figure a

Introduction

The overwhelming majority of gene transcription events do not result in protein translation. Whereas some non-protein-coding transcripts represent bona fide ‘transcriptional noise’ [1], others are functional non-coding RNAs, such as transfer RNAs (tRNAs), ribosomal RNAs, small RNAs (including micro-RNAs) and long non-coding RNAs (lncRNAs). Arbitrarily greater than 200 nucleotides in length, lncRNAs, like their protein-coding counterparts, are commonly transcribed by RNA polymerase II, spliced, 5′ capped and polyadenylated [2]. Recent years have witnessed a tremendous upsurge in interest in lncRNAs amongst all disciplines ranging from fundamental biology to human disease, including diabetes and its complications. LncRNAs may contribute to islet cell dysfunction [3] and they appear to play important roles in the regulation of metabolism [4]. However, being in its relative infancy, the lncRNA field has also experienced its share of controversy and contradiction [5, 6]. A case in point is the archetypal trans-acting modular scaffold lncRNA, HOX antisense intergenic RNA (HOTAIR) [7].

HOTAIR is a 2,148 nucleotide lncRNA that is transcribed in an antisense direction from an intergenic region between HOXC11 and HOXC12 in the HOXC cluster on chromosome 12 in humans [7]. HOXC is one of four gene clusters (HOXA-D) that encode transcription factors necessary for the determination of positional identity along body axes during development [7, 8]. HOTAIR is the first lncRNA reported to function in trans, regulating the expression of genes encoded from the HOXD locus located on chromosome 2 [7]. It is considered to influence gene transcription by acting as a modular scaffold, approximating chromatin modifying complexes to their target sites on the genome. The 5′ domain of HOTAIR binds polycomb repressive complex 2 (PRC2), which includes the histone methyltransferase, enhancer of zeste homologue 2 (EZH2), that catalyses the trimethylation of histone H3 on lysine residue 27 (H3K27me3), a marker of gene repression [7]. Concurrently, a 3′ domain of HOTAIR has been reported to bind the LSD1/CoREST/REST complex, which contains lysine-specific histone demethylase 1A (LSD1) (together with corepressor for element-1-silencing transcription factor [CoREST] and repressor element-1 silencing transcription factor [REST]), an enzyme that demethylates H3K4me2, which is typically found at sites of active gene transcription [9].

Unlike some other lncRNAs [10], orthologues of HOTAIR are present in other mammals, offering an opportunity to gauge the biological importance of a single lncRNA at the level of the whole organism. Hotair has 90% nucleotide conservation in mice [7], it is transcribed from the micro-syntenic location corresponding to its cognate human counterpart [11], and it similarly binds both EZH2 and LSD1 [12]. Targeted deletion of Hotair has been reported to cause homeotopic transformation of the spine and malformations of the wrist bone [12]. However, the variable phenotypes reported have led some researchers to call into question the importance of the in vivo actions of Hotair, at least during embryological development [6, 8, 11,12,13,14].

As the explication of lncRNA biology matures through the study of model transcripts such as HOTAIR, lncRNA dysregulation has been increasingly linked to the pathogenesis of complex chronic diseases, amongst them diabetic kidney disease (DKD) [15,16,17,18,19,20,21,22]. Given the trajectory of discovery, it seems highly likely that new candidates and new mechanisms will be unearthed and proposed in the future. Our earlier work has served to define a pivotal role for the EZH2-regulated histone mark, H3K27me3 in preventing diabetes-induced damage to glomerular podocytes [23, 24]. In light of the reported biological importance of HOTAIR–EZH2 interaction and of the importance of EZH2 to the maintenance of podocyte health, we set out to examine the actions of HOTAIR in podocytes both in the normal state and when challenged with diabetes.

Methods

Human studies

Human kidney tissue was obtained at the time of nephrectomy for conventional renal carcinoma and the clinical characteristics of the patients have been described before [24]. The study was approved by the Nova Scotia Health Authority Research Ethics Board (Halifax, NS, Canada) and the Research Ethics Board of St Michael’s Hospital. A waiver of consent based on impracticability criteria was provided by the Nova Scotia Health Authority Research Ethics Board.

