Human Antigen D (HuD) Promotes Tumorigenesis And Invasion of Small Cell Lung Cancer Via Targeting lncRNA LYPLAL1-DT /miR-204-5P/Prolin 2 Axis

Background: Small cell lung cancer (SCLC) is one of the most malignant tumors with poor prognosis. RNA-binding protein (RBP) human antigen D (HuD) has been indicated in the process of tumorigenesis and progression of lung tumors, as well as long noncoding RNAs (lncRNA). However, the role of HuD and lncRNA in SCLC remains unknown. Methods: Realtime PCR were used to examine the circulating levels of LYPLAL1-DT in the 46 SCLC patients and 18 normal controls. Assays of dual- luciferase reporter system, RNA pull-down were performed to determine that LYPLAL1-DT could sponge miR-204-5p to upregulate the expression of PFN2. Migration and invasion assay, CCK8 and colony formation assay were used to detect the malignant effect of HuD and LYPLAL1-DT. Tumor xenograft model was established and IHC assay was performed to determine how HuD and LAPLAL1-DT impact in vivo. Results: We revealed that HuD was highly expressed in SCLC tissues and cell lines. HuD boosts the proliferation, migration, invasion of SCLC cells by increasing the PFN2 mRNA stability, which promotes cytoskeleton formation. HuD also enhanced the stability of lncRNA LYPLAL1-DT, which expressed highly in the serum of patients with SCLC and acted as an oncogenic lncRNA in SCLC cells as conrmed in vitro and in vivo. Mechanistically, LYPLAL1-DT functioned as a competing endogenous RNA (ceRNA) for sponging miR-204-5p, leading to the upregulation of its target PFN2 to promote SCLC cell proliferation and invasion. In summary, our data reveal a regulatory pathway in which HuD stabilizes PFN2 mRNA and LYPLAL1-DT, which in turn increases PFN2 expression by binding to miR-204-5p, and ultimately promotes tumorigenesis and invasion in SCLC. Conclusions: Our ndings reveal novel regulatory axes involving HuD/PFN2 and lncRNA LYPLAL1-DT/miR-204-5p/PFN2 HuD-overexpression; HuD-KD: HuD-knockdown; siPFN2: small interference targeting PFN2; PFN2-OE: HuD upregulates the PFN2 expression. A and B, Effect of HuD on lamellipodia formation was showed in HuD overexpression and negative control H446 cells. The number of cells with lamellipodia were calculated and plotted on a histogram. C, Cellular fraction analysis was performed to detect the expression of F-actin and G-actin. Histogram showed the ratio of F-actin and G-actin between HuD-OE and HuD-OC groups. D, Western blot analysis showed that PFN2 but not PFN1 was upregulated in HuD-OE H446 cells compared with HuD-OC cells. E and F, Transwell migration assay and invasion assay evaluated the migration and invasion abilities of siPFN2 or siNC on HuD-OE cells. G and H, Transwell assay accessed the migration and invasion abilities of PFN2-OE or PFN2-OC on HuD-KD cells. I, IHC detected the expression level of PFN2 in HuD-OC or HuD-OE tumor tissues in xenograft tumor mice model. J, Western blot showed the expression level of PFN2 in HuD-OC or HuD-OE tumor tissues of mice model. Three independent experiments were performed, and data were presented as mean ± SD. *, P < 0.05, **, P < 0.01, ***, P < 0.001.


