Human CD14⁺ monocytes were purified from buffy coats of healthy donors using the anti-CD14 microbeads (Miltenyi Biotec), on the autoMACs Pro Separator (Miltenyi Biotec) as described in Supplementary Materials. In selected experiments 8 × 10⁶ cells were transfected with 200 pmol NRIR-specific Silencer Select siRNA (Mariotti et al., 2019) or Silencer Select negative control #2 (Ambion, Thermo Scientific), using the Human Monocyte Nucleofector Kit and the AMAXA Nucleofector II (Lonza), according to the manufacturer’s protocol.

LPS-stimulated monocytes were resuspended for 10 min on ice in RLN1 buffer (50 mM Tris-HCl pH 8, 140 mM NaCl, 1.5 mM MgCl₂, 0.5% NP-40) supplemented with 1 U/μl RNase Out (Invitrogen). After centrifugation at 300 g for 2 min, the supernatant was collected as the cytoplasmic fraction. The pellet was resuspended for 10 min on ice in RLN2 buffer (50 mM Tris-HCl, pH 8, 500 mM NaCl, 1.5 mM MgCl₂, and 0.5% NP-40) supplemented with 1 U/μl RNase Out. Chromatin was pelleted at 16,100 g for 3 min. The supernatant represented the nuclear fraction. All fractions were resuspended in RLT buffer, and RNA was purified as described below.

Total RNA was purified with the RNeasy Mini Kit (Qiagen). Gene expression was analyzed as reported in Supplementary Material. The primers used in this study are listed in Supplementary Table S1. Expression data were reported as MNE (Muller et al., 2002) after normalization over ACTIN B.

Identification of putative NRIR-binding proteins was performed using catRAPID Omics (parameters used: Homo sapiens library, RNA- and DNA-binding protein, full-length proteins (Livi et al., 2016), and the lncPRO algorithm (Lu et al., 2013) to identify interaction with the entire human proteome as reported in UniProt (https://www.uniprot.org/). Putative NRIR-binding proteins predicted by both algorithms were scored according to the lncPRO score, and the average calculated from the interaction score and the discriminative power obtained from catRAPID. Proteins common to lncPRO and catRAPID with a score higher than the average values were considered most probable NRIR-binding proteins. PANTHER protein families associated with the most probable NRIR-binding proteins were identified using the Gene List Analysis tool available at http://pantherdb.org/(Mi et al., 2013, 2019).

The interaction between NRIR and DNA was analyzed with LongTarget (He et al., 2015). DNA sequences of the promoter region (5 Kb upstream of the transcriptional start site) of CXCL10, CXCL11, DDX58, ESPTI1, IFI44, IFIT2, and MX1 were recovered from Ensembl (https://www.ensembl.org/). The interaction between NRIR and each promoter was analyzed with LongTarget using default parameters. Identification of enriched motifs either in Triplex Forming Oligonucleotide (TFO) or Triplex Target Site (TTS) was performed using the Multiple Em for Motif Enrichment (MEME) tool (Bailey and Elkan, 1994). Only motifs enriched with an E-value<0.05 were further investigated. The similarity between identified motifs was evaluated with the TomTom motif comparison tool with default parameters (Gupta et al., 2007). The presence of putative NRIR-binding sequences in the promoters of CXCL10, CXCL11, DDX58, ESPTI1, IFI44, IFIT2, and MX1 was assessed by using the Motif Alignment and Search Tool (MAST) from the MEME suite (Bailey et al., 2015). Enrichment analysis of putative NRIR-binding motifs in the promoters of NRIR-target genes was performed using the Simple Enrichment Analysis (SEA) tool available in the MEME suite (Bailey et al., 2015). Promoter regions of the genes not affected by the NRIR target (Mariotti et al., 2019) were used as background.

NRIR structure was predicted using the RNAfold tool of the ViennaRNA suite (Gruber et al., 2008) with the following parameters: p -d2 –noLP. The results of the minimum free energy-based prediction were used in this study. For selected proteins, the putative binding region in the NRIR sequence was identified using CatRAPID Fragment with default parameters (Livi et al., 2016).

Transcription factors (TFs) involved in the transcription of NRIR-target genes were obtained from the ENCODE project database (Dunham et al., 2012; Davis et al., 2018), as well as from published data. In order to identify the TFs involved in the transcription of NRIR-target genes in an unsupervised manner, a score was assigned to each TF according to the following rules:

• No information about the TF for the specific promoter: score 0

The matrix obtained by applying these scoring rules was subjected to hierarchical clustering using the hclust function in R, with column_split = 4.

