HOXA11-AS Regulates NIPAL3 via Hsa-miRNA-19a-3p, Hsa-miR-141-3p and Hsa-miR-140-5p in Keloid Fibroblast: Bioinformatic Analysis and In Vitro Validation


 Background: The long noncoding RNA HOXA11-AS is significantly increased in keloids with an unknown mechanism.Methods: Keloid-derived fibroblasts were primarily cultured from keloid explants (Sample 0). Small inference RNAs (siRNAs) against HOXA11-AS were transinfected in two keloid fibroblasts (Sample 1 and Sample 2). Nonspecific siRNA was transinfected into one keloid fibroblast (Sample 3). The blank plasmid was transinfected into one keloid fibroblast (Sample 4). Five keloid fibroblast samples were sequenced for lncRNAs, mRNAs and miRNAs. Differentially expressed (DE) mRNAs or lncRNAs were obtained in HOXA11-AS-knockdown keloid fibroblasts via DE RNAs from sample 1 intersected from sample 2, removing overlapping nonspecific interfered DE RNAs from sample 3 combined with sample 4. Using stepwise bioinformatics, dominantly functioning mRNAs downstream of HOXA11-AS were scored based on a first-step lncRNA-mRNA-protein (LMP) network in keloid fibroblasts, removing overlapping mRNAs from the nonspecific group. Screened genes were further validated by real-time PCR and western blotting. Validated mRNAs were overtaken to predict a sequence-matched competing endogenous ceRNA network involving the HOXA11-AS (down)- miRNA (up)- validated mRNA (down) pathway in keloid fibroblasts.Results: Six DE profiles (lncRNA, mRNA, miRNA) were obtained in the HOXA11-AS-knockdown group or in the nonspecific interference group. Fourteen dominantly functional genes were enriched in three biological pathways. SNED1, NIPAL3 and VTN were validated in keloid fibroblasts by knocking down HOXA11-AS via real-time PCR. Only NIPAL3 was predicted as a competing endogenous gene downstream of the HOXA11-AS pathway via three sponging miRNAs (hsa-miRNA-19a-3p, hsa-miR-141-3p, hsa-miR-140-5p).Conclusions: HOXA11-AS regulated NIPAL3 via three miRNAs in keloid fibroblasts.

removing overlapping nonspeci c interfered DE RNAs from sample 3 combined with sample 4. Using stepwise bioinformatics, dominantly functioning mRNAs downstream of HOXA11-AS were scored based on a rst-step lncRNA-mRNA-protein (LMP) network in keloid broblasts, removing overlapping mRNAs from the nonspeci c group. Screened genes were further validated by real-time PCR and western blotting.
Validated mRNAs were overtaken to predict a sequence-matched competing endogenous ceRNA network involving the HOXA11-AS (down)-miRNA (up)-validated mRNA (down) pathway in keloid broblasts.
Results: Six DE pro les (lncRNA, mRNA, miRNA) were obtained in the HOXA11-AS-knockdown group or in the nonspeci c interference group. Fourteen dominantly functional genes were enriched in three biological pathways. SNED1, NIPAL3 and VTN were validated in keloid broblasts by knocking down HOXA11-AS via real-time PCR. Only NIPAL3 was predicted as a competing endogenous gene downstream of the HOXA11-AS pathway via three sponging miRNAs (hsa-miRNA-19a-3p, hsa-miR-141-3p, hsa-miR-140-5p).

Background
Keloids are aberrantly proliferative benign tumors with remarkedly excessive deposition of broblasts and extracellular matrix (ECM) extending beyond the normal boundaries. Patients with genetic predisposition developed keloids during wound healing. The molecular pathways of how keloid broblasts proliferate into benign keloids remain largely unclear (1). Previous high-throughput pro ling has demonstrated that both encoding messenger RNA (mRNA) and long noncoding RNA (lncRNA) as well as microRNA (miRNA) cooperatively participate in the formation of keloids (2)(3)(4)(5)(6).
Long noncoding RNA of HOXA11 antisense RNA (HOXA11-AS) was shown to be markedly changed by in vivo array detection in keloids (7)(8)(9)(10). Additionally, using in vivo measurements and bioinformatics analysis, numerous microRNAs were shown to be endogenous competing molecular sponges downstream of the HOXA11-AS pathway in keloids (7)(8)(9). Notably, all these miRNAs in the HOXA11-AS pathways were heterogeneous, indicating a complicated molecular network. Moreover, HOXA11-AS was also shown to play a substantial role in various tumors (11). Interestingly, it could act as either a tumor promoter or tumor suppressor, depending on the type of cancer (reviewed by (11)). Moreover, even in one type of cancer (ovarian cancer), HOXA11-AS was found to be a tumor accelerator (12) or a tumor suppressor (13). All these studies indicated that HOXA11-AS regulated tumor cellular biology both extensively and intensively with a complicated molecular network (1). Therefore, HOXA11-AS might play as an essential "mediator" on benign keloids characterized by "paradoxical" hyperproliferation and high differentiation (14). In the present study, we used primary cultured keloid broblasts to investigate in vitro how HOXA11-AS regulates its target genes by a molecular network.

