Tumor progression has been significantly linked with EMT, and therefore it has captured increasing interest for researchers. Indeed, the association of EMT with cancer has been reported since the early 80s ^[1][16]. EMT refers to an escape mechanism that takes place in cancer cells and is associated with tumor development, invasion, and metastasis to secondary sites ^[2][17]. This process includes the transdifferentiation of epithelial cells to a mesenchymal phenotype which exerts invasive and migratory properties ^[3][18]. During the orchestration of EMT, three biological phenomena may occur simultaneously. The first one involves the disruption of cell-to-cell contact and epithelial tissue polarity. Second, the remodeling of the cytoskeleton leads to the restructuring of epithelial cells to fibroblast-like cells, and third, several changes in epithelial and mesenchymal markers occur. These changes concern alterations to EMT-related proteins, including a decrease in E-cadherin expression (an epithelial-specific factor) and an overexpression of N-cadherin and vimentin (mesenchymal-specific factors). Moreover, the switch in gene expression from epithelial to mesenchymal phenotype is also triggered by (a) EMT-inducible transcription factors (ZEB1, ZEB2, Snail, Slug, and Twist), (b) non-coding RNAs, (c) epigenetic modifications and posttranslational regulation, and (d) alternative splicing events ^[4][5][6][7][19,20,21,22]. Over the years, multiple triggering factors of EMT have been identified, including (i) microenvironment signals (such as hypoxia and oxidative stress) ^[8][23], (ii) growth factors [epidermal growth factor (EGF), and fibroblast growth factor (FGF)] ^[9][10][24,25], and cytokines secreted by the tumor microenvironment [tumor necrosis factor-alpha (TNF-α), IL-8, IL-60] ^[11][12][13][26,27,28], (iii) immune responses (such as elevated expression of immune checkpoints, namely PD1, PD-L1, LAG3, and CTLA4) ^[14][29], and (iv) regulation of drug- and radio-resistance-related genes ^[8][23]. Indeed, there is an intriguing relationship between EMT and the immune system. There is evidence that the cells of the innate and adaptive immune system (i.e., macrophages, natural killers, and Tregs) promote EMT ^[15][16][30,31], whereas mesenchymal-like tumor cells possess immunosuppressing functions ^[17][32]. More importantly, surgical debulking of OC has been linked to significant reductions in circulating Tregs and increases in CD8+ T-cells, supporting the idea that upfront cytoreduction has a valuable systemic effect by improving immunological state ^[18][33]. Furthermore, tumor-infiltrating lymphocytes (TILs) have been also associated with optimal primary surgical cytoreduction ^[19][34]. A recent comprehensive review by Huang et al. thoroughly describes all the molecular mechanisms of EMT involved in tumor progression and metastasis ^[20][6].

Wu et al. first reported that increased levels of E-cadherin can induce cell-to-cell cohesion, preventing cellular migration in human OC cells ^[21][35]. This highlighted the importance of EMT in tumor cell shedding into the peritoneum and intra-abdominal metastases. Indeed, as opposed to other epithelial malignancies that disseminate via the canonical invasion–metastasis cascade mechanism (which includes the hematogenic or lymphatic route) ^[22][36], epithelial ovarian cancer (EOC) metastasizes predominantly via a transcoleomic route of spread. Briefly, this route involves the peritoneal cavity with direct cancer cell shedding (both as single cells and as cell aggregates) from the primary tumor into the peritoneal fluid. Subsequently, their adherence to intraperitoneal tissues leads to their anchoring into the sub-mesothelial matrix, where they generate secondary lesions ^[23][24][37,38]. This unique peritoneal microenvironment exhibits an exceptional phenotypic plasticity, promoting epithelial-to-mesenchymal and mesenchymal-to-epithelial transitions. More specifically, OC cells exist in the peritoneal cavity in a heterogenous variety of phenotypes along the EMT spectrum (epithelial, hybrid, and mesenchymal), with the ability to dynamically shift from one state to another, responding to micro-environmental stimuli. Epithelial-type cell aggregates have been associated with radio- and chemo-resistance as well as resistance to apoptosis. Several bioactive soluble factors [lipid lysophosphatidic acid (LPA), growth factors, and matrix metalloproteinases] facilitate the generation of a soluble E-cadherin fragment which may trigger the disruption of intercellular junctions between cells with epithelial and hybrid phenotypes and promote cell exfoliation from the primary tumor. On the other hand, mesenchymal-type cell aggregates have been associated with metastasis via the invasion of peritoneal matrix. The latter can be also enhanced by the soluble E-cadherin fragment. Finally, the hybrid-type cells exhibit stemness properties with tumor initiation and growth potential. In mixed-cell aggregates (hybrid–epithelial or hybrid–mesenchymal), hybrid cells may differentiate to the adjacent cell phenotype and share their functions ^[25][39].

