expressed from mouse chromosome 12 at levels twofold or higher in RNA exosome–deficient B cells (in comparison with the wild-type B cell transcriptome), these bidirectionally transcribed regions are located closest to the IgH locus. We named these three regions as lncCSRIgA, locus A, and locus B. We performed high-throughput chromosome capture (Hi-C) in B cells and found that these three regions exist in a TAD. For the purpose of continuity here, we named this topologically associating cluster as TADlncCSRIgA. The TADlncCSRIgA is located about 2.6 Mb away from the IgH 3′RR (Fig. 1C), and other TADs are found to be present between TADlncCSRIgA and TADIgH (fig. S1A). We looked closely at the TADlncCSRIgA (Fig. 1D) and observed that the bidirectional transcription from locus A to locus B is placed next to regions that interact with the lncCSRIgA locus (Fig. 1, B and E). Thus, we considered that locus A and locus B interact with lncCSRIgA inside the TADlncCSRIgA. These observations are reproducible both in the gut-derived B cell line CH12F3 and in splenic primary B cells (Fig. 1F), indicating the existence of a functional TAD across different forms of B cells.

Fig. 1 lncCSRIgA exists in a TAD containing two additional interactors (loci A and B).

(A) A linear map of the location of the IgH, specifically the IgM and IgG1 regions. The lncCSRIgA region, locus A, and locus B are transcribed regions that express RNA exosome substrate ncRNAs. (B) RNA-seq used for identification of bidirectionally transcribed lncCSRIgA, locus A, and locus B in (A). The expressed ncRNAs are identified in RNA exosome–proficient wild-type (WT) B cells (Exosc3WT/WT) and RNA exosome–deficient B cells (Exosc3COIN/COIN). (C) Map of the TADlncCSRIgA separated from the IgH constant region genes; the 3RR forms a TAD with intronic enhancer Eμ. (D and E) Visualizations of TADlncCSRIgA at different resolutions. The green arrowhead indicates locus A; the orange arrowhead indicates lncCSRIgA, and the blue arrowhead indicates locus B. (F) The TADlncCSRIgA is identified in both CH12F3 cells and primary B cells, with primary B cell TAD on the left and CH12F3 cell TAD on the right. The black square indicates interacting area of lncCSRIgA and locus A, and the blue square indicates interacting area of lncCSRIgA and locus B. (G) Interactions between lncCSRIgA, locus A, and locus B are conserved in mouse (left-hand side) and human (right-hand side) B cell–derived cell lines. The regions mm9 chr12:111,615,000-112,170,000 of the CH12-LX cell line and hg19 chr14:102,100,000-102,900,000 of the GM12878 cell line Hi-C contact matrix were visualized in Juicebox. (H) Rare variants [single-nucleotide variant (SNV) and InDels] in selective IgA-deficient human patients align in remarkable congruence with locus lncCSRIgA. Ten patients with sIgAD having rare variants observed on both alleles of the lncCSRIgA locus are shown; the frequency of these rare variants is two- to fourfold higher than the local population (Sweden) and is shown to be significant (+P ≤ 0.05).

lncCSRIgA TAD is conserved in human cells

The human and mouse genomes harbor a similar but not quite overlapping IgH locus structure (15, 31). We compared regions neighboring the 3′RR of mouse and human genomes for TAD formation in B cell lines and observed partly conserved TADs [Fig. 1G (left) for a mouse B cell line and Fig. 1G (right) for a human B cell line] using insulation scores as a parameter [details in legends of fig. S1 (B and C) and Supplementary Materials and Methods]. Thus, on the basis of the overall conservation of TAD distribution neighboring IgH, we wanted to investigate whether the TADlncCSRIgA is conserved in humans. As shown in Fig. 1G and fig. S1 (B and C), the interactions of lncCSRIgA/A, lncCSRIgA/B, and A/B occur and are conserved in the mouse and human TADs. Similarly, the distribution of CTCF sites can be inferred from the comparison of fig. S1D and fig. S1E because the presence and interactions of conserved CTCF sites would be expected to exist in such conserved loops seen between lncCSRIgA, locus A, and locus B. Conserved CTCF sites can be mapped and predicted to be convergently or divergently oriented (fig. S1, D and E). Overall, the TADs formed from TADIgH to TADlncCSRIgA are quite conserved in structure/organization. The genes expressed in the TADlncCSRIgA include Ppp2r5c, Dync1h1, Hsp90aa1, Mok, orphan gene17000001k19Rik, Wdr20, Zfp839, and an lncRNA neighboring the Mok gene lncRNAMok1. In some human patients with selective IgA deficiency (sIgAD), there is an accumulation of rare variant single-nucleotide polymorphisms and insertions/deletions (InDels) in the lncCSRIgA locus (Fig. 1H and fig. S1F). These variants are different from those found in a well-known previous genome-wide association study that identified common variants (32).

TADlncCSRIgA is important for Ig gene diversification and IgA expression in gut PPs of mice