Methods
Clinical characteristics of patients undergoing complete keloid excision

Harvest Keloid-derived Fibroblasts From Keloid Explant
Fresh excised keloid samples were deeply punched into a 1 Mimi meter cube to obtain multiple keloid mini explants. Each mini-keloid explant was immediately fragmented into tiny pieces for primary culture of keloid-derived broblasts based on a previous approach (15). During primary culturing, fresh tiny keloid fragments were washed with phosphate-buffered saline (PBS; Cultilab, SP, Brazil), penicillin (100 Ul/ml; Gibco, Carlsbad, CA, USA) and streptomycin (100 µm/ml; Gibco) and incubated in Dulbecco's modi ed Eagle's medium (DMEM; Cultilab) for surface adherence. Consistent culture was conducted in DMEM 15% fetal bovine serum (FBS; Cultilab), penicillin (100 UI/ml; Gibco) and streptomycin (100 µg/ml; Gibco) at stable pH under a 5% CO2 atmosphere at 37°C. Keloid-derived broblasts were harvested from successfully cultured keloid mini explants for siRNA transfection. HOXA11-AS interference or nonspeci c interference in keloid broblasts via small interfering RNA (siRNA) Strains at passages 2-3 of primarily cultured keloid broblasts were used for siRNA interference using Transfection. The sequence of human lncRNA HOXA11-AS was TCCACAGCCTTTGCAGGCGGAATATCGGAATAAAGTGGGTCCAGGC. All siRNAs were designed and offered as commercial products from Sigma-Aldrich (Shanghai, China). Three 100 nM siRNA plasmids targeting HOXA11-AS were simultaneously transfected into two keloid broblasts (sample 1 and sample 2).
Simultaneously, plasmids harboring scrambled nonspeci c RNAs were transinfected using the same reagents and protocol (sample 3). Three 100 nM nonspeci c interfered siRNAs were transinfected simultaneously. A blank plasmid harboring no siRNA was transfected as a blank control (sample 4).
Normal broblasts were simultaneously detected in the normal group (baseline, sample 0). All transfected samples were validated by real-time PCR by comparing the relative expression to normal samples. All experiments were repeated three times.
Systematic sequencing of lncRNAs, mRNAs and lncRNAs in ve samples of keloid broblasts The sequences of lncRNAs, mRNAs and lncRNAs were measured in ve samples of keloid broblasts. Brie y, total RNA or miRNA of respective groups of cells was abstracted (Invitrogen Life Technologies), ampli ed and transcribed into uorescent cRNA (Arraystar, Rockville, MD). The labeled cRNAs were hybridized and deeply sequenced. Bioinformatic analysis by construction of a lncRNA-mRNA-protein (LMP) interaction network to screen dominantly functioning proteins in keloid broblasts with HOXA11-AS knockdown Using bioinformatic prediction, two lncRNA-mRNA networks (LMNs) were constructed in keloid broblasts with HOXA11-AS knockdown or in keloid broblasts with nonspeci c interference based on a human lncRNA-mRNA interrelated database (www.starbase.com) using a human protein-protein interaction (PPI) network as a "mediate" network. The rationale was that the dominant DE lncRNAs and DE mRNAs were both closely linked to PPI networks. Therefore, a proximity-based rst neighbor network (FNN) was predicted in keloid broblasts with HOXA11-AS knockdown or in keloid broblasts with nonspeci c interference. By this linkage, lncRNA-mRNA-protein (LMP) interaction networks were predicted in keloid broblasts with HOXA11-AS knockdown or in keloid broblasts with nonspeci c interference. Within two LMP networks, proteins were ranked by the random-walk PageRank (PR) algorithm, which has been commonly used in calculating the importance of webpages being visited based on both count and quality, leading to PR-ranked proteins. Proteins were considered dominantly functioning proteins when their PR values were over average. Finally, the dominantly functional genes from the two networks were intersected in removing overlapping mRNAs that were attributive to nonspeci c interference in keloid broblasts.
Validation of screened genes by quantitative real-time polymerase chain reaction (PCR) and western blotting Total RNA was extracted from normal keloid broblasts and from keloid broblasts with HOXA11-AS knockdown or with nonspeci c interference using an RNA extraction kit according to the manufacturer's protocol (Arraystar, Shanghai). Total cDNA was reverse transcribed from each sample using SuperScript III Reverse Transcriptase (Invitrogen). A standard curve was depicted by a diluted cDNA template gradient for each screened mRNA for qRT-PCR (Arraystar, Shanghai Kangcheng). The respective cRNA sample was used as a template for ampli cation. The sequences for primers of screened genes are shown in Supplementary Table S1. The qualitive concentration of each target gene was the relative percentage of gene concentration/GAPDH concentration. All real-time PCR have been conducted by three times.
Biomathematics analysis by constructing a lncRNA-mRNA-miRNA competing endogenous RNA (ceRNA) network using lncRNA-mRNA interrelationships combining mRNA-miRNA network interrelationships For the screened mRNAs by real-time PCR and/or western blotting in keloid broblasts with HOXA11-AS knockdown, both lncRNA-mRNA interrelationships and mRNA-miRNA interrelationships were constructed.
An expression-based competing endogenous interrelationship was downregulated lncRNA (knockdown of HOXA11-AS)-upregulated miRNA(s) (DE miRNAs)-downregulated mRNA(s) (all validated mRNAs). The DE miRNAs were identi ed between normal keloid broblasts (sample 0) and keloid broblasts with HOXA11-AS knockdown (sample 1 and sample 2). Among them, a sequence-based prediction of miRNAs as competitive sponges against the sequence of HOXA11-AS was further performed using the TargetScan 5.0 online tool. The outcome miRNAs undergoing expression and sequence dual validation were further identi ed by removing overlapping miRNAs that were also shown in the nonspeci c group. DE miRNAs in the nonspeci c group were obtained by union DE miRNAs between normal keloid broblasts (sample 0) and keloid broblasts with nonspeci c interference (sample 4) and DE miRNAs between normal broblasts at keloid broblasts with blank plasmid interference (sample 5).