Background
Small cell lung cancer (SCLC) accounts for 13%~15% of all lung cancers and rates the sixth most common cause of cancer-related mortality [33]. SCLC is characterized by rapid tumor growth, high vascularity, genomic instability, and early metastatic dissemination [23]. Over the last 30 years there has been a distinct paucity of signi cant breakthrough in SCLC therapy. The patient's survival level stays at 20% in the limited period and 2% in the extensive period [19]. Hence, we need to explore the intensive molecular mechanism involved in SCLC and identify novel molecular biomarkers and to unveil the mechanistic panoramic of SCLC.
RNA-binding protein (RBP) human antigen D (HuD), also known as embryonic lethal, abnormal vision like 4 (ELAVL4), is mainly expressed in differentiated neurons. Multiple studies have demonstrated that HuD controls neuronal commitment, plasticity and axonal growth [17]. It was revealed that apoptosis, differentiation of neurons and development of nervous system increased in HuD knockout (-/-) mice [1,17]. As an RNA-binding protein, HuD carries three RNA recognition motif (RRM) domains and binds to the AU-rich region of mRNA 3'-UTR, implicating its role in the regulation of mRNA stability, alternative splicing, alternative polyadenylation, RNA localization, and translation [16]. The association between HuD and SCLC was rst mentioned in 1994 that HuD is positively expressed in SCLC tissues and 18 SCLC cell lines but nearly undetectable in normal lung tissues [32]. Subsequently, it was reported that DNA vaccination against HuD antigen suppressed HuD-expressing tumor growth in a murine SCLC model [26]. Graus F et.al showed that the presence of Hu-Ab in the patients of SCLC is a strong and independent predictor of complete response to treatment [12]. More recently research showed approximately 10% of SCLC develop a paraneoplastic syndrome (PNS) due to a detectable serum autoantibody [16]. Our previous study showed that HuD antigen could act as a potential diagnostic criterium for SCLC [36]. All above evidences implicate that HuD may be a pivotal factor in SCLC. Unfortunately, the underlying mechanism of HuD pertained to the tumorigenesis and progression of SCLC has not seen any breakthrough over past 10 years.
Long noncoding RNAs (lncRNA) are a class of RNA molecules consisting of more than 200 nucleotides without protein-coding potential. Accumulating evidences indicate that lncRNAs constitute a regulatory system that functions at the transcriptional and posttranscriptional levels [13]. LncRNAs are involved in various biological processes, such as cell proliferation, differentiation, and metastasis [25]. Dysregulation of lncRNA has implication in the occurrence and development of tumors [15]. However, only a fraction of the lncRNAs in the genome have been deciphered and yet their functions in tumorigenesis, especially in SCLC, remains largely unknown [20]. Profound insight into the lncRNA-dependent gene-regulatory mechanisms will provide useful prognostic biomarkers and allow opportunity of developing precise therapy for SCLC. We recently screened lncRNA by RNA-sequencing and found lncRNA LYPLAL1-DT could protect endothelial cell from high glucose and in ammatory injury (data not published). However, the role of LYPLAL1-DT in high vascularity characterized SCLC is absolutely unknown.
Pro lin is a small actin-binding protein (12-17kDa) that promotes actin polymerization [21]. In mammals, four pro lin isoforms have been reported (pro lin1-4). Pro lin2 (PFN2) is tissue-speci c and mainly expressed in the brain whereas pro lin1 is universally present in all cell types and tissues [39]. Deletion of PFN2 led to increased neurotransmitter exocytosis and novel-seeking behavior in mice [28]. The neural cell adhesion molecule (NCAM) controls the proliferation and differentiation of neural progenitor cells by interacting with PFN2 [14]. A growing number of studies show that PFN2 is involved in the process of tumorigenesis. Kim et al. showed that PFN2 promoted migration, invasion and stemness of human colorectal cancer cells [18]. In non-small cell lung cancer, inhibition of PFN2 by miRNA-30a-5p can suppress epithelial-mesenchymal transition [40]. We previously reported that PFN2 is highly expressed in SCLC and could improve tumor growth and metastasis via exosome and angiogenesis [2]. Consequently, we have been curious about the mechanism of high expression of PFN2 in SCLC.
In the current study, we found that HuD could increase tumor growth and invasion by stabilizing PFN2 mRNA and a new lncRNA LYPLAL1-DT, the biological function of which has not yet been clari ed in tumorigenesis. As a competing endogenous RNA (ceRNA), LYPLAL1-DT sponges miR-204-5p to increase PFN2 expression, which boosts the malignant phenotype of SCLC. Our work elucidates HuD/LYPLAL1-DT/miR-204-5p/PFN2 axis in the tumorigenesis and progression of SCLC and provides prognostic indicators as well as a promising therapeutic target for SCLC patients.