Mouse studies

Podocin-cre+ (B6.Cg-Tg(NPHS2-cre)295Lbh/J) mice [25] were obtained from The Jackson Laboratory (Bar Harbor, ME, USA) and were bred with HotairLoxP/+ mice [12] obtained from H. Chang (Stanford University School of Medicine, Stanford, CA, USA) to generate Podocin-cre+Hotair+/+ (HotairCtrl) and Podocin-cre+Hotairfl/fl (HotairpodKO) mice. These mice were studied at 10–14 weeks of age under non-diabetic conditions. To measure glomerular Hotair expression in experimental DKD, C57BL/6 mice (The Jackson Laboratory) were rendered diabetic with i.p. injections of streptozotocin (STZ; 55 mg/kg in 0.1 mol/l citrate buffer, pH 4.5) after a 4 h fast for 5 consecutive days. Non-diabetic control mice received an equal volume of citrate buffer. Blood glucose was measured 12 days after the first i.p. injection and mice were followed for a further 12 weeks. Additionally, male db/m and db/db mice (The Jackson Laboratory) were studied at 20 weeks of age. To examine the effects of podocyte-specific Hotair knockout under diabetic conditions, male HotairCtrl and HotairpodKO mice were rendered diabetic with STZ as described above and were studied 12 weeks after the first i.p. injection of STZ (i.e. 10 weeks after confirmation of diabetes). Mouse phenotyping methods are described in electronic supplementary material (ESM) Methods. All experimental procedures adhered to the guidelines of the Canadian Council on Animal Care and were approved by St Michael’s Hospital Animal Care Committee.

Digoxigenin-labelled in situ hybridisation and RNAscope in situ hybridisation

Digoxigenin-labelled in situ hybridisation was performed on formalin-fixed, paraffin-embedded human or mouse kidney tissues as described in ESM Methods. RNAscope in situ hybridisation (Advanced Cell Diagnostics, Hayward, CA, USA) was performed according to the manufacturer’s instructions and using custom software as previously reported [26] and as described in ESM Methods.

Isolation of podocyte-enriched mouse glomeruli

Podocyte-enriched glomerular cell extracts were isolated from mouse kidneys using the CELLection Biotin Binder Kit (Thermo Fisher Scientific, Rockford, IL, USA) as described in ESM Methods.

Immunofluorescence microscopy

For immunofluorescence microscopy, mouse slides were stained with antibodies in the following concentration: mouse nephrin 1:100 (AF3159, R & D Systems, Minneapolis, MN, USA), secondary antibody Alexa Fluor 488 donkey anti-goat 1:100 (Thermo Fisher Scientific). Co-staining was with the following antibodies: EZH2 1:100 (#5246, Cell Signaling Technology, Danvers, MA, USA), LSD1 1:100 (#2139, Cell Signaling Technology), H3K27me3 1:100 (#9733, Cell Signaling Technology), H3K4me1 (ab8895, Abcam, Cambridge, MA, USA), H3K4me2 1:100 (ab7766, Abcam), H3K4me3 1:100 (ab213224, Abcam), secondary antibody Alexa Fluor 647 donkey anti-rabbit 1:100 (Thermo Fisher Scientific). Slides were viewed using a Zeiss LSM 700 confocal microscope and analysis was conducted on approximately 30 podocytes from at least six glomeruli per slide using ImageJ 1.46r software (National Institutes of Health, Bethesda, MD, USA).

Transmission electron microscopy

Images of the entire glomerular profile were taken through three randomly selected glomeruli from 5–8 mice per group (Philips CM100, Electron Microscopy Research Services; Newcastle University, Newcastle, UK). Podocyte density was estimated by the method of Weibel and Gomez [27]. Podocyte density was multiplied by mean glomerular volume (determined in periodic acid–Schiff [PAS] stained kidney sections [28]) to give an estimate of podocyte number per glomerulus. Glomerular basement membrane thickness and podocyte foot process width were measured in a masked manner in approximately ten micrographs (magnification ×5800) from each glomerular profile using ImageJ (National Institutes of Health).

Quantitative reverse transcription-PCR

Primers were obtained from Integrated DNA Technologies and had the following sequences: Hotair forward AGGGTCCCCAACATCGGTAGA, reverse TGCGGTGGAGATAGATGTGC or forward AGCTGAGAAGGCCTGAATGA, reverse AAGGGGTGAACAGTGATCTG; Rplp0 forward GCGTCCTGGCATTGTCTGT, reverse GAAGGCCTTGACCTTTTCAGTAAG; Rpl13a forward GCTCTCAAGGTTGTTCGGCTGA, reverse AGATCTGCTTCTTCTTCCGATA. Real-time quantitative reverse transcription- PCR (qRT-PCR) was performed on a Viia 7 real-time PCR system (Thermo Fisher Scientific) using SYBR Green and data analysis was performed using the comparative Ct method.