Several EMT regulators have been identified in OC and described in depth elsewhere ^[25][39]. In brief, besides non-coding RNAs, various components of the ascitic fluid, epigenetic changes, post transcriptional modifications, and biomechanical forces have been proposed to trigger EMT. For example, the role of LPA in EMT promotion has been widely established. Gil et al. reported that LPA favors the mesenchymal phenotype with enhanced invasion potential, via the shedding of an 80-kDa E-cadherin-soluble fragment in a urokinase plasminogen activator (uPA)-dependent manner ^[26][40]. EMT induction by LPA in OC has been also associated with (i) SIRT1 downregulation, a known ZEB1 inactivator and EMT suppressor ^[27][41] and (ii) Slug/Snail2 upregulation via the Gαi2, Src and HIF1α signaling nexus ^[28][42]. In addition, LPA has been correlated with matrix metalloproteinase-9 (MMP-9) overexpression and MMP-9-associated E-cadherin ectodomain shedding, which results in the disruption of cell–cell junctions between cancer cells ^[29][43]. Following the dysfunction of junctional integrity, LPA further stimulates EMT via the β1-integrin-dependent activation of Wnt/β-catenin signaling pathway ^[30][44]. Finally, it has been suggested that MMP-2 activation by LPA leads to enhanced migratory and invasive potential of epithelial OC cells ^[31][45].

Besides LPA, growth factors [such as epidermal growth factor (EGF), transforming growth factor beta (TGF-β) and hepatocyte growth factor (HGF)] which are also present in the ovarian tumor microenvironment have been associated with EMT induction through (i) the promotion of the cleavage of Ecad fragments to abrogate cell–cell adhesion and (ii) the activation of complex signaling pathways ^[32][33][34][46,47,48]. Moreover, MMPs facilitate EMT by inducing the cleavage of the Ecad ectodomain and suppressing the tight junction protein zona occludens-1 (ZO-1). These processes lead to impaired intercellular cohesion and activation of the Wnt signaling pathway, resulting in enhanced mesenchymal properties and thus increased invasive potential in the tumor cells ^[35][36][37][49,50,51]. Simultaneously, EMT-related factors such as Snail and β-catenin have been reported to regulate MMP expression ^[38][39][52,53]. Another component of ascitic fluid, the Wnt5a protein has been also associated with EMT shifts ^[40][54]. Furthermore, EOC-associated malignant ascites are enriched in IL-6 and IL-8, which have been believed to play a pro-EMT role in OC ^[41][42][55,56]. Notably, increased expression of both these cytokines has been also associated with LPA via the transcriptional activation of IL gene promoters ^[43][57].

In addition to the above, several EMT-associated epigenetic modifications have been described over the years. For instance, epigenetic silencing of CDH1 (the gene that encodes e-cadherin) can result from (i) hypermethylation via the 5′ CpG island ^[44][58] (ii) methylation via ZEB1 through the recruitment of DNA methyltransferase 1 (DNMT1) to the CDH1 promoter ^[45][59] and (iii) mono-, di and trimethylation of the CDH1 promoter ^[46][60]. On the other hand, reactivation of CDH1 has been linked with deacetylation of Slug ^[47][61]. Another example of EMT-related epigenetic change in OC involves the TGF-β-induced methylation of several genes such as (a) CDH1, FGFBP1, SNAI3, and MMP9, resulting in their downregulation; and (b) MMP2, MMP3, ZEB2, TGFB2 and SNAI1, resulting in their upregulation ^[48][62].