Identification of TFBSs enriched in NRIR-target genes was performed using PSCAN (Zambelli et al., 2009). The TF motifs available in Transfac were used in the analysis, and the promoters of the 47 genes unaffected by NRIR-silencing were used as background (Mariotti et al., 2019). For subsequent analysis, only TFBSs significantly (p-value < 0.05) enriched in the NRIR target were considered.

Nuclear RNA ImmunoPrecipitation (nRIP) was performed according to Zhao, (2015), with minor modifications. For each immunoprecipitation, 10⁷ CD14⁺ monocytes were treated with 100 ng/ml LPS for 4 h and harvested in ice-cold PBS prior to cell lysis. The protocol for cell lysis and nRIP analysis is detailed in Supplementary Material. Data are expressed as a percentage of the non-immunoprecipitated RNA (% of input) (Castellucci et al., 2015).

Data are expressed as mean ± SEM, unless otherwise indicated. Statistical evaluation was determined using two-way analysis of variance (ANOVA), followed by Bonferroni’s multiple comparisons test. p-value<0.05 was considered significant. GraphPad 8.0 was used.

Additional details are described in Supplementary Materials and Methods.

Recently, we have demonstrated that NRIR plays a role in the induction of the expression of 15 out of 56 interferon-stimulated genes (ISGs) in human monocytes activated by LPS (Mariotti et al., 2019). In order to determine whether the upregulation of LPS-induced ISGs mediated by NRIR occurs at the transcriptional or post-transcriptional level, the expression of the primary transcript (PT) of the fifteen NRIR-target genes was analyzed by RT-qPCR using specific primer pairs in the same NRIR-silenced monocytes used in our previous work (Mariotti et al., 2019) (Supplementary Figure S1A). Under these conditions, the induction of 12 ISG transcriptions by LPS was impaired as compared to cells transfected with a scramble siRNA (Figure 1A-L). Specifically, a significant decrease of LPS-induced PTs for CXCL10 (-66.92%±11.05%, Figure 1A), DDX58 (-39.68%±9.88%, Figure 1B), MX1 (-41.96%±13.16%, Figure 1C), CXCL11 (-66.06%±9.04%, Figure 1D), EPSTI1 (-37.81%±11.47%, Figure 1E), IFIT2 (-57.10%±10.94%, Figure 1F), and IFI44 (-45.75%±11.59%, Figure 1G), was detected. Consistently, even though not statistically significant, reduction of CCL8 (-43.83%±15.74%, Figure 1H), APOBEC3A (-51.78%±12.51%, Figure 1I), USP18 (-60.86%±3.78%, Figure 1J), IFIH1 (-37.21%±12.54%, Figure 1K), and OAS2 (-50.49%±7.32%, Figure 1L) PTs was also observed 4 h after LPS stimulation in NRIR-silenced monocytes. Transcription of OAS3, IFITM3, and ISG15, previously shown as NRIR-target genes (Mariotti et al., 2019), could not be analyzed for technical reasons (see Supplementary Material and methods). Transcriptional upregulation of additional LPS-induced ISGs, namely, IRF7, IFIT1, and OASL, comprising the 41 protein-coding genes previously identified as induced by LPS in an NRIR-independent manner (Mariotti et al., 2019), was unaltered in NRIR-silenced monocytes (Supplementary Figure 1B–F), thus suggesting that NRIR is required for transcription of a subset of LPS-induced ISGs.

To support the role of NRIR in the upregulation of LPS-induced ISG transcription, localization of NRIR was investigated, by RT-qPCR, in cytoplasmic, nuclear, and chromatin-associated fractions obtained from LPS-stimulated human monocytes (Figure 1M). Data showed that NRIR is predominantly (87.64%±1.95%) localized in monocyte nuclei, with 32.72%±2.76% of it associated with chromatin. The purity of cellular fractions was demonstrated by the distribution of IL-6 mRNA and IL-6 PT, which were properly enriched in monocyte cytoplasm and nucleoplasm, respectively (Figure 1M).