Statistical Analysis
All experiments were repeated three times. All data are displayed as the mean ± standard deviation (S.D.). P < 0.05 was used to show statistical signi cance.

Results
Primarily cultured keloid broblasts in vitro from keloid explants in vivo Keloid-derived broblasts were shown to be successfully grown in a few mini-keloid explants (4/22).
These four samples generally originated from relative edges of the keloid, indicating that central broblasts had been losing proliferative capability (14). Moreover, primary culturing presented a considerably slow growth curve. The slow proliferation of primarily cultured keloid broblasts reached its peak velocity after 10 days and then plateaued for a long time. Such a proliferation curve could re ect the con ned hyperproliferation of keloids in vivo (14,16).
Using optical microscopy, typical keloid broblasts were clearly visualized after 10 days of culture, with low-to-moderate proliferation (Fig. 1A). Notably, high nutritional culture medium (DMEM 15% fetal bovine) was required to consistently sustain keloid broblasts with low proliferation. To display the extent of proliferation of these cultured keloid broblasts, a proliferative marker of Ki-67 antibody (green) was stained with nuclei stained with DAPI (blue). Using confocal microscopy, sparse Ki-17 positivity was shown in cultured keloid-derived broblasts (Fig. 1B). Globally, the growth of keloid broblasts is slow and sustainable.

Validation of the siRNA interference of HOXA11-AS in keloid broblasts
We validated the effect of interference of HOXA11-AS using real-time PCR between the two groups. It was signi cantly reduced in keloid broblasts with HOXA11-AS knockdown (Supplemental Fig. 1S), indicating that siRNA interference was successful.