Conditionally immortalised mouse podocytes

Differentiated, conditionally immortalised mouse podocytes were cultured as previously described [29]. Cells were serum starved for 4 h and incubated in normal media (5.6 mmol/l glucose) or in 25 mmol/l glucose (high glucose) for 48 h. Mannitol (19.4 mmol/l) added to normal media served as the osmotic control. For p65 knockdown, cells were transfected with 50 nmol/l mouse sequence-specific short interference RNA (siRNA) for p65 (sc-29411, Santa Cruz Biotechnology, Dallas, TX, USA) or scrambled siRNA (AM4611, Thermo Fisher Scientific) for 4 h prior to exposure to high glucose for 48 h. Chromatin immunoprecipitation by RNA purification (ChIRP) was performed using an EZ-Magna ChIRP Interactome Kit (EMD Millipore, Etobicoke, ON, Canada). Biotinylated Hotair probes were designed and synthesised by Biosearch Technologies (Petaluma, CA, USA) to cover the entire length of Hotair (ESM Table 1). The LacZ probeset was from Millipore. Primer sequences for Hotair targets (Hoxd1, Hoxd3, Hoxd11, Dlk1, Plag1, Dcn and H19) and control genes (Vamp5, Emp2) were as reported in [12] and were from Integrated DNA Technologies (Coralville, IA, USA). Values were normalised relative to LacZ binding compared with input DNA [30]. Chromatin immunoprecipitation (ChIP) was performed using a Magna ChIP kit (EMD Millipore) with an antibody against p65 (1:100 dilution; #8242, Cell Signaling Technology) or an equal concentration of normal rabbit IgG control (sc-2027, Santa Cruz Biotechnology). qRT-PCR was performed using primers specific for the mouse Hotair promoter (forward CCCAGCCAGGTAGGTAGAGT, reverse GAAGGGGCTGATGGATGCTT).

Immunoblotting

was performed with antibodies in the following concentrations: nephrin 1:1000 (AF3159, R & D Systems), p65 1:100 (#8242, Cell Signaling Technology), β-actin 1:10,000 (#A1978, Sigma-Aldrich, Oakville, ON, Canada). Densitometry was performed using ImageJ.

Mesangial matrix index

Mesangial matrix accumulation was calculated in approximately 30 glomeruli from PAS-stained kidney sections by investigators masked to the study groups, as previously described [31].

Immunohistochemistry

Immunostaining of mouse kidney tissue was performed as previously described [31] with the following antibodies: collagen IV 1:500 (AB756P, EMD Millipore), α-smooth muscle actin (α-SMA) 1:400 (ab5694, Abcam), nephrin 1:200 (AF3159, R&D Systems). The proportion of glomerular area positively immunostaining was determined in 30 random glomerular profiles in each kidney section using ImageScope for collagen IV and α-SMA and HALO version 2.3.2089.23 (Indica Labs, Corrales, NM, USA) for nephrin immunostaining.

Bioinformatics

Transcriptomic data were extracted from the NCBI Gene Expression Omnibus (GEO) database, accessible under the series number GSE66494. The details of the dataset are described in [32]. In brief, we studied the ‘discovery cohort’ consisting of transcriptomic data derived using the Agilent Whole Human Genome Microarray 4x44K from biopsy tissue from 48 patients with chronic kidney disease (CKD) compared with tissue from five control participants. Differential expression of HOTAIR, HOXC11, TGF-β1 (also known as TGFB1), CTGF (also known as CCN2) and CCL2 was assessed using limma, an R/Bioconductor software package. The false discovery rate was controlled using the Benjamini–Hochberg method. A p value of ≤0.05 was used to identify differentially expressed genes. Pearson correlation coefficient was calculated using Microsoft Excel. Boxplots and heatmaps were generated using limma.

Statistics

Data are expressed as means ± SD. Statistical significance was determined by one-way ANOVA followed by Fisher least significant difference post hoc test for multiple groups comparison or by two-tailed Student’s t test for comparison between two groups, unless otherwise stated. All statistical analyses were performed using GraphPad Prism 6 for Mac OS X (GraphPad Software, San Diego, CA, USA) unless otherwise stated.

Results

HOTAIR is expressed by human glomerular podocytes in vivo

To determine whether HOTAIR is expressed in adult human kidneys, we first performed digoxigenin-labelled in situ hybridisation on normal human kidney tissue. HOTAIR riboprobe binding was observed in scattered glomerular cells, parietal epithelial cells of Bowman’s capsule and tubule epithelial cells, especially tubule epithelial cells of distal convoluted tubules and cortical collecting ducts (Fig. 1a, b). Next, by combining RNAscope in situ hybridisation for HOTAIR and immunohistochemistry for the slit pore protein, nephrin, we confirmed HOTAIR expression by adult human glomerular podocytes (Fig. 1c). Finally, diminishing the likelihood of non-specific probe binding, we observed a significantly larger number of RNAscope puncta using the HOTAIR probeset compared with a probeset directed against the bacterial gene, dapB (negative control) where glomerular puncta were almost completely unapparent (Fig. 1d).