LncRNAs can regulate transcription by a variety of mechanisms as they engage with proteins such as transcription factors, histone deacetylases, methyl-transferases, or chromatin remodeling complexes (Akkipeddi et al., 2020). Given the capability of NRIR to regulate transcription of its target genes, transcription factors (TFs) were among the expected NRIR-binding proteins. To support this hypothesis, in silico analysis of putative NRIR-binding proteins was performed by catRAPID and lncPRO. This analysis returned a list of 2,596 proteins predicted as putative NRIR-interacting proteins that were plotted as a function of their prediction scores (Figure 1N). The 330 proteins with a score higher than the average score for both the algorithms (Figure 1N, blue dots) were then subjected to functional classification using PANTHER (Supplementary Table S2). Remarkably, 63% of the proteins (n=139) were classified as involved in transcriptional regulation, and 37% (n=82) were functionally grouped as associated with post-transcriptional regulation (Figure 1O). Notably, 87% (n=121) of the proteins classified as a regulator of transcription specifically belong to the transcription factor class, thus strongly hinting to transcription factors as the most likely NRIR-interacting proteins (Figure 1O).

At the transcriptional level, a model is emerging whereby an lncRNA bridges DNA and protein by binding to chromatin and working as a scaffold for transcription factors and/or modifying protein complexes (Singh and Prasanth, 2013). LncRNA:protein and lncRNA:DNA interactions are mediated by interacting elements located in specific structural domains (Wang and Chang, 2011). NRIR secondary structure was investigated using RNAfold with the minimum free energy (MFE) and partition function. NRIR was predicted as a highly structured lncRNA composed of four main domains (D1 to D4) starting from a central four-way junction (Figure 2 and Supplementary Table S3). D1 and D2 appeared as two complex domains, composed of a total of three internal junctions: one five-way in D1, one five-way and one four-way junction in D2; nine terminal loops, 13 internal loops, and 13 helices (Figure 2). Differently, D3 and D4 were characterized by a lower degree of complexity than D1 and D2. D3 consisted of one helix, one external loop, and two internal loops (Figure 2). D4 was composed of two three-way internal junctions, three terminal loops, five internal loops, and five helices. Overall, the presence of highly complex domains (D1 and D2) and fewer complex domains (D3 and D4) suggested the co-existence of DNA-binding and protein-binding domains in NRIR.

To fulfill its role as a transcriptional regulator, an lncRNA must contain a DNA-binding domain that mediates the formation of RNA:DNA triplex structures. The identification of putative DNA-binding motifs in NRIR and putative NRIR-binding sites in the promoters of the seven genes significantly modulated by NRIR (i.e. CXCL10, DDX58, MX1, CXCL11, EPSTI1, IFIT2, and IFI44, Figures 1A–G) was performed in silico by the use of LongTarget (He et al., 2015). LongTarget analysis returned a number of putative NRIR–DNA interactions that were grouped into triplex-forming oligonucleotides (TFO) identified in the NRIR sequence. TFO1, the best TFO of all the TFOs that were generated, was then identified. In parallel, LongTarget provided a list of triplex target sites (TTSs), complementary to the identified TFO, that represent the putative NRIR-binding sequences in the genomic regions analyzed (He et al., 2015). Analysis of promoter sequences of the selected NRIR-target genes led to identification of 220 NRIR: DNA interactions associated with TFO1 (Supplementary Table S4). All 220 sequences were subjected to a motif enrichment analysis by MEME in order to identify a shared motif. Three major motifs (herein called NRIR DNA-binding motifs, DBM) were identified (Supplementary Table S5), among which DBM-1 showed the best features in terms of enrichment (E-value=7.0×e^-3316) and length (31 nts) (Figure 3A and Supplementary Table S5). Noticeably, DBM-2 and DBM-3 displayed a significant (p-value=8.98×e^-5 and 6.71×e^-2, respectively) homology with DBM-1 as evaluated by TomTom (Supplementary Table S5), suggesting that these two motifs may be part of the DBM-1 rather than distinct DBMs. Remarkably, and consistent with NRIR secondary structure, the DBM-1 is localized in the first helix of D4 (Figure 2) at position 45–75 nt (Supplementary Table S3), thus strengthening the likelihood that D4 represents the putative DNA-binding domain of NRIR.