HOXA11-AS markedly altered the expression of mRNAs in keloid broblasts
Heatmap expression ( Fig. 2A and 2B) and DE pro ling of keloid broblasts with HOXA11-AS knockdown ( Fig. 2C) or with nonspeci c interference ( Fig. 2D) is shown. The counts of DE mRNAs in two keloid broblasts with HOXA11-AS knockdown (sample 1 and sample 2) were similar to those in normal keloid broblasts (sample 0). Intersection of the two sets produced a total of 1,396 DE mRNAs in keloid broblasts with HOXA11-AS knockdown. Figure 2B indicates that DE mRNAs in keloid broblasts with nonspeci c interference were much higher than those with blank plasmid interference, indicating that nonspeci c sequences could substantially interfere with gene pro ling. The union of two sets produced 1,626 DE genes in keloid broblasts due to nonspeci c interference. Figure 3 included the heatmap express ( Fig. 3A and 3B) and DE lncRNAs in keloid broblasts with HOXA11-AS knockdown (Fig. 3C) or with nonspeci c interference (Fig. 3D). Only 39 DE lncRNAs were shown between keloid broblasts with HOXA11-AS knockdown (sample 1 and sample 2) and normal keloid broblasts (sample 0) (Fig. 3C). These DE lncRNAs were attributed to cross-talk between lncRNAs via an overlapping miRNA ne-tuning network. A total of 99 DE lncRNAs were shown in keloid broblasts with nonspeci c interference combined with blank plasmid interference (Fig. 3D). All DE pro les are shown in Supplementary document 1S.

Hoxa11-as Mildly Altered Lncrna Expression In Keloid Fibroblasts
Fourteen dominantly functioning genes in keloid broblasts with HOXA11-AS knockdown based on human lncRNA-mRNA-protein (LMP) interaction network prediction with PageRank scoring To screen potentially functional mRNAs regulated by HOXA11-AS in keloid broblasts, a stepwise bioinformatics analysis was performed. First, a lncRNA-mRNA network (LMN) was constructed using DE pro ling in keloid broblasts with HOXA11-AS knockdown. The LMN network contained 39 lncRNA nodes (red) and 63 mRNA nodes (orange) (Fig. 4A upper). Second, using the aid of a human protein-proteininteraction network, proximity-based "neighbor" proteins were selected as dominant (green band). Third, the green band was integrated into the constructed LMN network to predict a lncRNA-mRNA-protein (LMP) network (Fig. 4B upper). The LMP network contained 929 nodes and 6,572 edges, including the constructed LMN (39 lncRNA, red node; 63 mRNA, orange node) and rst-step dominant genes (green band). Fourth, using the advanced PageRank (PR) random walk algorithm, all screened functional mRNAs were scored. A total of 23 mRNAs ranked above the average PR score were screened as dominantly functioning genes in keloid broblasts by HOXA11-AS knockdown (Fig. 4C upper). Next, the biological enrichments removed 7 unenriched mRNAs from 23 screened mRNAs in keloid broblasts with HOXA11-AS knockdown, leading to 16 dominantly functioning proteins in keloid broblasts with HOXA11-AS knockdown (Fig. 4D upper). Thus, HOXA11-AS might regulate 16 genes that function in keloid broblast proliferation.
To remove nonspeci c interference, using the same protocol, an LMN containing 28 lncRNAs and 30 mRNAs was predicted using DE pro ling in keloid broblasts by nonspeci c interference (Fig. 4A, lower). An integrated LMP network was then constructed (Fig. 4B lower). A total of 14 mRNAs characterized by PageRank above the average score were screened in keloid broblasts by nonspeci c interference (Fig. 4C lower). As pathway enrichment was meaningless for nonspeci c interference, all 14 screened mRNAs remained the next step (Fig. 4D lower). After two mRNAs that emerged in both keloid broblasts by HOXA11-AS knockdown and in keloid broblasts with nonspeci c interference, they were removed as potential "noise". Consequently, a total of 14 mRNAs were thought to function as targets in the HOX11A-AS-involved molecular network during keloid formation (Fig. 4E).
Considering the long noncoding trait, these genes were supposed to be indirectly impacted within an enormous network, some of which were similar to Domino downstream.
Validation of candidate mRNAs in keloid broblasts by HOX11A-AS knockdown using qPCR and western blotting Using quantitative PCR, three genes (SED1, NIPAL3 and VTN) were shown to be signi cantly changed between keloid broblasts with HOXA11-AS knockdown and normal keloid broblasts. Therefore, these genes were supposed to be the dominant targets downstream of the HOXA11-AS1 regulatory pathway (Fig. 5B). Original datasets and curves of these fourteen qPCR assays are available on request.
However, western blotting detection showed negative expression of SED1, NIPAL3 and VTN either in normal keloid broblasts or in keloid broblasts with HOXA11-AS knockdown. These negative results were not shown. As all three genes belonged to endogenously expressed superfamilies with very low expression, trace expression could not be detected using western blotting because it harbors much lower sensitivity but higher speci city than PCR.
Screening potential molecular sponging miRNAs for HOXA11-AS to regulate three validated genes by competing endogenous RNA (ceRNA) network in keloid broblasts For three validated three genes with very low expression in keloid broblasts, a question arose: how are these genes regulated by HOXA11-AS? One of the most common regulatory mechanisms is the expression of lncRNAs combined with miRNA(s) to attenuate the inhibitory sponging effects of miRNAs on target genes. Therefore, we attempted to conduct another bioinformatic analysis to screen potential sponging miRNAs in keloid broblasts (Fig. 6).
First, we predicted 126 sequence-matched miRNAs that might interact with NIPAL3 (104 miRNAs), SNED1 (13 miRNAs) and VTN (9 miRNAs), referring to the human mRNA-miRNA competing endogenous RNA database among the DE miRNA pro les (2,17). Simultaneously, we conducted expression-based molecular interrelationships to identify the HOXA11-AS molecular pair. According to Down (HOXA11-AS knockdown)-Up (DE miRNAs)-Down (validated genes) in keloid broblast, upregulated VTN was removed, remaining downregulated NIPAL3 and SNED1. Sequence-based validation and expression-based dual validation addressed 10 upregulated candidate miRNAs with sequences competing with NIPAL3 among the DE miRNA pro les in keloid broblasts.
To further remove nonspeci c interference, DE miRNA pro ling in keloid broblasts with nonspeci c interference was analyzed, obtaining 495 upregulated DE miRNAs (Supplementary List 1). Among them, 7 miRNAs that also emerged in 10 candidate miRNAs were removed as nonspeci c interference. Consequently, three upregulated miRNAs, including hsa-miR-19a-3p, hsa-miR-141-3p, and hsa-miR-140-5p, were predicted as molecular sponges to speci cally downregulate the expression of NIPAL3 by the downregulation of HOXA11-AS in keloid broblasts.