To dissect the role of TADlncCSRIgA, we generated a knockout mouse model in which lncCSRIgA expression was genetically ablated. For a detailed schematic presentation of the knockout allele map and the genotyping, see fig. S2A. Figure S2B shows a sequencing test of the knockout of the lncCSRIgA region in the mouse genome; highlighted areas represent the 3′ arm and 5′ arm of the targeting vector and the replacement of the lncCSRIgA locus with an FRT site. Bone marrow–derived pro–B cells, pre–B cells, and IgM+B220+ immature B cells demonstrate comparable levels between littermate lncCSRIgA−/− and lncCSRIgA+/+ mice (fig. S2D). Among splenic B cells, there was no difference in the distribution of follicular B cells and marginal zone B cells (fig. S2E). Thus, the lncCSRIgA−/− mice did not demonstrate any defects in B cell development. Analysis of circulating serum for antibody isotypes through an enzyme-linked immunosorbent assay approach showed reduced levels of IgA (fig. S2C) without reduction in the expression of IgG1 and IgG2a. However, differences of serum Ig levels are often diminished because of serum accumulation in the animal over long periods of time, which complicates the analyses of class switching isotypes. Thus, we wanted to investigate the specific effect on IgA CSR in the PP B cells in the intestine. PP B cells isolated from lncCSRIgA−/− mice show reduced IgA CSR (Fig. 2B). To ascertain whether the decreased level of IgA expression in the PP of lncCSRIgA−/− mice was indeed detectable in the B cells resident in the small intestine tissues, we performed immunostaining. It is quite clear that in both the PP (Fig. 2C, top) and the small intestine villi (Fig. 2C, bottom), there are reduced numbers of IgA+ B cells in the lncCSRIgA−/− mice, in comparison with the lncCSRIgAWT/WT mice. IgA expression is critical for microbiome homeostasis in the intestinal tract (3336). It has been previously shown that the level of some bacteria, including segmented filamentous bacteria (SFB), is enriched in the gut of mice in the absence of IgA expression (37). Thus, we performed overall microbiome profiling and species-specific analyses. As shown in Fig. 2D, principal components analyses of overall gene expression of microbiota isolated from the lncCSRIgAWT/WT versus lncCSRIgA−/− mice clearly showed a difference because the fecal material 16S ribosomal RNA (rRNA) transcriptomes of lncCSRIgAWT/WT mice clustered separately from the lncCSRIgA−/− mice. These differences were observed in lncCSRIgAWT/WT and lncCSRIgA −/− littermates that were cohoused. Moreover, microbiota-specific primers were used to analyze the content of various well-established microbiome components colonized in these mice; we observed a clear increase in SFB levels and Enterobacteriaceae levels in the lncRNAIgA−/− mice (fig. S3A). Increased SFB and Enterobacteriaceae levels are correlated with inflammation in the gut (35, 38, 39). Thus, we looked for signs of increased inflammation in the gut of the lncCSRIgA−/− mice. We find that in lncCSRIgA−/− mice, the structure of the intestinal villi is disrupted (fig. S3B, left) and locations of clear inflammation were observed more frequently in intestinal Swiss rolls, compared with what is seen in lncCSRIgAWT/WT littermate control cohoused mice (fig. S3B, right).

Fig. 2 lncCSRIgA knockout B cells have specific attenuated IgA CSR in PPs and physiological consequences.

(A) A mouse model was generated in which the lncCSRIgA locus was ablated [targeting strategy shown in (A) and fig. S2A]. (B) B cells isolated from the PPs of these mice show a significant decrease in CSR efficiency (P ≤ 0.01, n = 5 littermate pairs). Left: IgA CSR in B cells isolated from PPs of lncCSRIgAWT/WT and lncCSRIgA−/− mice, along with CSR analyzed in PP B cells from AID−/− mice. Right: Data quantitation, where **P ≤ 0.01, by Student’s t test. (C) Immunohistochemical staining for IgA in the small intestine of lncCSRIgAWT/WT and lncCSRIgA−/− mice; staining was performed from sections showing PP (top) and intestinal villi (bottom). (D) A principal components analysis of the metagenomics of the intestinal bacterial flora of lncCSRIgAWT/WT, lncCSRIgA−/−, and AID−/− mice by high-throughput sequencing of 16S rRNA. Note that lncCSRIgAWT/WT, lncCSRIgA−/−, and AID−/− metagenomics data cosegregate.

To evaluate whether, in lncCSRIgA−/− mice, the defect in IgA CSR in PP B cells is cell intrinsic and whether CSR to other isotypes is perturbed, we performed cytokine-induced ex vivo CSR assays. We observed that lncCSRIgA−/− B cells have a clear CSR defect to IgA and IgG2b isotypes but not to IgG1 and IgG3 (Fig. 3, A and B). This was not due to changed AID expression (Fig. 3D) or altered B cell proliferation (Fig. 3C). Consistent with our observation with primary B cells, transformed B cell lines that undergo CSR only to IgA (CH12F3) also show a decreased efficiency in CSR from IgM to IgA (fig. S4, A and B) (22). We wanted to investigate whether there is altered SHM frequency in the lncCSRIgA−/− mice. For this purpose, we isolated germinal center B220+GL7+ B cells from the PPs of lncCSRIgA−/− and cohoused littermate control lncCSRIgAWT/WT mice [the fluorescence-activated cell sorting (FACS)–sorted population is shown in fig. S4C]. We isolated total PP B cell DNA and performed SHM sequencing (SHM-seq). As a control for the SHM-seq experiment, we also performed SHM-seq from AID−/− PP B cells. As shown in fig. S4D, the level of mutation in the PP B cells obtained from the lncCSRIgAWT/WT JH4 region is higher than that seen in the PP B cells from lncCSRIgA−/− mice. The overall mutation frequency at AID hotspots for G to A conversion is lower in the lncCSRIgA−/− B cells, in comparison with lncCSRIgAWT/WT cohoused littermate mice (fig. S4E). It is possible that the effect on SHM is due to overall decreased levels of IgA+ B cells, which, in littermate control mice, are selectively chosen for SHM.

Fig. 3 lncCSRIgA knockout B cells have decreased IgA and IgG2b CSR defects in ex vivo cytokine-stimulated cultures.

(A) Purified splenocytes from lncCSRIgAWT/WT, lncCSRIgA−/−, and AID−/− mice were cultured for 72 hours in media with components to promote expression of the different isotypes indicated. IgA expression in the lncCSRIgA−/− mice is specifically attenuated (along with partnering IgG2b). (B) Quantitation of data presented in (A). ns, not significant; **P ≤ 0.01 and ***P ≤ 0.001, by Student’s t test (C) Cell division of lncCSRIgAWT/WT, lncCSRIgA−/−, and AID−/− B cells grown with cytokine stimulation in ex vivo culture conditions measured by VPD450 dye dilution method. (D) Failure to undergo CSR in the lncCSRIgA−/− mice compared with the lncCSRIgAWT/WT mice is not owed to a diminution of AID protein expression as shown by Western blot against AID.

