DUSP11 suppresses nc886 expression

Since DUSP11 plays a role in 5′-maturation of Pol III-ncRNAs, we began our investigation by determining DUSP11’s effect on nc886 expression. First, we examined in vivo expression levels of DUSP11 and nc886. We had high-throughput sequencing of total RNA (RNA-seq) data for 36 pairs of a tumor tissue (“T”) and its adjacent non-tumor tissue (“N”) collected from head-neck squamous cell carcinoma (HNSCC) patients. DUSP11 expression levels were assessed from these data, and nc886 expression was measured by qRT-PCR. Another dataset was from our previous studies in which Illumina array and nc886 qRT-PCR had been conducted in T samples from 108 esophageal squamous cell carcinoma (ESCC) patients [17, 18]. In both cohorts, DUSP11 expression values distributed across a range of approximately eight-fold (see the y-axis of Fig. 1A, B), indicating variability in DUSP11 expression. Notably, DUSP11 expression levels were negatively correlated with those of nc886 (Fig. 1A, B).

Fig. 1

DUSP11 is variably expressed and suppresses nc886 expression. A, B Scatter plots showing correlation between DUSP11 and nc886 expression levels in clinical samples. T and N: a tumor sample and its adjacent non-tumor tissue respectively. In panel A, DUSP11 levels are TPM values from RNA-seq and nc886 levels are 2−ΔΔCt values from qRT-PCR. In panel B. nc886 levels normalized first to 18S rRNA and then to the average of all T samples (–ΔΔCt in the x-axis). C Northern and Western blot of indicated genes. nc886, vtRNA1-1, and 5S rRNA were probed sequentially on the same blot. 5S rRNA and β-actin: loading controls. All samples were harvested at 48 h after transfection of siDUSP11 or siControl. For Western blot, molecular size markers are indicated on the right. D Quantitation of the Northern signal in panel C. Each band was quantitated by Image J. nc886 and vtRNA1-1 values were normalized to 5S rRNA, log2-transformed, and plotted on a heat map. E A diagram illustrating steps of organoid culture. qRT-PCR of indicated genes is shown below each step. All samples were harvested at 48 h after siRNA transfection

These in vivo data, coupled with DUSP11’s role in the 5′-maturation of Pol III-ncRNAs, suggested that DUSP11 suppressed nc886 expression. To obtain experimental evidence, we transfected a small interfering RNA (siRNA) targeting DUSP11 (“siDUSP11”) into several cell lines. Knockdown (KD) was efficient, as shown by the almost complete disappearance of DUSP11 protein with siDUSP11 compared to a negative control siRNA (“siControl”) (Fig. 1C). nc886 expression increased upon DUSP11 KD in all the cell lines examined here (Fig. 1C, D). They varied in tissue origin and included different cell types (epithelial cells, fibroblasts, and myofibroblasts), suggesting that the suppression of nc886 by DUSP11 is a ubiquitous phenomenon. Consistent with previous reports [8, 10, 11], the expression of a canonical vtRNA, vtRNA1-1, also increased upon DUSP11 KD in most of the cell lines. These cell culture data were corroborated by an organoid experiment. We delivered siDUSP11 to organoids and observed elevated nc886 in hepatocyte organoids as well as in hepatic progenitor cells (Figs. 1E and S1).

The increase of nc886 upon DUSP11 KD is attributed to its enhanced RNA stability

Our in vivo, in vitro and organoid data unequivocally proved that DUSP11 suppresses nc886 expression. Next, we interrogated which step of nc886 expression is affected by DUSP11. The steady-state expression level of nc886 is determined by regulation at multiple levels, including epigenetic mechanisms, transcription factors (TFs), and RNA stability [reviewed in [16]].

Since DUSP11 is an enzyme acting on RNA [6, 7], we first interrogated whether DUSP11 affects the stability of nc886, by measuring nc886 after shutdown of transcription by treating cells with actinomycin D (ActD) (Figs. 2A, B and S2). The half-life of nc886 was calculated to be 77 min when DUSP11 was present, but was substantially extended upon DUSP11 KD (270 min in Fig. 2B). For comparison, the half-life of vtRNA1-1 was 55 min, and it also extended to 142 min by DUSP11 KD. These data indicated that DUSP11 destabilizes nc886, thereby lowering its steady-state expression level.