Results
HuD contributes to migration, invasion and cell proliferation of SCLC in vitro and in vivo Studies have shown that HuD is 100% expressed in SCLC [7]. We detected the expression of HuD on 6 SCLC tissues and 5 normal lung tissues with immunohistochemical analysis. Data showed that the expression of HuD increased signi cantly in the SCLC tissues (6/6) compared with normal tissues (0/5) (Fig. 1A). To investigate the role of HuD in SCLC, we constructed stable HuD-overexpression (HuD-OE) and HuD-knockdown (HuD-KD) SCLC cell lines via lentivirus infection (Supplementary Fig. S1A and S1B).
Transwell assay revealed that the migration and invasion ability were signi cantly increased in HuD-OE cells and decreased in HuD-KD cells (Fig. 1B, C). RTCA showed that HuD overexpression could increase the proliferation of H446 cells whereas HuD knockdown had the opposite effect ( Fig. 1D and E). We also found that overexpression of HuD boosted a signi cant increasement in colony formation (Fig. 1F). The mice xenograft tumor model showed that HuD-OE signi cantly promoted tumor growth (Fig. 1G). Ki67 immunostaining showed that the uorescence intensity and Ki67-positive cells were both much higher in HuD-OE group (Fig. 1H, 1I and 1J). These ndings demonstrate that HuD initiates the growth and invasion of SCLC cancer cells.

HuD augments migration and invasion of SCLC by upregulating PFN2 expression
In eukaryotic cells, actin dynamics plays a critical role in cell invasion and morphological changes [29].
To determine whether HuD is able to mediate the regulation of actin dynamics, the soluble fraction (cytoplasm) and insoluble fraction (membrane and cytoskeleton) were isolated from HuD-OE and HuD-OC cells and the lamellipodia formation was detected. After scratching for 12 h, cell lamellipodia toward nude area were prominent in migrating HuD-OE cells while it was unobvious in control cells ( Fig. 2A). Statistical analysis also demonstrated cell numbers with lamellipodia were signi cantly increased in HuD-OE cells (Fig. 2B). Furthermore, F-actin, which is present mainly in insoluble fractions, was much higher in HuD-OE cells than in HuD-OC cells; whereas G-actin, which exists mostly in soluble fractions, was similar (Fig. 2C).
Pro lins are well-known actin-binding proteins and regulate the dynamics of actin, thereby playing a key role in vivo cell motility [9]. In the current study, we found that HuD promote PFN2 but not PFN1 expression in SCLC cell (Fig. 2D). Notably, by analyzing the dataset provide by George J et al., we found PFN2 expression exhibited a positive correlation with HuD expression in SCLC clinical samples (R = 0.315) [11]. We suspected that HuD may promote migration and invasion of SCLC through upregulating PFN2, then con rmed by small interference RNAs targeting PFN2 (siPFN2). We showed that the migration and invasion of HuD-OE cells were reduced markedly when cell were transfected with siPFN2 ( Fig. 2E and 2F). Furthermore, transient transfection of PFN2-OE plasmid into HuD-KD cells reversed the decreasing trend of migration and invasion of the cells (Fig. 2G and 2H). These data collectively indicate that HuD exerts its effect via upregulating PFN2. Furthermore, we transplanted HuD-OC cells or HuD-OE cells into SCID-NOD mice. IHC assay and western blot assay showed that PFN2 signi cantly increased in HuD-OE group ( Fig. 2I and 2J).
HuD promotes the stability of PFN2 mRNA by binding to its 3' -UTR HuD could directly and speci cally enhance the translation e ciency of mRNAs known to be involved in motor neuron differentiation and axonogenesis [35]. RIP assay was performed to demonstrated that HuD directly bound to the PFN2 mRNA (Fig. 3A). Additionally, when de novo synthesis of mRNA was blocked with actinomycin D, the decay rate of PFN2 was much slower in HuD-OE SCLC cells than in control cells (Fig. 3B). To clarify the binding site of HuD, we cloned the whole 3'UTR region and the fragments (S1, S2, S3) of PFN2 mRNA ( Fig. 3C and 3D). Luciferase reporter analysis showed that HuD overexpression signi cantly increased the luciferase activity of whole PFN2 3'-UTR and PFN2 S3, while PFN2 S1 and PFN2 S2 activities exhibited slightly change (Fig. 3E). Our results demonstrated that HuD could increase PFN2 mRNA stability and expression by directly binding to the distal part of 3'-UTR.
HuD stabilized lncRNA LYPLAL1-DT, which enhances the proliferation, migration and invasion of SCLC cells Deregulation of lncRNA plays a vital role in the occurrence and development of tumors, including small cell lung cancer [22,44]. Most of lncRNAs requires the interaction with one or more RNA-binding proteins With these in vitro ndings, we went on to determine the circulating levels of LYPLAL1-DT in the patients based on a collection of 46 serum samples from SCLC patients and 18 normal controls. We found that the level of LYPLAL1-DT was signi cantly higher in the SCLC group (Fig. 4H). To verify the effects of LYPLAL1-DT on growth, migration and invasion of SCLC, we rst detected the expression level in two common SCLC cell lines of H446 and H69, found that LYPLAL1-DT expressed higher in H69 than in H446 ( Fig. S3A). Then we developed a stably LYPLAL1-DT-overexpressing H446 cell line (Fig. 4I). CCK-8 assay showed that LYPLAL1-DT overexpression signi cantly increased the proliferation of H446 cells. However, knockdown of LYPLAL1-DT in H69 cell (Fig.S3C) signi cantly lowered the ability of proliferation of SCLC cell (Fig. 4J). Moreover, migration and invasion capacity were accessed by transwell, and wound-healing assay. The results revealed that LYPLAL1-DT overexpression dramatically boosted these malignant phenotypes (Fig. 4K, 4L and Fig.S3B).