Organization of TADlncCSR-IgA

Our efforts turned toward understanding the molecular mechanism(s) underpinning the role of the lncCSRIgA locus and the TADlncCSRIgA in altering antibody gene diversification in the gut. First, to evaluate whether the interaction of the lncCSRIgA locus regulates CSR by interacting with locus A and/or locus B (see Fig. 1 for details of lncCSRIgA interaction with A and B loci), we generated multiple locus A and locus B knockout/mutant CH12F3 cell lines. Figure 4A demonstrates the locations of locus A and locus B, with locus A located 128 kb and locus B located 250 kb away from lncCSRIgA. Both locus A and locus B have neighboring CTCF binding sites that may have a role in promoting lncCSRIgA interaction with locus A and locus B. Using chromatin immunoprecipitation sequencing (ChIP-seq) data for CTCF and RAD21 (cohesin subunit) DNA binding in B cells, we found that there are CTCF- and RAD21-bound CTCF sites neighboring lncCSRIgA, locus A, and locus B. (Fig. 4A). CTCF sites flank locus A (locus ACTCF) and locus B (locus BCTCF), whereas 7 kb further from lncCSRIgA, a CTCF (lncCSRIgACTCF)– and a RAD21-occupied site exist (Fig. 4A). To characterize the lncCSRIgA sequence for regulatory properties, we checked deoxyribonuclease I (DNase I) hypersensitivity, enhancer marks (H3K27Ac and H3K4me1), and MED1 binding. We found that the lncCSRIgA locus shows all these marks (as does the 3′RR with which it interacts), establishing its role as a potential transcription enhancer sequence (Fig. 4B).

Fig. 4 CTCF binding sites neighboring locus A and lncCSRIgA interact to promote CSR, whereas sites neighboring locus B and lncCSRIgA interact to suppress CSR.

(A) The expression of ncRNAs from Exosc3WT/WT and Exosc3COIN/COIN cells and the presence of CTCF, RAD21 (cohesin), and MED1 binding. (B) Detection of enhancer marks H3K4me1 and H3K27Ac overlapping the lncCSRIgA locus and CTCF, MED1, and RAD21 binding. (C and D) (D) is a zoom-in of (C) showing the enhanced CTCFlocusB-CTCFlncCSR interaction (black arrow) in B cells obtained from lncCSRIgA−/− mice in comparison with those obtained from lncCSRIgAWT/WT mice. (E) A quantitative estimate of CTCFlocusB-CTCFlncCSR interaction of (C) from three biological repeats. (F) Quantitation of the results of CTCFlocusB-CTCFlncCSR interactions with 4C analyses with baits placed immediately outside of lncCSRIgA (H3) or next to CTCFlncCSRIgA (H4 bait). (G) Quantitation of 4C-seq results from (F).

The interaction of CTCF sites flanking lncCSRIgA, locus A, and locus B sequences regulates TAD dynamics that drive IgA CSR

We next wanted to determine whether lncCSRIgA and associated regions physically interact with the 3RR of the TADIgH, located about 2.6 Mb away. Using 3C experiments, we were able to show that the CTCFlncCSR (neighboring the lncCSRIgA locus) is able to interact with the HS4 region of 3′RR, with deletion of lncCSRIgA leading to a decrease in this interaction frequency (fig. S5A). These inter-TAD interaction frequencies are weaker in comparison with intra-TAD interactions but are clearly detectable in our 3C assays (fig. S5A). Overlap of the lncCSRIgA regions with published DNA break sites, as observed in END-seq assays, clearly shows that both the lncCSRIgA locus and the CTCFlncCSRIgA accumulate DNA double-strand breaks in an etoposide treatment–dependent fashion (which prevents repair of Top2B-mediated DNA breaks) (fig. S5B). All these observations point toward a long-range interaction between the TADlncCSRIgA and the TADIgH that potentially collaborates or competes with short-range intra-TADlncCSRIgA interactions.

We wished to understand in greater detail the molecular interactions between lncCSRIgA, locus A, locus B, and 3′RR in the regulation of IgA CSR in the TADlncCSRIgA. Therefore, we wanted to investigate whether the loss of lncCSRIgA (in lncCSRIgA−/− B cells) effectively leads to altered interaction of the neighboring CTCFlncCSRIgA site to the locus B region. We generated Hi-C maps of B cells obtained from lncCSRIgAWT/WT and lncCSRIgA−/− mice and observed very comparable Hi-C map patterns in the overall B cell genome. Thus, the lncCSRIgA−/− mutation does not have a genome-wide defect in B cell genome organization (fig. S5D). Next, we compared Hi-C maps of lncCSRIgAWT/WT and lncCSRIgA−/− mice at genome coordinates covering the TADlncCSRIgA and TADIgH. We found an increased interaction of the lncCSRIgA/CTCFlncCSRIgA region with the B locus in lncCSRIgA−/− B cells compared with lncCSRIgAWT/WT B cells (Fig. 4, C to E), the observation being quantitated over multiple independent experiments in Fig. 4E using B cells of littermate lncCSRIgAWT/WT and lncCSRIgA−/− mice. These observations indicate that the CTCF neighboring the lncCSRIgA locus (CTCFlncCSRIgA) interacts with both the locus A and the locus B sequences, with the interaction with the locus B sequence becoming stronger when the lncCSRIgA locus is deleted.

Next, we developed a 4C sequencing (4C-seq) assay with baits immediately adjacent to the lncCSRIgA locus (H3 bait) and one next to the CTCF site located 3 to 4 kb downstream (H4 bait). In our 4C-seq assays, we used the H3 bait and learned that the lncCSRIgA region physically interacts with the locus B region. Consistent with the Hi-C data of Fig. 4 (C to E), this interaction of lncCSRIgA and locus B is increased in the lncCSRIgA−/− B cells (Fig. 4, F and G, top). The interactions between lncCSRIgA and locus B should be occurring via neighboring CTCF sites [portrayed in Fig. 4 (A and B)]. To interrogate CTCF interactions, the H4 bait (neighboring the CTCF site) was found to bind with the locus B sequence more efficiently in lncCSRIgA−/− B cells (Fig. 4, F and G, bottom). Both in the Hi-C assays and the 4C-seq assays, the effect of CTCFlncCSR to locus A was not detectably altered (fig. S5C, top). In addition, from RNA sequencing (RNA-seq) of PP B cells, we did not observe a marked change in local gene expression overlapping locus A or locus B in lncCSRIgA−/− B cells (fig. S5C, bottom). Overall, these observations point toward an interaction of CTCFlncCSR with CTCFlocusB, with transcription of the lncCSRIgA locus being important for weakening the strength of this interaction in vivo (in lncCSRIgA−/− cells).