Fig. 2

DUSP11 reduces nc886 stability without altering its 5′-structure. A, B ActD was treated at 48 h after transfecting siRNAs. Northern hybridization of nc886, vtRNA1-1, and 5S rRNA (A). A graph showing nc886 and vtRNA signals quantitated by Image J (B). The values of siControl and siDUSP11 at 0 h were set as being one. The dashed red horizontal line indicates 0.5 of y-axis. Intersection points of this line to each graph are designated by arrows with half-life for nc886 and by arrowheads for vtRNA1-1. C Northern hybridization and Western blot of indicated genes. All other descriptions are the same as Fig. 1C. D A diagram illustrating nc886-expressing constructs. E Northern hybridization of nc886 at 24 h after transfection of nc886 plasmids (illustrated in D) into 293 T cells. EtBr staining is shown as a loading control. F A box plot depicting fc of nc886 and snU6 expression upon DUSP11 KD. “endo nc886”, Northern quantitation of seven cell lines in Fig. 1C; “U6 nc886”, Northern quantitation of three cell lines in C; “snU6 RNA”, snU6 expression levels of cell lines in C measured by qRT-PCR. G A diagram illustrating the substrate specificity of Terminator exonuclease. H, I Northern hybridization of indicate RNAs after Terminator assays. In panel H, “synthetic 5′-P RNA” is a spike-in RNA that was designed to be detected by the vtRNA1-1 probe (see Materials and Methods). J A graph showing the proportion of 5′-PPP RNA. Each band in panels H-I was quantitated by Image J. A total RNA amount (5′-P and 5′-PPP) and a 5′-PPP RNA amount were estimated from the band intensity of the -Terminator and + Terminator reaction respectively, to calculate the proportion of 5′-PPP RNA (y-axis). An average of three cell lines is shown, with SD and p-value. “n.s.”, not significant. K. Northern hybridization of indicated RNAs after Terminator assays. A cartoon illustrating nc886 “wild-type” and “mut_46-56” with respect to PKR binding and inhibition (leftmost panel). A 10% native acrylamide gel showing these two RNAs and vtRNA1-1, all produced by in vitro transcription. Two nc886 conformers are indicated by arrows (2nd left panel). Northern hybridization with the indicated probes and quantification (right two panels). 293 T cells transfected with each indicated RNA together with “control-RNA” were harvested at 4 h to isolate total RNA for Terminator assays (see Materials and Methods). nc886 bands were quantitated by Image J and displayed on a bar graph

A CpG island is present in the nc886 promoter region and hypermethylation of this region results in heterochromatin formation, leading to the complete silencing of nc886 expression. This epigenetic silencing occurs in a significant fraction of cell lines, including 293 T. In 293 T cells, nc886 expression remained silenced even after DUSP11 KD (Fig. 2C). In contrast, an increase in vtRNA1-1 expression was observed in the same batch of DUSP11 KD, confirming the KD efficacy. Not only 293 T and its derivative 293FT cell lines, nc886 is silenced also in the Hep3B. From each of them, we made a derivative cell line stably expressing nc886. When constructing an nc886-expressing plasmid, we inserted a 101 nucleotide (nt)-long DNA segment corresponding to the transcribed region. Among Pol III-ncRNAs, nc886 has a gene-internal type 2 promoter whose cis-elements (A and B boxes) exist within the transcribed sequence (Fig. 2D) [16, 19]. Despite existence of the A and B boxes, the 101-nt DNA could not drive nc886 expression when cloned into a promoter-less vector. Instead, nc886 was well expressed when cloned downstream of a heterologous promoter U6, which is a gene-external type 3 promoter for small nuclear RNA U6 (snU6) (Fig. 2D, E). Thus, nc886 expression was entirely driven by the U6 promoter in derivative cell lines from 293 T, 293FT, and Hep3B (designated 293 T-U6:nc886, 293FT-U6:nc886, and Hep3B-U6:nc886 respectively). In all of them, nc886 expression was increased upon DUSP11 KD (Fig. 2C and F). In contrast, the level of snU6, whose expression was naturally driven from its own U6 promoter, was barely affected by DUSP11 KD (Fig. 2F). All the results from nc886-silenced cells and their derivatives (Fig. 2C–F) accorded with our earlier (Fig. 2A, B) data, demonstrating that nc886’ stability is the sole step affected by DUSP11.