LYPLAL1-DT acts as a competitive endogenous RNA with miR-204-5p to upregulate PFN2 expression
Recently, an increasing amount of evidence has demonstrated that lncRNAs function as ceRNA to sequester miRNAs, which decreases miRNA targets at the level of posttranscription regulation [41]. To determine whether LYPLAL1-DT works as a ceRNA, miRNAs potentially bound to it were predicted using the program of miRcode (mircode.org) and Starbase (http://starbase.sysu.edu.cn/). Among 4 potential miRNAs with high scores, miR-204-5p is the most signi cantly downregulated one in LYPLAL1-DT-OE cells (Fig. 5C, Fig. S4). Intriguingly, PFN2 is one of candidate regulatory targets miR-204-5p harbors (Fig.   5A), and the level of miR-204-5p decreased in the serum of SCLC patients compared with healthy volunteers (Fig. 5B). Consistently, the expression level of miR-204-5p was signi cantly reduced in LYPLAL1-DT-OE cells compared with LYPLAL1-DT-OC cells (Fig. 5C). We detected higher expression of PFN2 in the serum of SCLC patients and in LYPLAL1-DT-OE cells than their normal controls, respectively ( Fig. 5D and 5E). Nevertheless, miR-204-5p mimic transfection reversed the increased PFN2 expression in LYPLAL1-DT-OE cells (Fig. 5F). Furthermore, luciferase reporter assay showed that the luciferase activity of wild type LYPLAL1-DT or PFN2 was decreased by miR-204-5p mimics, while mutant LYPLAL1-DT or PFN2 showed no signi cant response ( Fig. 5G and 5H). The proliferation ability enhanced by LYPLAL1-DT overexpression could also reduce by either siPFN2 or miR-204-5p mimic transfection ( Fig. 5I and 5J). Further RIP assay indicated that LYPLAL1-DT and PFN2 could bind with the RISC formed by miR-204-5p and Ago2 (Fig. 5K). However, when knocking down LYPLAL1-DT in HuD-OE cells, we found the expression level of PFN2 was remarkably reduced, similar to HuD-OC + LYPLAL1-DT-OC cells ( Fig. 5L and 5M), which highlighted the key effect of LYPLAL1-DT in HuD-regulatory pathway. Collectively, our data veri ed that LYPLAL1-DT sponged miR-204-5p to upregulate PFN2 expression.