Fig. 1
figure 1

HOTAIR is expressed by kidney cells in normal adult human kidney tissue, including glomerular podocytes. (a, b) Representative digoxigenin-labelled in situ hybridisation for HOTAIR (purple) in normal human kidney tissue (a, scale bar, 100 μm; b, scale bar, 50 μm; n = 12). HOTAIR is present in tubule epithelial cells (especially distal tubules and cortical collecting ducts), parietal epithelial cells and glomerular cells. (c) Representative RNAscope in situ hybridisation for HOTAIR (red puncta) and immunohistochemistry for nephrin (brown) in a human glomerulus (scale bar, 50 μm). The black arrow labels RNAscope HOTAIR probeset binding in the nucleus of a peripherally arranged nephrin-positive cell (podocyte) in the zoomed-in image (n = 6). (d) Representative RNAscope in situ hybridisation (red puncta) in human glomeruli for HOTAIR, the bacterial transcript dapB (negative control) or the housekeeper PPIB (positive control) and quantification of RNAscope puncta (n = 6/probeset). Original images taken with a ×63 objective (scale bars, 20 μm); blue is DAPI. The white arrows point to HOTAIR RNAscope puncta. Values are mean ± SD. *p < 0.05 by two-tailed Student’s t test

Generation of podocyte-specific Hotair knockout mice

As we had observed in adult human kidneys, when labelling mouse kidneys with an RNAscope probeset, we observed Hotair expression in adult mouse glomerular podocytes (Fig. 2a, b). In both human (Fig. 1) and mouse (Fig. 2) kidneys, Hotair was observed to be primarily nuclear in its localisation. To determine the function of Hotair in glomerular podocytes, we generated podocyte-specific Hotair knockout mice. We studied two groups of mice, Podocin-cre+Hotairfl/fl mice and Podocin-cre+Hotair+/+ mice, henceforward referred to as HotairpodKO and HotairCtrl, respectively. Comparison of Hotair expression levels in podocyte-enriched glomerular cell fractions with Hotair expression in whole kidneys revealed an approximate 70% reduction in glomerular cell Hotair in HotairpodKO mice (Fig. 2c–e). Further confirming efficient excision of Hotair from podocytes, by dual immunofluorescence we observed absence of Hotair probeset binding in nephrin-positive glomerular cells in the kidneys of HotairpodKO mice, accompanied by a reduction in glomerular Hotair levels (Fig. 2f).

Fig. 2
figure 2

Hotair expression in mouse kidneys and generation of podocyte-specific Hotair knockout (HotairpodKO) mice. (a) Representative RNAscope in situ hybridisation (red puncta) in mouse glomeruli for Hotair, the bacterial transcript dapB (negative control) or the housekeeper Polr2a (positive control) and quantification of RNAscope puncta (n = 6/probeset). Original images taken with a ×63 objective (scale bars, 10 μm); blue is DAPI. (b) Representative RNAscope in situ hybridisation for Hotair (red puncta) and immunohistochemistry for nephrin (brown) in a mouse glomerulus (scale bar, 25 μm). The arrow labels Hotair probeset binding in the nucleus of a peripherally arranged nephrin-positive cell (podocyte) in the zoomed-in image (n = 5). (c) Immunofluorescence microscopy for nephrin (red) in cell extracts isolated from mouse kidneys using a biotinylated anti-nephrin antibody and streptavidin coated beads, showing enrichment of the cell extracts for nephrin-positive cells (podocytes). Original image taken with a ×63 objective (scale bar, 20 μm) (n = 8). (d) Immunoblotting of cell-lysates of mouse kidneys isolated using a biotinylated anti-nephrin antibody (bound) in comparison with cells not bound to magnetic beads (unbound) showing enrichment for the podocyte protein, nephrin, in bound cells (1 and 2 are replicates). (e) Ratio of Hotair in whole kidneys and podocyte-enriched glomerular cell extracts from HotairCtrl (n = 6) and HotairpodKO mice (n = 4), determined by qRT-PCR showing relative depletion of Hotair from podocyte-enriched (glomerular cell) extracts in HotairpodKO mice (values normalised to Rpl13a). (f) Representative RNAscope in situ hybridisation (red puncta) for Hotair in glomeruli from HotairCtrl and HotairpodKO mice. The zoomed-in images show Hotair transcript (white arrows) in nephrin-positive peripherally arranged cells (podocytes) in the glomerulus from the HotairCtrl mouse and absence of Hotair from podocytes in the glomerulus from the HotairpodKO mouse. Original images taken with a ×63 objective (scale bars, 10 μm). Quantification of Hotair RNAscope puncta in glomeruli from HotairCtrl and HotairpodKO mice (n = 8/group). Values are mean ± SD. *p < 0.05 by two-tailed Student’s t test

Knockout of Hotair from podocytes has no effect on albuminuria, glomerular structure, podocyte number or podocyte ultrastructure