nc886 is not a substrate for DUSP11’s 5′-phosphatase activity

Since DUSP11 is an RNA 5′-phosphatase, we next examined nc886’s 5′-end structure. In the case of vtRNAs and hepatitis C virus (HCV) RNA, it has been reported that their expression is low in the presence of DUSP11 because DUSP11 converts them from a 5′-PPP-form to a 5′-P form that is more susceptible to degradation by 5′-exonucleases such as Xrn1 or Xrn2 [8, 20, 21].

So, we assessed the 5′-phosphorylation status of nc886 by using an experiment with an enzyme called Terminator, a 5′ to 3′ exonuclease (“Terminator assay”). Terminator degrades 5′-P-RNA but not 5′-PPP-RNA (Fig. 2G). In “siControl” samples of 293 T-U6:nc886 cells (Fig. 2H, lane 1–2), Terminator degraded nc886 as efficiently as vtRNA1-1 and a synthetic 5′-P RNA, indicating that nc886 was of 5′-P in the presence of DUSP11. Importantly, DUSP11 KD did not affect the sensitivity of nc886 to Terminator, indicating that nc886 was remained 5′-P even in the absence of DUSP11 (Fig. 2H, lane 3–4). DUSP11 KD resulted in the accumulation of the 5′-PPP form of vtRNA1-1, as shown by its resistance to Terminator in the same batch Terminator reaction and on the same blot (Fig. 2H, lane 3–4). In the case of vtRNA1-1, its 5′-PPP form has been reported to accumulate in the absence of DUSP11 in several publications [8, 10, 11].

Since we obtained an unexpected result in 293 T-U6:nc886 cells and these cells express nc886 from a heterologous promoter, we wanted to confirm this in cells that naturally express nc886. To this end, we performed the same experiments in TE-8 and WPMY-1, and the result was essentially the same (Fig. 2I, J). nc886 was sensitive to Terminator both regardless of siControl or siDUSP11. In contrast, vtRNA1-1 was sensitive in siControl, but became resistant upon DUSP11 KD.

Taken together, our data clearly demonstrated that nc886 is not a substrate for the phosphatase activity of DUSP11. This result was contrary to our initial expectation because the known substrates of DUSP11 are Pol III-ncRNAs, including vtRNA1-1, the closest paralog of nc886. We next sought to elucidate the reason for this unusual property of nc886. According to previous results from our laboratory and the Conn laboratory [12, 13], nc886 adopts two conformations, each with distinct properties in PKR binding and activation (left panel, Fig. 2K). When in vitro transcribed nc886 was run on a native gel, these two conformers were resolved (2nd left panel, Fig. 2K). A slowly migrating conformer (“conformer-1”) is a PKR-binding form, whereas a fast migrating conformer (“conformer-2”) is not [12]. Accordingly, nc886 conformer-1 was only seen in wild-type nc886, but not in a mutant nc886 (nc886_46-56) defective in PKR binding. Compared to vtRNA1-1 (98 nt long), which is close in size to nc886 (101 nt), nc886 conformer-2 migrated similarly, but conformer-1 migrated much slower, indicating a highly structured nature of conformer-1.

We thought that such a structure might be the reason why nc886 is a poor substrate for DUSP11. To test this idea, we transfected nc886 wild-type RNA and the nc886_46-56 mutant RNA into DUSP11-expressing 293 T cells and examined the 5′-phosphorylation status (right two panels in Figs. 2K and S3). A significant fraction of the nc886_46-56 mutant RNA, as well as another in vitro transcribed RNA included as a control for transfection efficiency and equal loading, was converted to 5′-P as indicated by degradation by Terminator. In comparison, the wild-type nc886 RNA largely remained as the Terminator-resistant 5′-PPP form. Since the majority of the wild-type and the mutant nc886 was in conformers -1 and -2 respectively, these data suggest that DUSP11 poorly recognizes the secondary structure of nc886, which is distinct from other RNAs.