Effect on CSR in TADlncCSRIgA mutants

We wanted to understand what the effects of the loss of locus A, locus B, and CTCFlncCSRIgA on IgA CSR are. Locus A and locus B are placed 120 and 250 kb from the lncCSRIgA locus, respectively (Fig. 5A), allowing us to make deletion mutants using CRISPR-Cas9 gene editing technology. We generated CH12F3 cell lines that lack each of these elements. We performed CSR assays from each of these lines to evaluate the effect on CSR; loss of locus B sequence leads to increased IgA CSR (Fig. 5B), whereas loss of locus A sequence leads to decreased IgA CSR (Fig. 5C). These observations were obtained from three independently isolated CH12F3 clones. Individual flow cytometry plots demonstrating the experiments in Fig. 5 (B and C) are shown in fig. S6 (A and B). Lack of lncCSRIgA or locus A sequence does not ablate CSR but, rather, reduces efficiency. Last, in the absence of CTCFlncCSR, there is a decreased IgA CSR (Fig. 5, D and E), indicating that CTCFlncCSRIgA could be a pivotal site for directing interactions inside the TADlncCSR so that efficient IgA CSR can occur.

Fig. 5 Characterization of various TADlncCSR mutants for CSR.

(A) A cartoon demonstrating the locations of locus A, lncCSRIgA, and locus B relative to the 3′RR of the IgH locus. (B) CSR efficiency increases in locus B mutants in comparison with WT (CH12F3) controls, but this increase is lost in lncCSRIgA−/−, locus B−/− double mutants. Three individual clones are shown (**P ≤ 0.01, by Student’s t test). (C) CSR to IgA is decreased in CH12F3 cells that are deleted of locus A (**P ≤ 0.01, and ***P ≤ 0.001, by Student’s t test); three independent clones are portrayed. (D and E) CRISPR-Cas9–mediated editing of CTCFlncCSRIgA site leads to decreased CSR in CH12F3 cells. (D) FACS plots of one experiment. (E) Representation of four separate biological repeats of (D). The experiment was done with four independently derived clones (n = 4; each clone was assayed three times).

lncRNA-CSRIgA recruits CTCF cofactors to the CTCFlncCSRIgA site and partly regulates IgA CSR

At this point, we wanted to ask whether the lncRNA-CSRIgA (exosome-sensitive transcript from the lncCSRIgA locus) is important for orchestrating IgA CSR. Using lentiviral transduction, we introduced lncRNA-CSRIgA in lncCSRIgA−/− CH12F3 cells [as control, we separately inserted an inverted sequence to lncRNA-CSRIgA (INV) in CH12F3 cells]. The transduced lncRNA-CSRIgA vector expressed green fluorescent protein (GFP), and these cells were therefore easily identified by flow cytometry. The overexpression of lncRNA-CSRIgA in lncCSRIgA−/− cells rescued the CSR defect (Fig. 6, A and B), but neither the vector control alone nor the inverted version of the lncRNA-CSRIgA was able to do so (Fig. 6, A and B). The expression level of the lncRNA-CSRIgA is 10-fold or higher than the endogenous levels of lncRNA-CSRIgA expressed in these cells (Fig. 6C). We then wanted to determine whether the knockdown of the lncRNA-CSRIgA affects CSR. We introduced CRISPR-Cas13a along with two different sets of lncRNA-CSRIgA–directed guide RNAs (gRNAs) into the normal CH12F3 cells. As can be seen in Fig. 6D, both sets of gRNAs successfully knocked down lncRNA-CSRIgA in the CH12F3 cells. The expression of IgSμ germline transcripts and AID expression was not perturbed by the introduction of the Cas13a-based knockdown system, but the expression of lncRNA-CSRIgA was significantly reduced (Fig. 6D). In these conditions, IgA CSR was decreased by 30 to 45% (Fig. 6E). The 30 to 45% decrease of IgA CSR after lncRNA-CSRIgA knockdown was consistent with the loss of the complete DNA locus lncCSRIgA, where the effect is slightly stronger (in the range of 50 to 70%). Thus, both through overexpression (Fig. 6, A to C) and knockdown (Fig. 6, D and E) of lncRNA-CSRIgA, we were able to show a role of the RNA moiety expressed in the lncCSRIgA DNA locus in influencing CSR.

Fig. 6 The noncoding enhancer RNA, lncRNA CSRIgA, regulates the pivotal function of CTCFlncCSR.

(A) IgA CSR deficiency in lncCSRIgA−/− CH12F3 cells can be rescued by overexpression of lentivirally transduced lncRNA-CSRIgA. FACS plots of cytokine-stimulated CH12F3 B cells that are parental clones (CH12WT), CH12/lncCSRIgA−/−, and rescued either by an inverted sequence of lncCSR (lncRNA-CSRInv) or by the physiological lncRNA-CSRIgA (representative of three independent clones) are shown. (B) Plot summarizing multiple experiments that evaluate CSR rescue of lncCSRIgA−/− CH12 cells transduced with lncCSR-RNAIgA. The number of independently performed experiments is shown in the figure by black circles. **P ≤ 0.01 and ***P ≤ 0.001, by Student’s t test. (C) Quantitation by qRT-PCR of lncCSRIgA expression in lncCSRIgA−/− cells after lentiviral infection of the lncCSRIgA−/− cells. Note the logarithmic scale on the vertical axis. (D) Degradation of lncRNA-CSRIgA with a Cas13a/gRNA approach reduces IgA CSR efficiency. gRNAs specific to the lncRNA-CSRIgA were constructed to specifically knock down the lncRNA-CSRIgA (right) but did not alter the AID expression levels (left) or IgSμ transcription (center). **P ≤ 0.01 and ***P ≤ 0.001, by Student’s t test (E) Decreased IgA CSR in CH12F3 cells that express two different gRNA pairs that knocked down lncRNA-CSRRIgA expression: FACS plots (left), percent CSR decrease (middle), and ratio of CSR decrease in comparison with Cas13 with empty gRNA vector control (right). (F) Coimmunoprecipitation of a stem-loop–associated lncRNA-CSRIgA after expression in 293 T cells validates interaction with SMC3 (top blot), PARP1 (middle blot), and SUPT16H (bottom blot). (G) CTCF protein association to CTCF binding sites flanking locus A (CTCFLocusA), lncCSRIgA (CTCFlncCSRIgA), and locus B (CTCFLocusB) was analyzed by ChIP-seq experiments from B cells isolated from lncCSRIgAWT/WT and lncCSRIgA−/− littermate mice. (H) ChIP experiment to confirm that the recruitment of lncRNA-CSRIgA interacting factors (identified by mass spectrometry) SMC3, SUPT16H, and PARP1 to CTCFlncCSRIgA is strongly decreased. Reintroduction of lncRNA-CSRIgA in the lncCSRIgA−/− CH12F3 cells partly rescues the recruitment of the three factors to CTCFlncCSRIgA.