LYPLAL-DT promotes tumorigenesis and invasion of SCLC via miR-204-5p/PFN2 axis in vivo
To clarify the oncogenic function of LYPLAL1-DT in vivo, LYPLAL1-DT-OE cells and control cells were implanted subcutaneously into NOD/SCID mice. LYPLAL1-DT overexpression promote the formation of tumors in vivo with a signi cantly increase in both tumor size and weight (Fig. 6A). Ki67 staining shown that the proliferation of LYPLAL1-DT-OE cells was higher than controls ( Fig. 6B and 6C). To further de ne the LYPLAL-DT/miR-204-5p/PFN2 axis, we evaluated the expression levels of miR-204-5p and PFN2 in mice xenograft tumor tissues. As shown in Fig. 6D, 6E and 6F, miR-204-5p was reduced sharply while PFN2 upregulated remarkably. Immunohistochemical staining manifested that PFN2 levels were increased in LYPLAL1-DT-OE group than the controls (Fig. 6G).

Discussion
As one of the most malignant tumors, SCLC treatment faces big challenge due to its predisposition for early dissemination, quickly development and early metastasis, even lack of e cient therapeutic target [31]. Many studies have shown that SCLC express high-levels of HuD protein [4,38]. Some data demonstrated that alternation of Hu genes might be involved in the tumorigenesis and/or progression of neuroendocrine lung tumors [5]. Sasahira T et al. found that HuD was observed in 36.6% of oral squamous cell carcinoma and signi cantly associated with histological differentiation, nodal metastasis and mode of invasion, suggesting that HuD may play some role in tumor invasion or metastasis [30]. Ectopic overexpression of HuD inhibits MYCN RNA decay, thereby contributes to the malignant phenotype of neuroblastoma cells [24]. In this study, we rst revealed HuD promoted cellular proliferation, migration, and invasion, which indicating its oncogenic character. Further test discovered that HuD induced lamentous actin (F-actin) reorganization, which provided an essential condition for SCLC migration and invasion, for altering the dynamics of F-actin/G-actin turnover is a key step for cell migration and invasion. HuD belongs to an RNA binding protein family, ELAV-like family. ELAV-like proteins contain three ribonucleoprotein-2/ribonucleotide-1 ribonucleotide recognition motifs. By binding to both AU-rich element and the poly(A) tail, they are able to affect the processing of transcripts, alternative splicing, etc. [24]. Study have shown that HuR (ELAVL1) participates in the export of target mRNA to the cytoplasmic compartment, thus protecting them from degradation. Compared with other family members, HuD is highlighted its function in neural system. Studies showed that in vivo knockdown of HuD impaired learning performance and especially inhibited up-regulation of GAP-43 mRNA involved in synaptic plasticity remodeling [27]. HuD promotes neuronal differentiation of neural stem/progenitor cells (NSCs) in the adult subventricular zone by stabilizing special adenine-thymine (AT)-rich DNA-binding protein 1 mRNA [37]. Moreover, HuD can interact with circRNAs and regulate their expression and transport, which could control neuronal differentiation and synaptic plasticity [8]. Here, we de ned rstly that HuD could binding PFN2 distal part of 3'-UTR, then predicted and further con rm HuD stabilize AU-rich element of lncRNA LYPLAL1-DT. These data newly expanded and enrich the role of HuD as an RNA binding protein.