In summary, DUSP11 represses most Pol III-ncRNAs probably by destabilizing them. It appears that DUSP11 does this by 5′-dephosphorylation for vtRNA1-1 and most others but by a different mechanism for nc886.

DUSP11 KD sensitizes cells to higher IFN responses against PAMP, and this sensitization is mitigated by nc886

So far, we undoubtedly demonstrated that nc886 expression is increased without alteration of its 5′-end while some other Pol III-ncRNAs accumulated in a 5′-PPP-form, in the absence of DUSP11. Regarding this intriguing result, we sought to find biological significance. It has been documented that cells are immune-sensitized upon DUSP11 KD and elicit stronger immune responses against incoming pathogens [9,10,11, 22]. To determine nc886’s role in this context, we conducted combinatorial transfection of siDUSP11 and a synthetic PAMP-RNA (shortly “PAMP”) into nc886-expressing (“293 T-U6:nc886”) and control cell lines (“293 T-vector”) (Figs. 3 and S4; see also Fig. S5 for DUSP11 and nc886 levels in this experiment). In “293 T-vector” cells, PAMP induced the IFN response, which was assessed by IFN-β expression (bar 1–4 in Figs. 3A and S4). The fold-induction was higher in cells pre-treated with siDUSP11 than siControl (> threefold, compare bar 2 vs 4 in Figs. 3A and S4). DUSP11 KD alone, without PAMP, had a marginal impact in IFN-β induction (compare bar 1 vs 3 in Figs. 3A and S4). These data corroborated the aforementioned immune-sensitization upon DUSP11 KD [9, 11].

Fig. 3

The increased nc886 by DUSP11 KD suppresses innate immune responses. siRNAs were transfected at 0 h, PAMP (or vehicle control) was treated at 36 h, and cells were harvested at 48 h. A qRT-PCR of IFN-β. y-axis: 2−ΔΔCt value with the leftmost bar set to 1. –RT reactions were performed in parallel and subtracted from + RT values. The p-value in red letters represents the statistical significance of the siDUSP11/siControl values between the two PAMP-treated cell lines. More specifically, three siDUSP11/siControl values of 293 T-vector (the ratio of bar 4 to bar 2) and those of 293 T-U6:nc886 (the ratio of bar 8 to bar 6) were calculated and were subjected to Student's t-test. B A dot plot showing the relative TPM values of 42 ISGs that were upregulated by more than twofold in siDUSP11 with PAMP compared to siControl without PAMP in 293 T-vector cells. The TPM values of these ISGs were normalized to siControl in 293 T-vector without PAMP as shown in Fig. 3A. The p-value in red letters represents the statistical significance of the difference between the two cell lines and was calculated as follows: siDUSP11/siControl values for these 42 ISGs were calculated and the 42 ratio values of 293 T-vector were analyzed against 42 values of 293 T-U6:nc886 using Student's t-test. C A heat map depicting TF activity from MSigDB analysis. Z-scores for each of four samples indicate the change of a TF activity upon PAMP-treatment. Among 1137 TFs, those of Z-score >  + 4 in 293 T-vector cells with siDUSP11 (= column 2) were selected and displayed. D A heat map of Z-scores for KEGG pathways (selected from total 186). All other descriptions are the same as panel C, except for the Z-score cut-off (> + 3 here) and the color scale (see scale bar below). E Northern hybridization and Western blot of indicated genes. All descriptions are the same as Fig. 1C, except that EtBr staining is shown for equal loading. F A graph showing the expression of PAMP-triggered IFN-β and ISGs calculated from 2−ΔΔCt values of qRT-PCR. The fold-induction values of a siDUSP11 sample relative to the corresponding siControl sample were calculated (y-axis). siRNAs were transfected at 0 h, PAMP (or vehicle control) was treated at 40 h, and cells were harvested at 48 h. G Western blot of DUSP11 (top panel), quantification (bottom panel), and qRT-PCR of nc886 (bottom panel), after KSHV infection. At 24 h post-infection, > 90% of the cells were infected. In the lower graph, nc886 expression is 2−ΔΔCt values (left y-axis), with the value from Huh7 cells with mock infection at 6 h set to 1. DUSP11 and β-actin bands in the top Western blot were quantified using Image J, to calculate DUSP11/ β-actin, which is a normalized DUSP11 expression value. The ratio of normalized values between KSHV and mock infection (right y-axis) in Huh-7 cells was shown at each time point, with the ratio at 6 h set to 1. H nc886 expression, as measured by qRT-PCR, at day 1, 3, and 5 after infection of rKSHV.219 at the indicated multiplicity of infection (MOI). LEC are KSHV-permissive (upper panel); BEC are non-permissive and were included as a negative control (lower panel). I KSHV infectivity at each indicated MOI (x-axis) at 24 h post-infection, as measured by GFP expression using flow cytometry. J A dot plot showing the expression of 81 ISGs (that were upregulated in DUSP11-low patients) between nc886-low and nc886-high HNSCC patients. Each dot indicates an average of TPM values (in three nc886-low patients or in seven nc886-high patients) that were calculated from RNA-seq data