At this point, we wished to determine the role of the lncCSRIgA locus in modulating interaction of CTCFlncCSRIgA with locus A and locus B sequences. For this purpose, we identified lncCSRIgA-transcribed lncRNA (lncRNA-CSRIgA) interacting proteins by expressing a stem-loop tagged lncCSRIgA in human embryonic kidney (HEK) 293 T cells and performing biochemical purification of associated proteins in 293 T cells that express MS2 protein (MS2 protein interacts with stem-loops; see schematic in fig. S7A). To distinguish proteins that bind to the MS2 bead matrix or to the stem-loop tag from proteins that bind to the lncRNA-CSRIgA ncRNA bait, we performed controls of purification with expression of the untagged lncCSRIgA (lncRNA-CSRIgA alone) or the MS2 tag fused with a GFP RNA (MS2-GFP-alone), along with the experimental tagged lncRNA-CSRIgA (MS2-GFP-lncCSRIgA). We purified the RNA-protein complex and evaluated interacting proteins using mass spectrometry. This approach provides a tool for evaluating proteins that bind to ncRNAs independent of transcription and locus positioning. We eliminated nonspecific binding proteins, considering the controls used, and mapped the proteins that specifically interacted with lncRNA-CSRIgA ncRNA. We found that Poly [ADP-Ribose] polymerase I (PARP1), SUPT16H, and structural maintenance of chromosomes (SMC3) are specific interactors with lncRNA-CSRIgA (fig. S7, B and C, and table S4). We evaluated whether these interactions could be confirmed through an immunoprecipitation-coupled Western blot. As can be seen in Fig. 6F, MS2-fused lncCSRIgA interacts with SUPT16H, SMC3, and PARP1 in coimmunoprecipitation assays. SUPT16H is a component of the FACT complex that is important for recruiting epigenetic regulators at promoters and enhancers (40). PARP1 is important for promoting DNA double-strand break formation at DNA regulatory elements (41), and SMC3 is a component of the cohesin complex (42). To ascertain whether, in B cells, these proteins regulate the activity of the CTCFlncCSR site, we performed ChIP with antibodies against CTCF and all three interacting proteins in lncCSRIgAWT/WT and lncCSRIgA−/− B cells (Fig. 6, G and H). We confirmed that in lncCSRIgA−/− B cells, there is a modest decrease in CTCF occupancy at the CTCFlncCSRIgA site but no detectable change at CTCFA or CTCFB or control V gene-associated CTCF sites (Fig. 6G). We found a significant loss of SMC3, SUPT16H, and PARP1 at CTCFlncCSR in the lncCSRIgA−/− B cells (Fig. 6H). Moreover, in cells that have overexpression of lncRNA-CSRIgA (overexpressing cells sorted through the coexpression of GFP), there is a substantial rescue of SUPT16H, SMC3, and PARP1 recruitment at the CTCFlncCSR region (Fig. 6H). Together, these observations indicate that recruitment of CTCF cofactors (FACT subunit SUPT16H, PARP1, and cohesin subunit SMC3) may be provoked upon expression of lncRNA-CSRIgA inside the TADlncCSRIgA and may be important for modulating CSR. Perhaps, the overexpression of lncRNA-CSRIgA through artificial means overrides its requirement of localized expression inside the TADlncCSR. When expressed at physiological levels, the lncRNA-CSRIgA should be expressed at close proximity to the CTCFlncCSR site to influence SUPT16H, PARP1, and cohesin recruitment.