Accordingly, various modifications have been reported to regulate EMT at the posttranscriptional level. Briefly, phosphorylation of SNAI1 by protein kinases leads to its degradation and thus increased E-cadherin expression, whereas inhibition of these enzymes results in decreased levels of E-cadherin and EMT promotion ^[49][63]. Moreover, Snail1 glycosylation prevents its degradation, thus inducing EMT ^[50][64]. The mediation of EMT in the posttranscriptional setting has been also linked with the impairment of critical molecules of the HIPPO signaling pathway (namely YAP/TAZ, MAT1/2, and LATS1/2) due to abnormal regulation of posttranscriptional modifications ^[51][65]. Finally, sumoylation of Smad-Interacting Protein 1 (SIP1) has been correlated with E-cadherin downregulation ^[52][66]. The following sections will describe EMT-related mechanisms in ovarian cancer initiation and progression, focusing on the role of long non-coding RNAs.

LncRNAs represent a subtype of non-coding RNAs with distinct regulation properties in cancer biology, including cellular proliferation, invasion, and metastasis ^[53][8]. Although a growing number of studies have demonstrated their implications for protein-encoding genes, their role in terms of regulating EMT/MET in ovarian cancer has been relatively recent.

Recently, lncRNA AC005224.4 was found to exhibit significant oncogenic properties, affecting EMT-associated markers in OC cell lines. The authors suggested that this lncRNA promotes OC cell proliferation, metastasis, and EMT via sponging miR-140-3p and thus increases SNAI2 expression ^[54][67]. Plasmacytoma-variant translocation 1 (PVT1) enhanced the proliferative and migrative potential of OC cells through the modulation of the expression of connective tissue growth factor (CTGF), which in turn also promotes the EMT process ^[55][69]. Moreover, SNHG17 knockdown was found to repress EMT in OC cells via the regulation of miR-485p/AKT1 axis ^[56][71]. Another lncRNA, OIP5-AS1, was upregulated in OC. Downregulation of OIP-AS1 repressed EMT, migration, and invasion while facilitating apoptosis of OC cells. More specifically, this lncRNA was reported to function as a competing endogenous RNA (ceRNA) by sponging miR-92a to regulate integrin alpha 6 (ITGA6) ^[57][72]. Similar conclusions were reported in a second study; in the latter, the investigators supported the idea that OIP-AS1 acts as ceRNA for miR-137, resulting in ZNF217 upregulation and EMT promotion ^[58][73]. HCG18 silencing suppressed proliferation, migration, and EMT in OC cells and reduced ki67 levels in vivo. This lncRNA also functions as a ceRNA for miR-29a/b, resulting in TRAF4/5 downregulation, which subsequently activates the NF-κB signaling pathway ^[59][74]. Similarly, LINC00909 acts as a ceRNA for MRC2 mRNA by sponging miR-23-3p and thereby induces EMT in OC cells ^[60][82]. Liu et al. proposed that hypoxia-induced MIR210HG was positively correlated with poor prognosis and cancer progression. Further results of their work demonstrated that knockdown of MIR210HG suppressed EMT and angiogenesis under hypoxic conditions ^[61][75]. Another widely studied lncRNA in cancer is MALAT1. According to a recent work, MALAT1 functions as a ceRNA interfering with miR-22 and consequently increasing the expression of Snail in OC ^[62][76]. EMT promotion in OC cells by MALAT1 has been also associated with the regulation of additional EMT hallmarks, such as ZEB2 ^[63][77] and RBFOX2 ^[64][78]. Moreover, MALAT1 expression has been negatively correlated with 5-year OS and it has been suggested that it induces EMT via the PI3K/AKT pathway ^[65][79]. Similarly, increased expression of the oncogene H19 promotes the OC-related EMT phenotype via direct binding to miR-140-5p and activation of the PI3K/AKT pathway ^[66][104]. In addition, H19 has been reported to facilitate TGF-β-induced EMT by sponging miR-370 in OC cells ^[67][103], while SRA overexpression regulated EMT-related markers and the NOTCH pathway ^[68][81]. A work conducted by Xue et al. linked lncBCAS1-4_1 upregulation with EMT and vitamin D signaling. According to the researchers, lncBCAS1-4_1 could eradicate the antitumor function of the 1α,25(OH)2D3 due to ZEB1 upregulation ^[69][80]. The effect of LINC00922 on OC cell proliferation, invasion, migration, and EMT was investigated by Wang et al.; in brief, they reported that this lncRNA competitively binds with miR-361-3p, resulting in the upregulation of CLDN1 and activation of the Wnt/β-catenin signaling pathway ^[70][83]. Indeed, the enhancement of the Wnt/β-catenin signaling pathway in OC has been associated with more than one lncRNA. For example, alongside the regulation of epithelial and mesenchymal markers, SNHG8 silencing resulted in downregulation of EMT-associated stem factors such as ALDH1, Nanog, SOX2, OCT4, CD44, and CD133 through the Wnt/β-catenin pathway ^[71][95]. In addition, HCP5 was reported to (a) serve as a sponge for miR-525-5p, preventing its further binding to polycomb repressive complex 1 (PRC1) and (b) promote the Wnt/β-catenin signaling pathway ^[72][96]. Moreover, Lou et al. found that ROR promoted EMT contribution to cell migration and invasion through the activation of the canonical Wnt/β-catenin signaling pathway ^[73][107]. It has also been suggested that this lncRNA acts as a sponge for the tumor-suppressive miR-145 and its target gene FLNB ^[74][106]. Finally, two more lncRNAs, HOXB-AS3 and CCAT2, exert their EMT-promoting functions through activation of the Wnt/β-catenin signaling pathway ^[75][76][110,119].