In nc886-expressing cells (“293 T-U6:nc886”), PAMP also induced the IFN response, but less robustly than control cells (compare bar 2 vs 6 in Figs. 3A and S4), assuring nc886’s suppressive role in the IFN response [14]. Most importantly, a siDUSP11/siControl ratio of “293 T-U6:nc886” cells was < twofold, which is significantly lower than that (> threefold) of “293 T-vector” cells. This difference indicated the contributing portion of “the increased 5′-P-nc886” by DUSP11 KD to the mitigating effect on IFN.

The IFN response leads to transcriptional changes in a set of genes, called ISGs. To comprehensively investigate ISGs, we conducted RNA-seq on the eight samples in Fig. 3A. There are 376 ISGs that were reported to be changed by infection of various pathogens [23], and we examined them in our RNA-seq data (Table S1). Among them, 42 genes were captured (Transcripts Per Million (TPM) > 1) in all the eight samples and were > twofold induced by PAMP and siDUSP11 in “293 T-vector”. We calculated fold-change (fc) values (PAMP-treated / untreated) of these 42 genes and displayed them on a dot plot and a heat map (Figs. 3B and S6, Table S1). siDUSP11-treated cells exhibited a higher induction of ISGs by PAMP than siControl (2.31 fold when comparing dataset 1 vs 2 of “293 T-vector”). Importantly, this stimulatory effect of DUSP11 KD on ISG induction was significantly attenuated in “293 T-U6:nc886” cells (1.54 fold in dataset 3 vs 4). This result, in addition to the IFN-β data, supports the immune-suppressive role of “the increased 5′-P-nc886” by DUSP11 KD.

We analyzed the RNA-seq data in the Molecular Signature Database (MSigDB; https://www.gsea-msigdb.org/gsea/msigdb) to identify which pathways or TF activities were activated or suppressed. In the MSigDB, there are a TF set (that contains 1137 TFs) and several pathway sets including the Kyoto Encyclopedia of Genes and Genomes (KEGG) set that contains 186 biological pathways (Table S2-S3). Z-scores of “PAMP-treated vs untreated” were calculated for each TF and pathway to indicate their activation or suppression upon PAMP treatment. We sorted 23 TFs and 17 KEGG pathways that were most activated by PAMP (Z-score >  + 4 and >  + 3, respectively) in siDUSP11-treated “293 T-vector” cells. Most of these TFs and pathways were implicated in immunity broadly (bold letters in Fig. 3C, D, Tables S2, S3), and among them, the majority were directly related to cellular innate immunity (bold red letters). A heat map of Z-scores for the four experimental sets consistently demonstrated nc886’s suppressive role in immune responses. Furthermore, the contrast between columns 1 vs 2 was more pronounced than 3 vs 4, clearly depicting how much “the increased 5′-P-nc886” by DUSP11 KD suppressed TFs or pathways that had been activated by PAMP. The same trend was observed in Z-score heat maps of the Biocarta and Reactome pathway sets (Figs. S7, S8, Tables S4, S5).