Although transcription at enhancer sites and the ensuing enhancer RNA transcripts at various regions of the mammalian genome have been extensively identified and reported on, the requirement of such for gene regulation and organism development is incompletely understood (22, 24, 4346). In this study, we provide evidence of both a bidirectionally transcribed enhancer locus and an associated ncRNA having roles in controlling TAD interaction dynamics, recruitment of genome organization regulatory factors at critical sites, and antibody gene diversification mechanisms. We show that the lncCSRIgA locus that transcribes the ncRNA (lncRNA-CSRIgA) has particular characteristics of an enhancer element that interacts with two different loci: locus A and locus B inside the TADlncCSRIgA. Molecular dissection and genome topological analyses of the lncCSRIgA locus using genetically manipulated mouse models and mutated cell lines identifies a potential mechanism where interaction of lncCSRIgA with locus B is reduced after B cell stimulation through an ncRNA transcription-dependent fashion, eventually leading to physiological levels of robust IgA CSR and SHM. The availability of the lncCSRIgA locus (free of interaction with locus B) may promote its interaction with the IgG 3′RR locus HS4 site (24), eventually leading to robust IgA and IgG2b CSR [modeled in fig. S7 (D and E)]. All these interactions are facilitated by CTCF sites flanking the locus A, locus B, and two sites next to lncCSRIgA. Ablation of the pivotal CTCF sites, CTCFlncCSRIgA, used in lncCSRIgA/locus A and lncCSRIgA/locus B interactions, leads to defective IgA CSR. Molecular examination of the lncRNA-CSRIgA ncRNA transcript aligns it with a mechanism of recruitment of cohesion subunit SMC3 protein, FACT subunit SUPT16H, and PARP1 at the pivotal CTCFlncCSRIgA of TADlncCSRIgA [Fig. 6 and modeled in fig. S7 (D and E)]. All three proteins could be associated with the regulation of CTCF site biology. Because of topological stress at enhancers and CTCF sites, there are etoposide-sensitive DNA double-strand breaks (fig. S5B) (47), and PARP1 may play a role in resolving these DNA breaks. SMC3 is a component of the cohesion complex that interacts with CTCF to regulate loop extrusion (4850), and the FACT complex (40) often functions to regulate local DNA topology. Future studies will provide additional information regarding the mechanistic aspects of these proteins binding RNA at a CTCF site. At this moment, we do not have tools to completely quantitatively segregate the effect of the lncCSRIgA DNA element and the lncRNA-CSRIgA RNA element on CSR. It is possible that in lncCSRIgA−/− B cells, the loss of regulation of the CTCFlncCSR site and the decreased efficiency of lncCSRIgA enhancer to interact with 3′RR together cause the defective IgA CSR. However, detailed mutational analyses of lncCSRIgA, CTCFlncCSR, locus A, locus B, and the HS4 site of 3′RR will be required to fully understand the underlying mechanisms of the phenotypes that we report in this study. Together, this study provides a rare example of the role of a conserved TAD, an enhancer sequence, and an enhancer-associated ncRNA in regulating intra-TAD interactions that ultimately affect an important biological process (adaptive immunity).

Last, we have identified a regulatory element that expresses a functional ncRNA to control IgH recombination in B cells that undergo CSR in the gut. The regulatory element is conserved in humans and contains polymorphisms at a frequency higher than normal controls in IgA-deficient patients (Fig. 1H and fig. S1F). We note that only a subset of patients with IgAD carry the lncCSRIgA locus mutant alleles; the genetics and mechanism of induction may well differ in different patients depending on the distribution of many other variant alleles genome-wide. For example, there are three major major histocompatibility complex susceptibility haplotypes seen in patients with IgAD, and these may potentially biologically interact with distinct sets of coding and noncoding sequences, including lncCSRIgA. In lncCSRIgA−/− mice, IgA expression and SHM efficiency are reduced but not completely abolished in PP B cells. Thus, we think that the inter-TAD and intra-TAD interactions controlled by lncCSRIgA locus transcription and lncRNA-CSRIgA expression have a regulatory effect on antibody gene diversification, but other compensatory mechanisms that control IgA CSR exist. The reduced IgA expression and SHM in B cells in the gut were sufficient to alter the homeostatic levels of microbiome components. It will be of interest to characterize other TADs surrounding the IgH or IgL loci to determine whether different regulatory DNA elements in tissue-specific antibody gene diversification processes can be identified.

This study entices us to question just how general is the possibility that enhancer locus expressed ncRNAs or other ncRNAs cause changes in genome topology by altering CTCF site action. Given our example here of lncCSRIgA affecting control of a neighboring CTCF site, we looked genome-wide to see how often CTCF sites are found next to noncoding RNA–expressing unidirectional or bidirectional elements (fig. S8A). More often than not, in various TAD datasets available (fig. S8B), three-way interacting CTCF sites are neighbored by an ncRNA-expressing element (fig. S8, C and D, and table S5). We postulate that at least in the case of lncCSRIgA transcription and regulation of CTCFlncCSR residing inside the TADlncCSR, the local concentration of ncRNA is increased to influence biological events and such effects could also be artificially reconstituted by overexpressing the ncRNA from a different region of the genome [as in Fig. 6 (A and B)]. The possibility exists that these ncRNAs provide some secondary structural properties that promote macromolecular complex formation. It is too premature to conclude but it is worth speculating whether ncRNAs control CTCF biology and cofactor interactions genome-wide by localized regulation, but this study does raise this question and opens up avenues for interrogating the role of ncRNA biology in genome topology control.


Hi-C sequencing, 4c-seq, and RNA-seq genomic datasets

Genomic datasets have been uploaded in the National Center for Biotechnology Information for public access after publication (accession no. PRJNA486392).

Experimental models and protocols

Mouse models

Exosc3COIN/COIN;Rosa26 CreERt2 and lncCSRIgA−/− mouse models were generated and bred according to the Institutional Animal Care and Use Committee guidelines at Columbia University, New York.

Cloning and site-directed mutagenesis for generation of stem-loop structure fused lncRNA-CSRIgA construct

To identify and characterize lncRNA-CSRIgA interacting proteins in ex vivo conditions, we prepared Plasmid-A lncCSRIgA-MS2 (a chimeric RNA containing lncCSRIgA followed at the 3′ end by 24 copies of the MS2 hairpin). Restriction site Bgl II of the pSP72 cloning vector containing lncCSRIgA was converted into an Spe I site using a protocol reported earlier (51) to accommodate 24× MS2 loops at the 3′ end of lncCSRIgA, with the 24× MS2 loop fragment digested from another vector with Bam HI and Spe I restriction enzymes. Plasmid-B hemagglutinin (HA)–pMS2 (pEGFP-1 Clontech vector backbone) expressed the MS2 coat protein containing the N terminus HA-tag and C terminus GFP fusion proteins.

lncRNA-CSRIgA immunoprecipitation

We cotransfected plasmids A and B into HEK-293 T cells using PEI reagents (2). We also used two cotransfection controls, namely, (i) plasmid-A alone and (ii) plasmid-B alone, respectively, in HEK-293 T cells under identical conditions. HEK-293 T cells were harvested 72 hours after transfection, washed twice with cold phosphate-buffered saline (PBS), pelleted and lysed with 500 μl per 25 million to 30 million cells of lysis buffer [20 mM tris-HCl (pH 7.5), 100 mM KCl, 5 mM MgCl2, 5% glycerol, 0.5% NP-40, 1× protease inhibitors (Roche), RiboLock RNase (ribonuclease) inhibitor (Fermentas; 40 U/ml), and 1 mM dithiothreitol (DTT)]. After a 30-min incubation on ice, cells were slowly homogenized using a glass douncer (300 to 500 times) to achieve at least 85 to 95% cell lysis and subsequently pelleted at 20,000g for 20 min at 4°C. The protein lysate was collected and incubated with equilibrated anti-HA agarose and Protein G Sepharose beads. Cell lysate along with beads was incubated overnight at 4°C, with rotation.