EMT-mediated OC progression has been also associated with a positive feedback loop of the lncRNA DSCR8 via the miR-98-5p/STAT3/HIF-1α nexus ^[77][92]. The activation of the STAT3 signaling pathway has been also attributed to the lncRNA ATB ^[78][102]. PAXIP1-AS1 promotes cell proliferation and EMT by sponging miR-6744-5p to regulate PCBP2 expression ^[79][84], whereas MAFG-AS1 was reported to facilitate EMT progress by upregulating the NFKB1-dependent IGF1 via miR-339-5p ^[80][86]. Another lncRNA, CTSLP8, was found to promote OC in vitro and in vivo by sponging miR-199a-5p. Mechanistically, the authors reported that lnc-CTSLP8 exerted its oncogenic effects through CTSL1 upregulation ^[81][87]. EMT-related markers were also identified as regulated by HOTIPP overexpression via activation of the MEK/ERK pathway ^[82][91]. Moreover, nuclear-enriched abundant transcript 1 (NEAT1) was found to sponge miR-1321 and thus regulate the expression of the tight junction protein 3 (TJP3) (an important factor in cancer development), resulting in EMT promotion, invasion, and migration of OC cells ^[83][93]. LINC00858 has also been associated with OC proliferation, motility, and EMT by sponging miR-134-5p to upregulate RAD18 ^[84][94]. The findings of another study demonstrate that E2F4as promoted tumor aggression by regulating EMT-associated mechanisms, thereby predicting patients’ prognosis ^[85][89]. Additionally, the EMT phenotype in OC cells has been found to be mediated by (a) MIAT by sponging miR-150-5p ^[86][99], (b) NORAD by regulating miR-199a-3p ^[87][100], and (c) LINC00963 by targeting miR-378g ^[88][101]. PTAL (also known as AC004988.1) has been also associated with facilitating EMT via the miR-101-FN1 axis ^[89][105]. Furthermore, FLVCR1-AS1 was found to bind directly to miR-513, thus downregulating its expression. In addition, the authors reported that miR-513 underexpression might be linked with YAP1 upregulation ^[90][109]. Similarly, according to Yan et al., DQ786243 could directly bind to miR-506 and thus increase cAMP responsive element binding protein 1 (CREB1) expression ^[91][115]. Another study, using bioinformatics analysis, noted that lncRNA PTAR expression was positively correlated with ZEB1 expression in OC; the same work also suggested that PTAR enhanced EMT by binding to miR-101-3p ^[92][116]. In addition, Linag et al. proposed that lncRNA PTAF regulated EMT in OC through the miR-25-SNAI2 axis ^[93][118].