So far, our data on 293 T-derived cells have come from a gain-of-function experiment, which we wanted to complement with a loss-of-function experiment. Of note, it is very challenging to construct an nc886 knockout (KO) cell line from a parental cell line expressing nc886, because depletion of nc886 leads to activation of PKR and consequently apoptosis. Nevertheless, we obtained an nc886-null derivative cell line from a hepatoma cell line Huh7 that expresses nc886, which was named Huh7i (nc886-inactivated) [24]. The expression of nc886 was increased upon DUSP11 KD in Huh7 cells and remained unexpressed in Huh7i (Fig. 3E), consistent with our results in Fig. 2A–F that DUSP11 KD induced nc886 in nc886-expressing cells. Having confirmed the increase in nc886 in Huh7 cells, we evaluated its effect on synthetic PAMP-induced IFN responses by measuring the expression of IFN-β and selected ISGs. In Huh7i cells, siDUSP11 stimulated the induction of IFN-β and ISGs by 1.5–2.3 fold, as compared to siControl (see y-axis values in Fig. 3F). Importantly, this fold induction was decreased in Huh7 cells, proving the suppressive role of “the increased 5′-P-nc886” on IFN responses.

Collectively, our data elucidated nc886’s role in DUSP11-mediated innate immunity. DUSP11 depletion primes cells to provoke robust IFN responses against PAMP and that this effect is mitigated by “the increased 5′-P-nc886”, which is another consequence of the DUSP11 depletion.

nc886, controlled by DUSP11 upon viral infection, suppresses ISGs and promotes viral replication

Encouraged by the results of the synthetic PAMP experiments, we extended our investigation to viral infection. We chose KSHV, because a change in DUSP11 expression during KSHV infection has been reported [10]. In this work, DUSP11 expression showed two opposite directions of DUSP11; a slight increase at hourly time points after KSHV infection and a decrease at daily time points after lytic reactivation of KSHV. We performed KSHV infection in Huh7 and Huh7i cells, the only cell line pair available to date for the nc886 loss-of-function approach. Although they are not a natural host for KSHV, we infected KSHV and confirmed that a majority of the cells were infected similarly to both cell lines (data not shown). Upon KSHV infection, DUSP11 expression increased at 24 h and then decreased at 48 h in Huh7 cells (upper panel in Fig. 3G). In contrast, nc886 expression decreased at 24 h and then increased at 48 h (bottom panel in Fig. 3G), consistent with the suppression of nc886 by DUSP11 reproducibly shown in our previous data (Figs. 1, 2). nc886 remained silenced in Huh7i cells independent of the change in DUSP11 expression upon KSHV infection.

nc886, which increased at 48 h post-infection and was associated with decreased DUSP11, suppressed ISGs (Fig. S9). We hypothesized that the induction of nc886 and the consequent suppression of ISGs are required for KSHV to propagate. Therefore, we performed a prolonged infection of KSHV in natural host cells. Primary lymphatic endothelial cells (LEC) are the primary target host cells in vivo [25]. We infected KSHV into LEC, along with non-permissive blood endothelial cells (BEC) as a negative control. The expression of nc886 was increased at day 3 and 5 in LEC but not in BEC (Fig. 3H). Three vtRNAs were also increased specifically in LEC, with fc similar to nc886 (Fig. S10). To determine nc886’s role in KSHV, we infected “293 T-vector” and “293 T-U6:nc886” cells. KSHV infectivity was significantly higher in nc886-expressing cells (Fig. 3I).