Subsequently, beads were pelleted and washed thrice using wash buffer [20 mM tris-HCl (pH 7.5), 100 mM KCl, 5 mM MgCl2, 0.01% NP-40, 1× protease inhibitors (Roche), RiboLock RNase inhibitor (Fermentas; 40 U/ml), and 1 mM DTT] for 15 min at 4°C and pelleted at 500g for 5 min. After pulldown and washes, beads were divided into two parts: One-half was used for protein analysis; the other half was used for the isolation of lncRNA-CSRIgA. For RNA analysis, the pulldown material was digested with DNase and proteinase, and the RNA was precipitated with sodium acetate/ethanol for further analysis. Pulldown enriched with lncRNA-CSRIgA was detected by quantitative reverse transcription polymerase chain reaction (qRT-PCR) amplification. For protein experiments, lncRNA-CSRIgA was eluted using 250 μl of HA-peptide (Thermo Fisher Scientific, catalog no. 26184; 1 mg/ml) per elution based on principles originally described (52). Supernatant containing the lncRNA-CSRIgA and its interacting proteins was collected. The protein complexes recovered from purification were fractionated on a 4 to 20% SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gradient gel. Unknown proteins were directly excised from the gel, and mass spectrometry was performed.

Western blot analysis for lncRNA-CSRIgA interacting proteins

The protein complexes recovered from lncRNA-CSRIgA immunoprecipitation were fractionated on a 4 to 20% SDS-PAGE gradient gel and transferred to polyvinylidene difluoride membrane. Immunoprecipitation products were analyzed using anti-SUPT16H (Abcam, 204303), PARP1 (Active Motif, 39559), and SMC3 (Abcam, 9263).

lncCSRIgA mouse generation

With CRISPR-Cas9 assistance, small homology arms were designed for gene targeting. A colony PCR approach was used to amplify both the 5′ and 3′ homology arms (with a length of about 1.0 and 0.95 kb) from a positively identified C57BL/6 BAC clone (RP23: 209D11). The FRT-flanked neomycin cassette replaced 1121 base pairs (bp) of the genomic sequence encoding the lncCSR-IgA. The targeting vector was confirmed by restriction analysis after each modification step and by sequencing using primers inside the selection cassette into the 3′ end of the 5′ homology arm and likewise into the 5′ end of the 3′ arm. Primers in the vector sequence were also used to read into the 5′ and 3′ homology arms. Ten micrograms of the targeting vector was linearized and then transfected by electroporation into iTL C57Bl6 (IC1) embryonic stem (ES) cells. After selection with G418 antibiotic, surviving clones were expanded for PCR analysis to identify recombinant ES clones. Expanded clones were reconfirmed by PCR, and then, secondary confirmation of positive clones identified by PCR was performed by Southern blot analysis. Briefly, DNA was digested with Afl II, separated on an agarose gel, and transferred to nylon membrane. The digested DNA was hybridized with a probe targeted against the 3′ external region. Positive clones were further confirmed by Southern blot analysis using an internal probe that involved digestion of the DNA with Bam HI, electrophoretic separation on an agarose gel, transfer to nylon membrane, and hybridization with a probe targeted against the 5′ internal region. A primary and a backup clone were identified and injected. Recombinant ES cells were injected into BALB/c blastocysts, and chimerism was determined visually by the distribution of the black coat color on the pups. Four chimeras were identified and mated with FLPe deleter mice to excise the neomycin cassette, and the resulting black pups from this mating were genotyped. These somatic neo deleted mice were then mated with wild-type C57BL/6 mice. Positive progeny were termed germline neo deleted (GND) mice and were also confirmed for absence of the FLP transgene and bred further for experiments. All generations of these GND mice were confirmed by PCR genotyping using the following primers: 28647 Seq For and 286473crv with 36 cycles of PCR performed at 94°30″, 60°30″, and 72°1′30″ using Ex Taq. Animals with lncCSR-IgA intact generate a PCR product of 1677 bp, whereas animals with lncCSR-IgA deleted and only an Frt site remaining will generate a PCR fragment of 635 bp. All mouse experiments were performed in adherence with protocols approved by the Columbia University Institutional Animal Care and Use Committee.

Chromatin immunoprecipitation

A total of 3 × 106 CH12 lncCSRIgAWT/WT or CH12 lncCSRIgA−/− cells or primary mouse lncCSRIgAWT/WT or lncCSRIgA−/− B cells were isolated and cultured as required and subjected to addition of formaldehyde [27 μl/ml of cells of a 37% solution (Sigma-Aldrich, F8775; 500 ml)] for 10 min with gentle inversion/rotation before the addition of 100 μl/ml of 1.375 M glycine to quench the formaldehyde with further gentle inversion/rotation for 10 min. Cells were washed several times with PBS before being suspended in SDS lysis buffer with protease inhibitors [1% SDS, 10 mM EDTA, and 50 mM tris plus 1 “complete mini” EDTA-free protease inhibitor solution tablet (Roche) per 10 ml of lysis buffer]. Sonication was performed on ice using a Branson Sonifier 250 apparatus for 25 cycles, each cycle comprising 20 s of sonication at duty cycle 30%, followed by a 2-min rest period. After gel electrophoresis to gauge approximate amounts of material, the lysate was centrifuged at maximum speed for 10 min at 4°C, and 2% of the material was snap-frozen as input. Lysates were diluted appropriately and precleared with protein A/G plus agarose solution and ChIP-grade rabbit IgG for 1 hour. Immunoprecipitation of lysates was performed overnight at 4°C with gentle rotation using indicated antibodies (anti-SMC3, anti-PARP1, and anti-SUPT16H). Sonicated herring sperm DNA (Promega, D1815) and protein A/G plus agarose solution were added for 90 min with continued rotation. Subsequently, the beads were washed by the standard series of washes [low salt, high salt, LiCl, and Tris-EDTA (TE)], and ChIP products were eluted, followed by RNase A treatment overnight at 60°C and proteinase K treatment for 12 hours at 37°C and followed by 2 hours at 55°C. ChIP DNA was purified by phenol-chloroform-isoamyl alcohol addition (and recovered using phase-lock gel heavy tubes) and then precipitated using ethanol. ChIP products were evaluated with appropriate primers that were specific for the regions interrogated (e.g., CTCFlncCSRIgA).