The association between lncRNA expression levels, EMT, and OS in OC patients has been also investigated in several studies. For example, LINC01969 was linked with OC progression and worse OS by regulating the miR-144-5p/LARP1 axis as a ceRNA ^[94][88]. Similarly, LINC01094 overexpression has been associated with higher FIGO stage, lymph node (LN) metastasis, and shorter OS in patients with OC. Conversely, LINC01094 silencing retarded OC cell proliferation, migration, invasion, and EMT by targeting miR-577 and activating the Wnt/β-catenin signaling pathway ^[95][90]. Liu et al. suggested that LINC01215 promotes EMT-related LN metastasis through the methylation and downregulation of the RUNX3 promoter ^[96][85]. Moreover, HOXA cluster antisense RNA 3 (HOXA-AS3) overexpression has been correlated with significantly shorter PFS and OS, whereas its downregulation has been found to inhibit the EMT process in vitro ^[97][68]. Interestingly, TC0101441 levels may be a potential independent prognostic marker of OS and DFS in EOC through the downregulation of KISS1 (a metastasis suppressor gene) ^[98][108]. Furthermore, downregulation of AOC4P has been positively correlated with FIGO stage and LN metastasis by promoting EMT ^[99][123]. Increased PVT1 expression has also been associated with advanced FIGO stage and poor prognosis for OC patients. On the other hand, silencing of PVT1 impaired cell proliferation, migration, and invasion in vitro. More specifically, the EMT-related promotion of OC by PVT1 involved its interaction with EZH2, which resulted in the repression of miR-214 expression ^[100][70]. Accordingly, LINC01296 overexpression was negatively linked with progression-free survival (PFS) and OS, whereas knockdown of LINC01296 inhibited EMT in OC ^[101][97]; the underlying mechanism of its oncogenic function involves the regulation of EMT through the LINC01296/miR-29c-3p axis ^[102][98]. A negative correlation between HOXD-AS1 expression and PFS and OS in EOC patients has been also reported. Two EMT-associated mechanisms were identified in current literature. More specifically, this lncRNA acts as a ceRNA (i) via the miR-186-5p/PIK3R3 pathway ^[103][111] and (ii) via the miR-133a-3p/Wnt/β-catenin pathway ^[104][112]. lncARSR expression was also correlated with FIGO stage, histological grade, LN metastasis, and poorer OS. In terms of mechanism, lncARSR was found to increase ZEB1 and ZEB2 expression by sponging members of the miR-200 family ^[105][117]. Following the conduction of an integrated analysis, Mitra et al. identified 3 EMT-related lncRNA (DNM3OS, MEG3 and MIAT) out of which DNM3OS was found to be associated with worse OS ^[106][120]. Another lncRNA, HOXA11, has been significantly associated with histological grade and preoperative CA-125, as well as EMT promotion and shorter PFS and OS, in patients with serous OC ^[107][121]. Accordingly, HOTAIR and CCAT1 expression were highly positively correlated with FIGO stage, histological grade, LN metastasis, and reduced OS ^[108][109][114,122]. The pro-metastatic effects of HOTAIR were partially mediated by the regulation of EMT-related genes ^[109][122] whereas CCAT1 promoted OC progression via a miR-152/miR-130b-ADAM17/WNT1/STAT3/ZEB1 regulatory network ^[108][114]. Finally, CCAT1 was found to act as a ceRNA by directly sponging miR-490-3p and thus elevating TGFβR1 and promoting EMT ^[110][113].