Next, we examined the renal phenotype of otherwise normal adult HotairpodKO and HotairCtrl mice (aged approximately 10–14 weeks). There was no difference in body weight (Fig. 3a), systolic BP (Fig. 3b) or kidney weight (Fig. 3c) between HotairCtrl and HotairpodKO mice, although kidney weight:body weight ratio was slightly reduced in HotairpodKO mice (Fig. 3d). Albuminuria (Fig. 3e, f), plasma creatinine (Fig. 3g) and blood urea nitrogen (Fig. 3h) did not differ between HotairCtrl and HotairpodKO mice. Light microscopic appearance of the kidneys of HotairpodKO mice was unremarkable (Fig. 3i), as was glomerular volume (Fig. 3j). By transmission electron microscopy, we saw no overall change in podocyte density (Fig. 3k), podocyte number (Fig. 3l) or podocyte foot process width (Fig. 3m). Likewise, there was no difference between HotairCtrl and HotairpodKO mice in podocyte levels of the Hotair binding partners EZH2 and LSD1 or the histone marks that they regulate (H3K27me3 and H3K4me1-3) (ESM Fig. 1).

Fig. 3
figure 3

Knockout of Hotair from podocytes has minimal renal phenotypic effects. (a) Body weight, (b) systolic BP, (c) mean kidney weight, (d) kidney weight:body weight ratio, (e) 24 h urine AER and (f) urine albumin:creatinine ratio in HotairCtrl (n = 8) and HotairpodKO (n = 8) mice aged 10–14 weeks. (g) Plasma creatinine and (h) blood urea nitrogen (BUN) in HotairCtrl (n = 7) and HotairpodKO (n = 8) mice. There was insufficient sample for measurement of plasma variables in one HotairCtrl mouse. (i) Representative H&E (scale bars, 100 μm; HotairCtrl [n = 8], HotairpodKO [n = 8]) and PAS (scale bars, 25 μm; HotairCtrl [n = 7], HotairpodKO [n = 7]) stained kidney sections from HotairCtrl and HotairpodKO mice showing unremarkable kidney architecture in HotairpodKO mice. (j) Glomerular volume, (k) podocyte density, (l) podocyte number and (m) podocyte foot process width in HotairCtrl (n = 7) and HotairpodKO (n = 5) mice. The representative transmission electron micrographs in (m) illustrate normal foot process architecture in the glomerulus from a HotairpodKO mouse (scale bars, 2 μm). Values are mean ± SD. *p < 0.05 by two-tailed Student’s t test

HOTAIR expression is increased in the kidneys of humans and mice with diabetes

Having detected that Hotair is present in (but dispensable to) normal mouse podocytes, we next queried whether the expression pattern of HOTAIR is altered in DKD. We determined the magnitude of HOTAIR transcript levels in glomerular cells after digoxigenin-labelled in situ hybridisation of kidney tissue from individuals without diabetes and from individuals with pathologically confirmed diabetic glomerulosclerosis, observing an approximate 80% increase in glomerular HOTAIR in the setting of human diabetic glomerulosclerosis (Fig. 4a). Next, we examined Hotair levels in experimental diabetes in mice. By qRT-PCR, we observed an increase in Hotair expression in the kidneys of mice with STZ-induced diabetes (ESM Table 2) or db/db mice (ESM Table 3) (Fig. 4b). Likewise, quantification of glomerular Hotair labelling by in situ hybridisation also revealed an increase in Hotair abundance in the setting of STZ-induced diabetes (Fig. 4c) or in db/db mice (Fig. 4d).

Fig. 4
figure 4

HOTAIR expression is increased in human and experimental DKD. (a) Digoxigenin-labelled in situ hybridisation for HOTAIR and quantification of glomerular HOTAIR in kidney tissue from individuals without diabetes (h_Control, n = 12) and individuals with DKD (h_DKD, n = 11) (scale bars, 50 μm). (b) qRT-PCR in STZ-diabetic C57BL/6 mice (STZ, n = 6) in comparison with age-matched controls (n = 6) or in db/db mice (n = 8) in comparison with non-diabetic db/m mice (n = 8). (c) Digoxigenin-labelled in situ hybridisation for Hotair and quantification of glomerular Hotair in kidney tissue from control (n = 6) and STZ-diabetic (n = 6) mice (scale bars, 25 μm). (d) Digoxigenin-labelled in situ hybridisation for Hotair and quantification of glomerular Hotair in kidney tissue from non-diabetic db/m (n = 6) and diabetic db/db (n = 6) mice (scale bars, 25 μm). Values are mean ± SD. *p < 0.05, **p < 0.01 by two-tailed Student’s t test

Exposure to high glucose increases Hotair expression by mouse podocytes in a p65-dependent manner