Although dozens of literatures have shown that lncRNAs play pivotal roles in tumorigenesis, the researches in SCLC are pretty limits. In the current study, we rstly found an absolutely new lncRNA LYPLAL1-DT which expressed highly in SCLC patient serum, could promote tumor growth and migration with gain-loss-of function by in vivo and in vitro test, strongly con rmed that LYPLAL1-DT would be a critical factor in SCLC development. A growing body of literatures regard that lncRNAs play the ceRNA role in tumor [3]. LncRNA PVT1 has been con rmed regulating gemcitabine resistance of pancreatic cancer by miR-619-5p/Pygo2 and miR-619-5p/ATG14 pathway [45]. A famous lncRNA HOTTIP was involved in SCLC tumorigenesis by sponging miR-574-5p and affecting the expression of EZH1 [34]. In the present study, we investigate whether LYPLAL1-DT can act as a ceRNA in SCLC. Firstly, we performed bioinformatic analysis and select lower expression of miR-204-5p in LYPLAL1-DT overexpression SCLC cell. Then luciferase assays and RNA immunoprecipitation assays with the Ago2 protein revealed that LYPLAL1-DT is engaged in complementary binding with miR-204-5p. MiR-204-5p is received concern just since 2020, and about 20 references reported it to be a tumor factor that can inhibit colorectal cancer cell growth and chemoresistance via exosome [43], promote apoptosis and decrease migration in gastric cancer [42], suppressed oral cancer cell aggressiveness [10], etc. However, no evidence has shown it function in SCLC up to now. Our current nding identi es miR-204-5p as the sponging target of LYPLAL1-DT in SCLC. Moreover, the miR-204-5p target gene PFN2 was con rmed in our previous study to activate Smad2/3 expression [2], and mediate endothelial cell migration via ERK pathway (data not published).
Furthermore, we demonstrated that miR-204-5p overexpression, PFN2 depletion signi cantly restored changes in LYPLAL1-DT overexpression. Taken together, our results indicated that the ectopic expression of LYPLAL1-DT was su cient to improve SCLC development as a ceRNA to regulate the miR-204-5p/PFN2 axis.
Emerging evidences showed that PFN2 is involved in the cancer progression. In ovarian cancer, PFN2 is activated in stem-A subgroup through non-canonical Wnt pathway and associated with poor clinical prognosis. In non-small cell lung cancer, PFN2 is overexpressed and suppresses the recruitment of HDAC1 to Smad2/Smad3 by preventing nuclear translocation of HDAC1, which leading to EMT of cancer cells. Our study demonstrated that signi cantly high expression of PFN2 could promote both migration and metastasis of SCLC [2]. In this study, we primarily revealed the mechanism of elevating PFN2 in SCLC by two branches and one trigger HuD. First, HuD could bind to the AU-rich region in the 3'-UTR of PFN2 mRNA and increased its half-life. Since mRNA half-life is a parameter re ecting the mRNA decay and translational processing, we propose that HuD can regulate the expression of PFN2 at posttranscriptional level. Second, PFN2 could be upregulated by LYPLAL1-DT with sponging miR-204-5p pathway which is still initiated by HuD binding and stabilizing LYPLAL1-DT. These results clarify the relationship between HuD and PFN2 via both directly stabilizing and LYPLAL1-DT/miR-204-5p pathway.
Some limitations still exist in our study. First, LYPLAL1-DT level has been just detected in patient serum, unfortunately did not con rm in SCLC tissue. Since the number of SCLC case is limit and it is di cult to obtained fresh SCLC tissue, especially the ones simultaneously matching with the serum sample. As a fallback solution, we con rmed highly expressed-HuD and PFN2 in the current study and our previous report, respectively. Second, LYPLAL1-DT is a kind of new lncRNA, lacking information in RNA database, plus short of SCLC RNA-sequencing database, we could not analyze the relationship of LYPLAL1-DT and clinical data of SCLC, such as overall survival rate, prognosis, etc. Last, HuD, LYPLAL1-DT, miR-204-5p or PFN2, which is/are the most e cient therapeutic target has not been explored, whereas limit our data application in clinical practice. It is also still nebulous in the upstream regulation factors of HuD. For ELAV-like proteins, it seems that various signaling pathways involved in their activation, such as MAPK signaling and PI3K signaling et al. All these questions inspire us to seek up further investigation and solution in the future.