In conclusion, our KSHV data support the suppression of nc886 by DUSP11. The interplay between virus, DUSP11 and nc886 appears to be very complex, as shown by the opposing trends between time points. Although the significance of the early event (increased DUSP11 and decreased nc886) remains to be elucidated, we were able to illustrate the importance of the DUSP11/nc886 interplay when KSHV survives the host immune response and replicates successfully. In this situation, it appears that KSHV suppresses DUSP11 expression (Fig. 3G) and induces nc886 expression (Fig. 3G, H) to use nc886 to inhibit the innate immune response (Fig. S9) and promote viral replication (Fig. 3I).

nc886 and DUSP11 are associated with the expression of ISGs in HNSCC patients

Inflammation, shaped by the innate immune response, is associated with cancer [26] and viral infection is an etiological factor in some cancers, including HNSCC [27]. We asked whether ISGs are also controlled by the DUSP11-nc886 interplay in cancer. To this end, we examined gene expression data in HNSCC patients (n = 36). Among them, ten patients showed decreased expression of DUSP11 in T relative to N (“DUSP11-low patients”) and are therefore expected to be immune-sensitized. Among these ten patients, those with nc886 increase (n = 7) outnumbered those with decreased nc886 (n = 3). We hypothesized that immune sensitization by low DUSP11 expression would be alleviated by nc886 in the seven patients, but not in the remaining three. From a total of 376 ISGs, we made a short list of 81 ISGs that were captured in RNA-seq and were increased in the DUSP11-low patients (Table S6). For each of these 81 ISGs, we calculated two averages; one from the three patients (“nc886-low”) and the other from the seven patients (“nc886-high”). When compared in a dot plot, nc886-high patients tended to have lower expression of those ISGs that were induced by DUSP11 reduction (Fig. 3J and S11, Table S6). This trend supported our idea that nc886 suppresses the sensitized immunity induced by DUSP11 reduction. The result from HNSCC patients also suggested that the interplay of DUSP11-nc886 might be involved in the regulation of ISGs during tumorigenesis, in addition to PAMP treatment and viral infection.

DUSP11 is variably expressed in humans as compared to mice, a species that lacks nc886

Our study has been based on experimental depletion of DUSP11 to evaluate nc886’s role. To understand the significance of our KD data in natural contexts, we investigated DUSP11 expression and regulation. In the vicinity of the transcription start site (+ 1 in Fig. 4A), there are several features including a promoter element, enhancer elements, a CpG island, and a peak of an open chromatin mark (H3K27Ac) (Fig. 4A). These features are consistent with DUSP11 being a housekeeping gene, as inferred from its role in the maturation of Pol III-ncRNAs. Nevertheless, we observed a high peak of ReMap density at the + 1 site (Fig. 4A-B). The ReMap density is a simplified illustration indicating the degree of TF binding. ReMap (https://remap.univ-amu.fr) has compiled Chromatin ImmunoPrecipitation followed by high-throughput sequencing (ChIP-seq) data from the Encyclopedia of DNA Elements (ENCODE) and Gene Expression Omnibus (GEO) databases. This feature, representing variable expression of DUSP11, suggested that there are DUSP11-low situations in which the innate immune phenotypes upon DUSP11 KD can occur naturally.

Fig. 4

DUSP11 is variably expressed in humans. A A screenshot of a DUSP11 region at the University of California, Santa Cruz (UCSC) genome browser, showing indicated tracks (information about its reference sequence, ReMap density, promoter, enhancers, and an epigenetic mark) in a 8 kb region flanking the transcription start site (designated + 1, vertical dotted line). B Screenshots of the USCS genome browser showing ReMap density at the genomic region (from upstream 1 kb to downstream + 1 kb of the transcription start site) of indicated genes. Brown arrow with a vertical line, + 1 site and the transcription direction; red arrow with number, peak position and height. C A rank plot depicting SD/average values of 12,449 human genes (see the main text for detailed information). DUSP11 (brown color) and other genes that remove 5′-PPP (blue color) are designated by arrows and vertical lines. D A rank plot depicting SD/average values of 10,902 mouse genes. All other descriptions are the same as C