To explore the possible mechanisms by which Hotair expression is upregulated by podocytes in the diabetic setting, we studied immortalised cultured mouse podocytes [29]. ChIRP demonstrated Hotair enrichment at known genomic sites [12] (Fig. 5a), whereas when podocytes were exposed to high glucose we observed a greater than doubling in Hotair levels (Fig. 5b). In considering how high glucose may upregulate HOTAIR, we were cognisant of work reporting a role for the p65 subunit of NF-κB in regulating the expression of human HOTAIR [33, 34]. We performed in silico analysis of the mouse Hotair promoter and identified three potential binding sites for NF-κB (ESM Fig. 2). By ChIP, we found p65 enrichment at the Hotair promoter in mouse podocytes (Fig. 5c). Supporting a role for p65 in mediating Hotair upregulation, siRNA directed against p65 negated the upregulation of Hotair induced by high glucose (Fig. 5d, e).

Fig. 5
figure 5

Hotair expression is increased in a p65-dependent manner in cultured mouse podocytes exposed to high glucose. (a) ChIRP for Hotair in cultured mouse podocytes. In comparison with LacZ binding (control), Hotair is enriched at genomic sites previously identified as being affected by Hotair binding (Hoxd1, Hoxd3, Hoxd11, Dlk1, Plag1, Dcn and H19), whereas there is no enrichment at unaffected sites (Vamp5, Emp2). (b) qRT-PCR for Hotair in mouse podocytes under control conditions (5.6 mmol/l glucose) or after exposure to high glucose (HG, 25 mmol/l) for 48 h or mannitol (osmotic control) (n = 3/condition). (c) ChIP for p65 at the Hotair promoter (IgG, n = 2; p65, n = 6). (d) Immunoblotting for p65 after transfection of mouse podocytes with p65 siRNA or scramble control for 48 h (scramble, n = 3; p65 siRNA, n = 3). (e) qRT-PCR for Hotair in mouse podocytes transfected with p65 siRNA for 4 h before exposure to HG for 48 h (control, n = 7; p65 siRNA, n = 8; HG, n = 8; p65 siRNA + HG, n = 8; mannitol, n = 7). Values are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 by one-way ANOVA followed by Fisher’s least significant difference post hoc test (b, e), two-tailed one-sample t test (c) or two-tailed Student’s t test (d)

Hotair knockout from podocytes does not affect the development of glomerular injury in diabetic mice

Because we had found HOTAIR to be upregulated in DKD, we set out to determine whether it plays a role in diabetes-associated glomerular injury. We rendered HotairCtrl and HotairpodKO mice diabetic with STZ. As expected, glomerular Hotair levels were increased in STZ-diabetic HotairCtrl mice in comparison with their non-diabetic counterparts, whereas glomerular Hotair expression was lower in STZ-diabetic HotairpodKO mice in comparison with STZ-diabetic HotairCtrl mice (ESM Fig. 3). Body weight was decreased (Fig. 6a) and HbA1c was increased (Fig. 6b) equivalently in STZ-diabetic HotairCtrl and STZ-diabetic HotairpodKO mice. Kidney weight was unchanged across experimental groups (Fig. 6c), whereas kidney weight:body weight ratio was reduced in non-diabetic HotairpodKO mice in comparison with STZ-diabetic HotairCtrl or STZ-diabetic HotairpodKO mice (Fig. 6d). Urine albumin excretion (Fig. 6e), glomerular basement membrane thickness (Fig. 6f) and mesangial matrix accumulation (Fig. 6g) were each equivalently increased in STZ-diabetic HotairCtrl and STZ-diabetic HotairpodKO mice. Likewise, glomerular collagen IV expression (ESM Fig. 4a) and α-SMA levels (ESM Fig. 4b) were increased to the same magnitude in STZ-diabetic HotairCtrl and STZ-diabetic HotairpodKO mice, whereas glomerular nephrin levels did not differ between the four experimental groups (ESM Fig. 4c).

Fig. 6
figure 6

Knockout of Hotair from podocytes has minimal effect on the renal phenotype of STZ-diabetic mice. Male HotairCtrl and HotairpodKO mice were rendered diabetic with STZ (or injected with citrate buffer, control) and were followed for 10 weeks. (a) Body weight, (b) HbA1c, (c) mean kidney weight, (d) kidney weight:body weight ratio and (e) 24 h urine AER. (f) Representative transmission electron micrographs with measurement of glomerular basement membrane thickness below (scale bars, 2 μm). (g) Representative PAS-stained kidney sections with semi-quantitative mesangial matrix index below (scale bars, 25 μm). In (ae, g), non-diabetic HotairCtrl, n = 7; STZ-diabetic HotairCtrl, n = 7; non-diabetic HotairpodKO, n = 5; STZ-diabetic HotairpodKO, n = 9. In (f), HotairCtrl, n = 7; STZ-diabetic HotairCtrl, n = 5; non-diabetic HotairpodKO, n = 5; STZ-diabetic HotairpodKO, n = 8. AU, arbitrary units. Values are mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001 by one-way ANOVA followed by Fisher’s least significant difference post hoc test