Conclusion
Our work demonstrates that HuD could promote the proliferation and migration of SCLC through two axes of the HuD/PFN2, as well as HuD/LYPLAL1-DT/miR-204-5p/PFN2. It provides tangible evidence of HuD function in SCLC pathogenesis, and demonstrate the upstream of PFN2 which could be developed as e cient biomarkers of therapy and prognosis for the deadly disease.

Materials And Methods
Tissue specimens and serum samples Six SCLC tissues and ve normal lung tissues were obtained from Beijing Shi Ji Tan hospital. All samples were con rmed as SCLC by pathologic examination. A para n-embedded tissue specimen was available for each included patient. Forty-six serum samples from SCLC patients and 18 normal controls were collected from Beijing Chest Hospital. All patients were informed consent prior to the collection of specimens according to the institutional guidelines. Under the protocol approved by the Institutional Review Board, informed consents were obtained from the patients or their guardians.

Cell culture
The lung cancer cell lines NCI-H446 were obtained from Chinese National Infrastructure of Cell Line Resource. Cell lines were maintained in RPMI1640 supplemented with 10% FBS (PAN). Other cell lines, including 293T, were cultured in DMEM and supplemented with 10% FBS. All cell lines were cultured at 37℃ with 5% CO2.

Lentivirus package and stably transfected cell lines construction
The overexpression lentivirus vector of HuD (HuD-OE), the knockdown (HuD-KD), overexpression of LYPLAL1-DT and their negative controls (HuD-OC, HuD-KC, LYPLAL1-DT) were constructed. For constructing stable expressing cells, SCLC cell lines H446 were respectively infected with lentivirus vectors for 48h. Then, cells were selected with 2 μg/ml puromycin (Invivogen, Cat#ant-pr-1) for 3 weeks. Further, realtime RT-PCR was used to examine the overexpression and knockdown e ciency of these cells.

Transient transfection
The PFN2 expression vector pLV-PFN2 and control vector were purchased from Abcam company. For the knockdown of PFN2, siRNAs or negative control siRNA (Abcam) were transfected into the cells by using Lipofectamine 3000 (Invitrogen, CA, USA, Cat# L3000015). The miR-204-5p mimic and negative controls were purchased from HANBIO Ltd (HANBIO, Shanghai, China). They were transfected at a nal concentration of 50nM via an RNA t reagent (HANBIO, Shanghai, China), following the manufacturer's instruction.

Cell proliferation assay
The cell proliferation rate was assessed using a real-time cell analyzer (RTCA). Cells were suspended in the culture medium, 8000 cells per well of the E-plate with 4 replicates. After incubation at room temperature for 15 min, the E-plate was placed onto the RTCA Station in the incubator for continuous recording. Cell index values were recorded every 15 min for 72 h.

Western blot analysis
The protein samples were separated using sodium dodecylsulphate polyacrylamide gel electrophoresis and transferred onto poly(vinylidene uoride) membranes (Millipore, Cat#IPVH00010). The membranes were blocked with 5% skimmed milk (BD, Cat#232100) for 1h at room temperature and incubated with primary antibody overnight at 4 °C. The membranes were subsequently incubated with secondary antibody for 1h at room temperature. Protein signals were detected using enhanced chemiluminescence. The antibodies used in this study are listed in Table 1. Mouse IgG Proteintech, Cat#B900620 Monoclonal ANTI-FLAG SIGMA, Cat#F1804-50UG Quantitative PCR analysis Total RNA was extracted using TRIzol (Vazyme, Cat#R401-01) and reverse transcribed using PrimeScriptTM RT Kit (Abcam, Cat#G490). qRT-PCR was performed using EvaGreen 5×qPCR MasterMix (Abcam, Cat# MasterMix-5S). mRNA level was quanti ed by the 2-ΔΔCt algorithm with β-actin as the normalizer gene. All the primers used are found in Table 2.

Confocal immuno uorescence
Cells were seeded in glass bottom cell culture dishes (NEST, Cat#801001) at 2×10 5 cells/well, xed with 4% paraformaldehyde and permeabilized in 0.1% Triton X-100/PBS. Coverslips were blocked for 1h by 1% BSA, washed and exposed to a 1:100 dilution of HuD antibody (Proteintech, Cat#24992-1-AP) overnight at 4 °C. Then the cells were incubated with 1 mg/ml diluted Hochest for 5 min at room temperature. The cells were photographed using scanning confocal microscope.

Actin cytoskeleton staining
Actin cytoskeleton staining was performed after scratching using DyLight™ 594 Phalloidin (CST, Cat#12877) according to the manufacturer's instructions. The cells were stained with phalloidin and the nuclei were stained with Hoechst 33358. The cells were then photographed with uorescent microscope with at least 5 random elds along the scratch lines in each culture dish were taken. The number of cells with lamellopodia was counted and the percent of cell with lamellopodia was calculated via analyzing 40× micrographs.