Discussion
Our previous study revealed a signi cantly increased expression of HOXA11-AS in keloids by array-based lncRNA comparison (10). In the present study, we further attempted to reveal the potential regulation of potential miRNA targets involved in HOXA11-AS during keloid formation in vitro.
When we attempted to culture keloid broblasts primarily from in vivo keloid explants, we observed slow in vitro proliferation post 10 days and then plateaued without a hyperproliferation peak during the 50 days scenario. Moreover, a high nutritional microenvironment was required to sustain slow proliferation. Therefore, we supposed that the excessive proliferation of keloids ceased due to lower nutritional status.
Ki-67 staining validated the weak capability of proliferation even with high nutritional supply. Taken together, we hypothesized that the in vitro primarily cultured keloid broblasts could re ect the subtle regulation between differentiation and proliferation in vivo.
We next investigated the function of the lncRNA HOXA11-AS in the keloid broblasts using a routine lossof-function design. We simultaneously transfected three siRNAs to "knockdown" the expression of HOXA11-AS in keloid broblasts; however, we could only signi cantly downregulate the expression of HOXA11-AS in keloid broblasts. To reduce the "noise" from nonspeci c interference, we constructed plasmids harboring nonspeci c sequences for transfection and plasmids harboring blank plasmids for transfection. To enhance the speci city of the bioinformatics-based outcome, we removed the union pro les of both the nonspeci c group and the blank group.
We then conducted a stepwise bioinformatic analysis. First, we obtained three DE pro les (lncRNA, mRNA and miRNA) by knocking down HOXA11-AS or by nonspeci c interference. We removed overlapping DE molecules between the two groups to remove "noise". Then, we constructed a lncRNA-mRNA predictive network to screen dominant mRNAs downstream of the HOXA11-AS pathway. Our rationale was that spatiotemporal proximity meant functional dominance. By this step, we screened 14 dominant mRNAs among DE mRNAs as candidate functional genes in the HOXA11-AS-knockdown keloid broblasts. Intriguingly, these genes were largely endogenous genes enriched in three pathways: one pathway on cellular response to growth factor stimulus and two pathways on transcriptional regulation of endogenous genes. Using cellular validation, merely three genes (NIPAL3, SNED1, VNT) were found to be signi cantly changed among the 14 candidate genes in keloid broblasts via real-time PCR. No protein of these genes could be detected by western blotting even in normal keloid broblasts without any molecule knockdown, largely attributive to the trace expression of these endogenous genes.
Supposing NIPAL3, SNED1, and VNT are targets downstream of HOXA11-AS, we attempted to identify miRNA "mediators" by prediction of the ceRNA network. It is well accepted that endogenous competing miRNAs imperatively regulate cellular processes through their 3'-UTR binding to respective target mRNAs. We predicted sequence-based competing matching network using a public database (18) (19). Also, we predicted expression-based matching network on DE pro les. The present expression-based network was downregulated (HOXA11-AS knockdown)-upregulated (among upregulated DE miRNAs)-downregulated (PCR-validated NIPAL3 and SNED1) in keloid broblasts. The dual validation had identi ed ten upregulated sponging miRNAs against one gene of NIPAL3. After removing seven miRNAs that were also differentially expressed in keloid broblasts with nonspeci c interference, we constructed a nal ceRNA network characterized by downregulated HOXA11-AS, three upregulated miRNAs (hsa-miR-19a-3p, hsa-miR-141-3p and hsa-miR-140-5p), and reduced NIPAL3 in keloid broblasts.
In previous studies, the aberrative expression of HOXA11-AS was shown to correlate with a bulk of functions, including proliferation, migration, invasion and epithelial mesenchymal transition, via dozens of different miRNAs, presenting hundreds of regulatory pathways in various tumors (reviewed by 11).
Three remarkable strengths of the research were as follows: rst, three DE pro les (lncRNAs, mRNAs and miRNAs) were systematically sequenced with synchronicity; second, siRNA knockdown of HOXA11-AS was performed with complete nonspeci c interference control to reduce "noise" during high-throughput data analysis; third, advanced bioinformatics analysis was performed to construct a lncRNA-miRNA-mRNA network from interior interrelationships, with high respect to the integrative complexity of molecular network regulation; and fourth, the miRNA-mRNA pairs were dually validated by expression and sequence, avoiding subjective bias.
Our studies have several limitations. First, the validation sample size was small. Second, western blotting could not detect protein-level alterations even in normal keloid broblasts without any knockdown, indicating very low expression of the target gene. Third, case-speci c bioinformatic analysis lacked a standard protocol for high-quality control.
To the best of our knowledge, our ndings were the rst to prove that multiple miRNA networks are involved in ceRNA with potentially "opposite" bu ng regulation. More studies are required to investigate the subtle "balance" of these sponging miRNAs in the HOXA11-AS-dependent network to reveal the mechanism of keloid broblasts.    Figure 2B) were shown. The differentially expressed (DE) mRNA pro les between sample 0 and sample 1 as well as between sample 0 and sample 2 were intersected to obtain DE pro les in keloid bro blasts with HOXA11-AS knockdown ( Figure 2C). The differentially expressed (DE) mRNA pro les between sample 0 and sample 1 as well as between sample 0 and sample 2 were united to obtain DE pro les in keloid broblasts with nonspeci c interference ( Figure 2D).