Thus, we wanted to determine how much DUSP11 levels actually vary and whether its variability implicates nc886. Since we were focusing on innate immunity that occurs in most normal cells, we looked into the Genotype-Tissue Expression portal (GTEx; https://gtexportal.org/home/) that has compiled RNA-seq data of 8371 normal human tissues. From a total of 56,200 genes, we shortlisted 12,449 genes by eliminating scarcely expressed ones according to an average TPM of the 8371 samples (cut-off = 5). We estimated the expression variability of these 12,449 genes by calculating a standard deviation (SD) divided by the average, sorted in ascending order of SD/average values (Table S7), and plotted the results. In this rank plot (Fig. 4C), DUSP11 was at the 4381st rank, exhibiting greater variability than other 5′-PPP removal enzymes (see ranks of RNMT, RNGTT and DROSHA). Additionally, TF binding density at the DUSP11 promoter region was higher compared to the other three enzymes (Figs. 4B and S12). Collectively, our analysis showed a relatively wide expression range of DUSP11, when considering that it is a RNA 5′-maturation enzyme.

nc886 is mainly conserved in primates and does not exist in mice [24]. We conducted the same analysis for RNA-seq data of normal mouse tissues (n = 39), to plot SD/average values of 10,902 genes that were selected (average TPM of 39 tissues > 1) from total 47,531 genes (Table S8). Notably, in opposition to the human result, DUSP11 expression was more constitutive than the other 5′-PPP removal enzymes (Fig. 4D).

DUSP11 controls gene expression and pathways, which are suppressed by nc886

The difference in the variability of DUSP11 expression between humans and mice was intriguing and could be explained if nc886 alleviates the sensitized immunity, which is caused by low expression of DUSP11 and is potentially cytotoxic. To prove this idea, our initial plan involved a combination of DUSP11 KD and ectopic expression of nc886. However, the challenge arose from the fact that DUSP11 KD increased nc886 expression (Fig. 1), making it difficult to distinguish the effects of DUSP11 KD from nc886 overexpression. Fortunately, during the exploration of various cell lines, we found HCT-116 to be a unique cell line where DUSP11 KD did not increase nc886 expression (Fig. 5A). Taking advantage of this unexpected finding, we proceeded with DUSP11 KD and nc886 overexpression. After confirming the efficiency of these manipulations (Fig. 5A), RNA-seq was performed to obtain TPM values for a total of 60,624 genes across the four samples.

Fig. 5

DUSP11 KD alone affects gene expression and pathways, which is mitigated by nc886. A Northern and Western blot of indicated genes. Transfection of plasmids at 0 h, transfection of siRNAs at 24 h, and cell harvest at 48 h. All other descriptions are the same as Fig. 1C. B, C Box plots displaying fc values of 2 > increased 163 genes (panel B) and 2 > decreased 105 genes (C) upon DUSP11 KD. The y-axis of panel B and C indicates fold-increase and fold-decrease respectively. D A Venn diagram showing the number of altered genes (fc > 2) upon DUSP11 KD. A p-value was calculated by the Fisher’s exact test. E, F Box plots for the 57 genes (18 increased and 39 decreased upon DUSP11 KD) from D. All other descriptions are the same as panel B, C. G A heat map depicting the change of Biocarta pathways upon DUSP11 KD. 19 pathways were selected with Z-score cut-off = 3.5 from the “vector control” set. H A summary cartoon illustrating our study

Our initial analysis aimed to identify differentially expressed genes (DEGs) upon DUSP11 KD. In the "vector control" group, 163 genes were found to be increased, and 105 genes were decreased, with a TPM fc cut-off of > 2. Crucially, the fc values of these DEGs were significantly smaller in the "nc886 plasmid" group than the "vector control" group (Fig. 5B, C, Table S9). To identify genes likely to be directly controlled by DUSP11, RNA-seq data from DUSP11 KD samples in 293 T cells were also analyzed, resulting in 118 genes increased and 221 genes decreased by > twofold (Table S10). This 293 T gene set (339 DEGs) significantly overlapped with the HCT116 set (268 DEGs), yielding 57 common DEGs likely to include DUSP11’s direct target genes (Fig. 5D, Table