The putative motifs recognized by DBM1 in the promoters of CXCL10, CXCL11, DDX58, ESPTI1, IFI44, IFIT2, and MX1 were identified by analyzing all the TTS complementary to TFO1. Motif enrichment analysis of TTS led to the identification of 18 significantly enriched motifs (E-value<0.05, Figure 3B), which from now on are called NRIR-binding sites (NBS). The presence of the 18 NBS in the promoter region of CXCL10, CXCL11, DDX58, ESPTI1, IFI44, IFIT2, and MX1 was assessed by in silico analysis using MAST. All the promoters contain at least 11 NBS. The number of sites for each NBS is shown in Supplementary Figure S2. In order to identify the NBS more likely involved in the capability of NRIR to specifically regulate the subset of LPS-induced ISGs, a relative motif enrichment analysis was performed using SEA. SEA identifies motifs that are relatively enriched in a set of input sequences, herein consisting of the promoter sequences of all the 15 NRIR-target genes, as compared to user-provided control sequences, herein represented by the promoter sequences of the 41 LPS-induced NRIR-independent genes (Mariotti et al., 2019). As a result, five NBS, namely NBS-3, NBS-4, NBS-9, NBS-12, and NBS-16, were significantly enriched in NRIR-target gene promoters (Figure 3C). Collectively, the identification of the DNA-binding domain 1 in NRIR and of five NRIR-binding sites in the promoter of NRIR-target genes strongly supported the role of NRIR as a transcriptional regulator of its target genes.

In order to identify TFs interacting with NRIR and concurrently involved in the regulation of NRIR-target ISGs, we first assessed the transcription factor-binding sites (TFBS) significantly over-represented in the promoters of the 15 NRIR-target genes. Fifty-nine TFBSs were found significantly enriched in NRIR-target gene promoters as compared to non-targeted gene promoters by PSCAN (Supplementary Table S6). In parallel, a search of the literature and of the ENCODE project database for proteins demonstrated to promote transcription of the 15 NRIR-target genes yielded a list of 158 TFs. Each of these 158 TFs was assigned a score, according to the criteria described in Material and methods. Hierarchical clustering analysis followed by a k-mean clustering uncovered a cluster of 34 TFs involved in the regulation of the majority of the NRIR-target genes (cluster 4, Figure 3D). The intersection of the list of 34 TFs with the list of the 59 TFBS enriched in NRIR-target genes returned eight TFs, namely, SP1, MYC, MAX, STAT1, TBP, STAT2, YY1, and IRF1 (Supplementary Figure S3), that were shared by the majority of NRIR-target genes and simultaneously were able to bind to the TFBS enriched in the NRIR-target gene promoters. Collectively, these data suggested a mechanism through which NRIR binding to the TFs selected according to the strategy described above promotes transcription of the selected NRIR-target genes.

The interaction profile of NRIR with each of the eight TFs identified was performed by catRAPID fragment. This analysis identified a principal protein binding element (PBE-1) spanning 300–480 nt (Supplementary Figure S4 and Supplementary Table S7), shared by all the TFs and overlapping the D2 and D3 domains in the NRIR structure (Figure 2). Interaction profiles of NRIR and MYC, STAT1, or SP1 detected a second protein binding element (PBE-2), spanning 660–730 nt (Supplementary Figure S4 and Supplementary Table S7), located in the D1 domain (Figure 2).

Among the eight TFs selected as described above, STAT1 and STAT2 have been previously demonstrated to be constitutively upregulated in monocytes from Systemic Sclerosis patients and to correlate with the expression of STAT-target genes (Van Der Kroef et al., 2019). In the same in vivo setting, NRIR has also been demonstrated to be constitutively upregulated, and to account for the type I IFN signature that is a hallmark of this pathology (Mariotti et al., 2019). Based on these experimental data, the existence of a causal link between STAT1/2 and NRIR was investigated by nuclear-RIP assay (nRIP). Freshly purified monocytes were treated with LPS for 4 h, and nuclear lysates were incubated with αSTAT1 or αSTAT2 antibodies or with IgG as a control. The presence of NRIR was assessed by RT-qPCR in αSTAT1, αSTAT2, and IgG immune precipitates (IPs). In parallel, in order to validate the nRIP assay, nuclear lysates were incubated with αp50 NF-κB antibodies, and the presence of PACER, a lncRNA already demonstrated to bind p50 NF-κB (Krawczyk and Emerson, 2014), was assessed by RT-qPCR. PACER was detected exclusively in αp50 NF-κB IPs, as it was undetectable in αSTAT1, αSTAT2, and IgG IPs. Remarkably, significant recruitment of NRIR was observed in both STAT1 and STAT2 IPs, but not in the control IgG IP nor in the αp50 NF-κB IP (Figure 3E), thus providing experimental validation of the predicted NRIR:STAT interactions.