Figure 3
Long noncoding RNA (lncRNA) expression heatmap and differentially expressed pro les in keloid broblasts Long noncoding RNA (lncRNA) expression heatmap of sample 0, sample 1 and sample 2 ( Figure 3A) as well as sample0, sample 3 and sample 4 ( Figure 3B) were shown. The differentially expressed (DE) mRNA pro les between sample 0 and sample 1 as well as between sample 0 and sample 2 were intersected to obtain DE pro les in keloid bro blasts with HOXA11-AS knockdown ( Figure 3C).
The differentially expressed (DE) mRNA pro les between sample 0 and sample 1 as well as between sample 0 and sample 2 were united to obtain DE pro les in keloid broblasts with nonspeci c interference ( Figure 3D).  (Figure 4D Lower). Among 16 screened mRNAs that were predicted to be involved in the HOXA11-AS pathway, 2 mRNAs that were also predicted to be involved in nonspeci c interference were removed as "noise", leaving 14 candidate mRNAs regulated by HOXA11-AS in keloid broblasts ( Figure 4E).

Figure 5
Pathway enrichment of 14 mRNAs and PCR validation in keloid broblasts with HOXA11-AS knockdown Three pathways were enriched in 14 putative candidate mRNAs in keloid broblasts involved in the HOXA11-AS pathway using GO online analysis ( Figure 5A). Three mRNAs, VTN, SNED1 and NIPAL3, were found to be signi cantly changed in keloid broblasts with HOXA11-AS knockdown compared with normal controls using qPCR validation ( Figure 5B).