4. EMT-Related lncRNAs: Current Evidence

4.1. EMT-Related lncRNAs in Cancer Diagnosis

Several EMT-related lncRNAs that were retrieved have already been evaluated for their value as diagnostic biomarkers in various cancer types. One of the best-recognized lncRNAs with high diagnostic accuracy is MALAT1. Moreover, it has been proposed as a biomarker for the early diagnosis of prostate cancer (PC); its expression was found to be positively correlated with high Gleason score and PSA level and it has been patented (CN104498495) as an auxiliary biomarker for early PC diagnosis, particularly for PSA “gray area” cases ^[111][112][138,139]. A meta-analysis by Chen et al. concluded that the differential expression of MALAT1 in various tumor types, including OC, can support its diagnostic potential ^[113][140]. NEAT1 has been proposed as a therapeutic target for OC ^[83][93]. However, this lncRNA has exhibited high diagnostic potential for NSCLC, displaying both high sensitivity and specificity ^[114][141]. Similarly, while HOTAIR has been associated with poor prognosis in OC ^[109][122], it also exhibits good characteristics as a diagnostic biomarker for multiple tumors, such as diffuse large B-cell lymphoma (DLBCL) (AUC: 0.71; sensitivity, 72.6%; specificity, 69.7%) ^[115][142], glioblastoma multiforme (GBM) (area under the ROC curve: 0.913; sensitivity, 86.1%; specificity, 87.5%) ^[116][143], and esophageal squamous cell carcinoma (ESCC) (AUC: 0.793) ^[117][144]. Notably, HOTAIR and MALAT1 are also under patent (CN105586399A) as diagnostic biomarkers for papillary thyroid cancer. Another relevant lncRNA with diagnostic potential is H19. More specifically, the AUC of plasma lncRNA H19 was found to be 0.81 in breast cancer patients (sensitivity, 56.7%; specificity, 86.7%), a finding which is higher than that of traditionally used breast cancer biomarkers such as carbohydrate antigen 153 (CA153) and carcinoembryonic antigen CEA ^[118][145]. Moreover, the diagnostic ability of H19 was high for gastric cancer (GC) (AUC: 0.838) ^[119][146] as well as papillary thyroid cancer ^[120][147]. The diagnostic and prognostic efficacy of exosomal HOTIPP has been also suggested for GC, displaying higher AUC compared to other biomarkers such as CEA, CA 19-9 and CA 72-4 ^[121][148]. Interestingly, the value of CCAT1 as a diagnostic biomarker is being currently investigated in a clinical trial (NCT04269746) in colorectal cancer (CRC). Indeed, the combination of CCAT1 and HOTAIR levels has been reported to display good diagnostic characteristics in CRC ^[122][149]. Finally, the additional evaluation of CCAT2 levels to CA-125 and squamous cell carcinoma antigen (SCC) levels was suggested to improve the diagnosis of cervical cancer ^[123][150].

4.2. EMT-Related lncRNAs in Cancer Prognosis

The aberrant expression of oncogenic lncRNAs correlates with unfavorable effects regarding OS, metastasis, and tumor stage and grade; thus lncRNAs could potentially serve as prognostic biomarkers. In the following paragraph, here will briefly discuss the prognostic value of EMT-related lncRNAs in multiple tumor types. In addition to OC ^[124][151], MALAT1 has been also proposed as a prognostic marker for CRC ^[125][152], hepatocellular carcinoma (HCC) ^[126][153], squamous cell lung cancer ^[127][154], BC ^[128][129][155,156], GC ^[130][157], esophageal cancer ^[131][158], and osteosarcoma ^[132][159]. Moreover, MALAT1 and HOTAIR levels have been suggested as a putative predictive duo of neuroendocrine transformation in PC ^[133][160]. The expression signature of HOTAIR has been also associated with poor prognosis in various cancer types, such as bladder cancer ^[134][161], GC ^[135][162], and CRC ^[136][163]. CCAT1 has been proposed as an EMT-related lncRNA with putative prognostic value in OC. Similarly, its expression has been correlated with aggressive malignant phenotypes in HCC ^[137][164], lung cancer ^[138][165], and cholangiocarcinoma ^[139][166]. Accordingly, H19 prognostic value has been assessed in acute myeloid leukemia, CRC, and melanoma ^{[140][141][142]}[167,168,169]. Finally, beyond its diagnostic value, HOTTIP displays additional potential as a prognostic biomarker. Indeed, its expression has been associated with poor prognosis in multiple tumor types ^{[143][144][145][146][147]}[170,171,172,173,174].

4.3. EMT-Related lncRNAs as Therapeutic Targets in Cancer

The therapeutic potential of EMT-related lncRNAs represents another very promising scenario. Their abnormal expression and important roles in crucial cellular processes make them putative targets for cancer therapy. For example, BC-819, a double stranded DNA plasmid including the promoter H19 and the gene for diphtheria toxin A, has been investigated in clinical trials ^[148][149][175,176]. Mechanistically, H19 facilitates the expression of diphtheria toxin in cancer tissues, thus inhibiting tumor growth ^[150][177]. Of note, a phase 1/2a clinical trial evaluated the optimal dose, preliminary efficacy, and safety of BC-819 following intraperitoneal administration in patients with recurrent ovarian/peritoneal cancer. According to the results, this agent exhibits a good safety profile, but further research investigating the combination of intraperitoneal chemotherapy and BC-819 should be assessed in phase 2 and 3 trials ^[149][176]. Although some other lncRNAs are being investigated for their therapeutic potential ^{[150][151][152][153]}[177,178,179,180], such treatment applications in the everyday clinical setting still have a long way to go.