HOTAIR expression parallels the expression of the adjacent HOXC11 gene in human CKD

Finally, to further understand the role (if any) that HOTAIR may play in human CKD, we performed a bioinformatic analysis of transcriptomic data from a publicly available dataset of 48 biopsy samples obtained from individuals with CKD (including but not limited to DKD) compared with five control kidney RNA samples (GSE66494) [32]. We examined the expression patterns of HOTAIR and the adjacent HOXC11 gene, together with three other genes important in the pathogenesis of CKD: TGF-β1, CTGF and CCL2 [35,36,37]. Differential gene expression analysis also showed upregulation of HOTAIR expression in these CKD samples (Fig. 7a). Furthermore, heatmapping revealed that HOTAIR expression patterns closely parallel those of the HOXC11 gene in human CKD, whereas other genes implicated in CKD pathogenesis (e.g. TGF-β1, CTGF and CCL2) did not exhibit such a coordinated pattern of expression (Fig. 7b). Indeed, by Pearson correlation analysis we observed an extremely high correlation between the expression levels of HOTAIR and HOXC11 in human CKD (r = 0.87, p ≤ 0.001; Fig. 7c).

Fig. 7
figure 7

HOTAIR expression closely mimics the expression pattern of the adjacent HOXC11 gene in human CKD. Transcriptomic data were analysed from dataset GSE66494 (discovery cohort) consisting of RNA extracted from kidney biopsy tissue from 48 individuals with CKD and five control participants [32]. (a) Relative gene expression in CKD tissue (in comparison with control) was examined for HOTAIR, the adjacent HOXC11 gene and three genes implicated in the development of CKD (TGF-β1, CTGF and CCL2). The boxplot displays the median values (bold lines) of each gene with interquartile ranges (upper and lower limit of the boxes). The whiskers show the maximum and minimum, excluding outlying values. CKD vs control: HOTAIR p = 3.61 × 10−5, HOXC11 p = 1.98 × 10−7, TGF-β1 p = 0.352, CTGF p = 0.17 × 10−5, CCL2 p = 7.11 × 10−8 by two-tailed Student’s t test. (b) Heatmap of expression of the five genes relative to control showing a similar pattern of expression of HOTAIR and HOXC11 across the 48 CKD samples, that is not shared amongst the three comparator genes (TGF-β1, CTGF and CCL2). Colour key shows log2 fold change. (c) Correlation between expression levels of HOTAIR and HOXC11, r = 0.87, p ≤ 0.001 (Pearson correlation coefficient)

Discussion

In the search for new treatment modalities for DKD, recent interest has turned to the potential impact of therapeutically targeting non-coding RNAs, including lncRNAs [38]. Here, we report that the lncRNA HOTAIR is expressed by glomerular podocytes and upregulated in DKD. However, genetic knockout of Hotair has little effect on podocyte ultrastructure, glomerular health or renal physiology, either under normal conditions or when mice are challenged with experimental diabetes. Rather HOTAIR expression parallels that of its genic neighbour, HOXC11. At a time when interest in lncRNA biology is escalating rapidly, the results serve as a counterweight to emphasise that, in complex chronic diseases like DKD, lncRNA dysregulation may be a bystander without necessarily contributing to disease pathogenesis.

HOTAIR was originally described as a trans-acting lncRNA that is required for PRC2 occupancy at the HOXD locus [7], its subsequent interaction with the H3K4me2-demethylating LSD1/CoREST/REST complex further arguing in favour of a gene regulatory function [9]. However, although histone patterning at the HOX loci was originally studied as a means of discerning how lncRNAs may affect gene transcription, it has become apparent that the HOXD locus is not the only site of HOTAIR binding on the genome. For instance, over 800 genome-wide HOTAIR occupancy sites have been identified in human breast cancer cells and hundreds of genes are induced or repressed with HOTAIR overexpression in cancer cells [39, 40]. We observed HOTAIR expression in normal adult mouse and human kidney tissue and in mouse podocytes grown in culture, and we observed that HOTAIR expression is consistently increased in human and mouse DKD. This prompted us to explore the regulatory control of HOTAIR in podocytes and to inquire as to whether HOTAIR upregulation may be causatively implicated in the pathogenesis of DKD. We found that podocyte HOTAIR levels are increased by high glucose and that this upregulation is dependent on the p65 subunit of the proinflammatory transcription factor, NF-κB. However, Hotair knockout from podocytes is largely inconsequential either under normal conditions or in diabetes, with the exception of a small but statistically significant diminution in kidney weight:body weight ratio in HotairpodKO mice. This subtle change could be a consequence of podocyte Hotair deletion or it could be a consequence of local cis effects at the HoxC locus. Either way, although numerically significant, its biological importance is uncertain in light of the absence of change in other parameters of renal, glomerular or podocyte structure or function.