CSR protocol

CH12F3 cells were plated at a concentration of 5 × 104 cells in 1 ml of appropriate medium [RPMI supplemented with 10% fetal bovine serum (FBS) and nonessential amino acids, sodium pyruvate, l-glutamine, Hepes, penicillin-streptomycin, and β-mercaptoethanol] and placed in one well of a cluster 12-well tissue culture plate (Corning Costar, 3512). Cells were plated in triplicate to reduce variation. Cells receiving stimulation were administered lipopolysaccharide (LPS), interleukin 4 (IL4), and transforming growth factor–β (TGF-β) (20 μg/ml, 20 ng/ml, and 1 ng/ml, respectively). All cells were placed in a 37°C tissue culture incubator with 5% CO2. At the time of examination, 200 μl of cells was removed from the well, washed with PBS/2% FBS, spun down, and resuspended in 100 μl of 2% FBS in PBS containing 0.1 μl of anti–IgA-phycoerythrin (PE) antibody. Cells were incubated with the antibody at 4°C for 30 min before being centrifuged and resuspended in 2% FBS/PBS and interrogated on an LSRFortessa (BD Biosciences) FACS machine.

CSR protocol of splenic B cells for Ig isotypes CSR

CD43-negative B cells from spleen were stimulated with LPS (20 μg/ml) (Sigma-Aldrich) for IgG3; LPS (20 μg/ml) and IL4 (20 ng/ml) (PeproTech) for IgG1; and LPS (10 μg/ml) (Sigma-Aldrich), TGF-β (2 ng/ml) (R&D Systems), and anti-IgD dextran (0.33 μg/ml) (Fina Biosolutions) for IgG2b and IgA CSR. CSR efficiency of each isotype was measured after 3 to 4 days. B220-PE (BioLegend)/IgG3-FITC (fluorescein isothiocyanate) (BD Biosciences) and B220-FITC (BD Biosciences)/IgG2b-PE (BioLegend), IgA-PE (eBioscience), or IgG1-PE (BD Biosciences) were used for staining cells for FACS analyses.


Materials and Methods

Fig. S1. TAD clusters neighboring 3′RR.

Fig. S2. lnCSRIgA deletion in the mouse genome.

Fig. S3. Physiological consequences of deletion of lncCSRIgA.

Fig. S4. CSR and SHM in lncCSRIgA−/− cells.

Fig. S5. TADIgH and TADlncCSR interaction and locus A and lncCSRIgA interaction.

Fig. S6. IgA CSR frequency of locus A and locus B mutant B cell lines.

Fig. S7. lncCSRIgA interaction with CTCF-associated cofactors and working model of action in TADlncCSRIgA.

Fig. S8. Interactions among multiple genomic regions infer potential coregulatory relationships between TADs or sub-TADs with ncRNA expression.

Fig. S9. Gating strategy for FACS analysis.

Fig. S10. Complete image (Fig. 3D).

Fig. S11. Complete image (Fig. 6F).

Table S1. List of reagents and resources.

Table S2. List of primers used.

Table S3. Table of conserved genes placed between mouse and human TADIgH and TADlncCSR.

Table S4. Table of lncRNA-CSR interacting proteins identified from two independently performed purifications.

Table S5. Table of interactions among multiple genomic regions infer potential coregulatory relationships.

Table S6. Raw data.

References (5361)


Acknowledgments: We thank T. Honjo (Kyoto University) for providing CH12-F3 cells and AID−/− mice, B. R. San Martin (IGCMB, Strasbourg) for ChIP-seq datasets, E. L. Aiden and A. Presser (Baylor College of Medicine) for advice regarding Hi-C data, and members of the Basu laboratory for valuable discussions; the Columbia Genome Center for high-throughput genome sequencing and the Department of Microbiology and Immunology core FACS facility as well as the Flow Cytometry Core, Columbia Stem Cell facility; and the Ingenious Targeting Laboratory for mouse model generation. Funding: This work was supported by grants to U.B. [NIAID (1R01AI099195 and Ro1AI134988), the Leukemia and Lymphoma Society, and the Pershing Square Sohn Cancer Research Alliance], B.L. (EMBO fellowship ALTF 906-2015), and L.N. (NIH grant T32 AI106711). Author contributions: G.R. designed, executed, analyzed, and interpreted the experiments and helped write the manuscript, assisted in all steps by J.L. who also undertook the deep sequencing and interaction studies. W.Z. undertook the bioinformatics analyses of most genomics experiments. P.K.G. performed experiments to determine lncRNA-CSRIgA interacting proteins and analyzed mass spectrometry experiments. Yuling Chen and H.D. performed mass spectrometric analyses of lncRNA-CSRIgA interacting proteins. B.L. analyzed the JH4 mutational analysis and executed the developmental studies. M.F. and L.H. provided data for the human homolog studies, and structural modeling thereof was contributed by Z.-P.L. Yiyun Chen and J.W. studied the interactions among multiple genomic regions affecting coregulatory relationships. U.B. participated in designing the research, interpreting the experiments, and authoring the manuscript. All authors contributed to writing the final manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All the detailed information of (i) each reagent used, (ii) all the pipelines and computational parameters used for analyses of high-throughput genomic datasets, (iii) the accession numbers and location of the raw sequence datasets themselves, (iv) the CRISPR-Cas9/Cas13 protocols used for genome or RNA editing, and (v) the results of the extraction of fecal RNA/DNA can be found in the Supplementary Materials.

Stay Connected to Science Immunology

Navigate This Article