Residing in the intergenic region between Hoxc11 and Hoxc12, in mice Hotair has a tissue distribution that resembles the distribution of transcripts derived from its genic neighbours [8]. During development, HOXC11 is found in the metanephric mesenchyme [41], the embryonic region from which podocytes are derived [42]. In an entirely separate, publicly available, human CKD dataset [32] we observed that whole kidney HOTAIR levels are increased and that the expression of HOTAIR closely parallels that of the neighbouring HOXC11 gene. Only two of the 48 samples in this dataset were derived from patients whose primary diagnosis was DKD [32]. Thus, although we observed the p65-dependent upregulation of Hotair by high glucose in cultured mouse podocytes, diabetes is not the sole cause of HOTAIR dysregulation in CKD and podocytes are unlikely to be the sole cell-type in which HOTAIR is dysregulated. Rather, aberrant HOTAIR expression in human CKD is more probably indicative of broader genic dysregulation, including that of the HOX cluster genes, the consequences of which are uncertain in the fully developed organism.

The results herein presented are generally aligned with the studies of skeletal malformations in Hotair knockout mice, which overall suggest that Hotair is not a major determinant of developmental identity [6, 8, 11, 12]. They are also aligned with recent reports that many lncRNAs are largely dispensable in either mice or zebrafish [5, 30]. However, they are generally misaligned with other contemporary studies that describe a pathogenetic role for other lncRNAs in DKD [15,16,17,18,19,20,21,22]. There are several potential reasons that may underlie the largely benign consequences of the knockout of Hotair from podocytes. For instance, the significance and specificity of PRC2 or LSD1/CoREST/REST binding of HOTAIR is uncertain, especially in vivo [6]. PRC2, for example, is recruited to chromatin by thousands of different RNA transcripts [43,44,45]. In this context, the deletion of just one interacting lncRNA partner from one cell-type may be quite insufficient to induce a discernible effect on renal physiology, and the deficiency may be compensated for by any number of other non-coding transcripts present in the target nucleus. The expression of Cre recombinase in Podocin-cre+ mice begins to occur during the late capillary loop stage of embryonic development [25] and, in some other settings, a renal phenotype resulting from the experimental disturbance of histone modifying processes in podocytes has only become apparent during the process of ageing [46]. Accordingly, the results of the present study do not preclude an important role for HOTAIR at either extreme of the life course.

Several limitations of the currently presented experiments warrant emphasis. First, our survey of the consequences of Hotair deletion on histone marks was limited to assessment of global levels of H3K27me3 and H3K4me1-3 in podocytes. Previous analyses have indicated that HOTAIR inactivation causes H3K4me3 gain and, to a lesser extent, H3K27me3 loss at target genes [12], and we cannot determine the extent to which the podocyte regional histone landscape was altered in HotairpodKO mice. Second, whereas we approached the function of HOTAIR through the lens of its best-characterised mode of action, that of a modular scaffold [7, 9], HOTAIR has more recently been reported to also exert chromatin-independent effects. For instance, HOTAIR interacts with E3 ubiquitin ligases [47] and it can also function as a competing endogenous RNA (ceRNA) or micro-RNA sponge [48]. Third, our study of HOTAIR function in vivo focused on the consequences of deletion of the lncRNA from podocytes, and the importance of HOTAIR to other kidney cell-types remains to be determined. Likewise, it is possible that inducible deletion of Hotair from podocytes may have yielded a different phenotype to the constitutive deletion. Finally, given that we observed upregulation of HOTAIR in humans and mice with DKD, it remains to be determined whether overexpression of Hotair is either sufficient to promote kidney disease development under normal conditions or to accelerate kidney disease progression in the setting of diabetes. These limitations notwithstanding, we can be confident, however, that in normal mice or mice with experimental diabetes, inactivation of the lncRNA Hotair does not influence the renal phenotype in a biologically meaningful manner.

In summary, despite evolutionary differences in its encoding genes [49], the lncRNA HOTAIR is expressed by glomerular podocytes in both humans and mice and its expression is upregulated in experimental and human DKD. However, in this glomerular cell-type and in this disease setting, HOTAIR appears to play a largely redundant role. LncRNAs are vast in number, low in abundance and overlapping in function [50]. The present study highlights that lncRNA dysregulation can occur in complex chronic diseases like DKD without necessarily contributing to disease pathogenesis.