Subcellular Isolation
The cell fractionation assay was performed according to the manufacturer's instructions (CST, Cat#9038). In brief, H446 cells were suspended in 0.5 ml of cold 1×PBS, aliquoted 100 µl of cell suspension for the whole cell lysate (WCL), sonicated and centrifuge for 5 min at 500 g. The pellet was resuspended in 500 μl of Membrane Isolation Buffer and centrifuged for 5 min at 8,000 g. The pellet was resuspended in 250 μl of Cytoskeleton/Nucleus Isolation Buffer.
Cell lysate was incubated with anti-HuD or Ago2 antibody (CST, USA) and IgG antibody at 4°C for 6h. A protein-RNA complex was captured and digested with 0.5 mg/ml proteinase K containing 0.1% sodium dodecyl sulphate (SDS) to extract RNA. The magnetic beads were repeatedly washed with RIP washing buffer to remove non-speci c adsorption as much as possible. Finally, the extracted RNA was subjected to mRNA level determination using qRT-PCR.
mRNA stability assay ActD (5 μg/ml, MCE, Cat#HY-17559), inhibiting the de novo RNA synthesis, was added to HuD-OE cells or its control cells. Total RNA was extracted at the indicated time points and PFN2 or expression was evaluated by qRT-PCR assay. The mRNA decay rate was calculated by comparing to the initial mRNA level before ActD addition.

Dual-luciferase reporter gene assay
The 3'-UTR of PFN2 mRNA and the mutant were cloned and inserted into pGL3 plasmid. Dual-luciferase reporter assay was performed following the manufacturer's instruction. The full length of the 3'-UTR of mRNA, the rst section of 3'-UTR (S1), the middle section of 3'-UTR (S2) and the last section of the 3'-UTR (S3) were cloned and inserted into the Xbal-Fsel restriction site downstream to re y luciferase (pGL3 promoter plasmid). 239T cells were seeded on 6-well plate and co-transfected with the constructed recombinant (2μg/well) or control plasmids and Renilla luciferase reporter (0.1μg/well) or HuD overexpression plasmids. After 48h, luciferase and Renilla luciferase activities were separately measured using the Dual-Luciferase Reporter Assay Kit following the manufacturer's instruction (Vazyme, Cat# DL101-01).

CCK-8 and colony formation assay
For CCK-8 assays, cells were seeded into 96-well plates at a density of 1,000 cells/well, and 10μL of CCK-8 (Vazyme, Nanjing, CHINA) was added per well in days 0 to 5. The cells were subsequently incubated at 37℃ for 2 hours, and the optical density was measured at 450nm. For the colony formation assays, 600 cells were inoculated into 6-well plates and cultured at 37℃ for 10 days. The colonies were the xed with methanol and stained with crystal violet.

Tumor xenograft model
Four-week-old NOD-SCID mice (Weitonglihua Biotechnology) were raised under speci c pathogen-free conditions. HuD-OC and HuD-OE-transfected cells or LYPLAL1-DT-OC and LYPLAL1-DT-OE were, respectively, injected subcutaneously into the ank region of the mice (5×10 6 cells/150 μL per ank, n = 5). The tumor volumes were measured every 7 days after inoculation. Eight weeks after injection, the mice were sacri ced, and the tumor nodules were harvested. The tumors, lungs, liver, and kidneys were isolated from the mice for further analysis.

Statistical analysis
Statistical analyses were performed using Graphpad Prism. Data were presented as mean ±SEM, and the statistical signi cance was determined by Student's t-test (t-test) or two-way ANOVA as indicated in the gure legend. Sample size (n) is also reported in the gure legend for each experiment, with n as the number of identically treated replicates. All tests were two-tailed, and P values < 0.05 were considered statistically signi cant.

Declarations Supplementary information
Supplementary information accompanies this paper as additional le : Fig. S1, S2, S3 and S4.
Ethics approval and consent to participate Availability of data and materials All remaining data are available within the article or available from the authors upon request.

Competing interests
The authors declare no con icts of interest.   HuD stabilizes PFN2 mRNA by directly binds to the end part of its 3'-UTR. A, RNA immunoprecipitation assay (RIP) indicated that HuD bound to the mRNA of PFN2. B, mRNA stability of PFN2 was measured in H446, HuD-OC and HuD-OE cells. C, Sequence of 3'-UTR of PFN2 mRNA showed the S1-S3 fragments and highlighted the AU-rich elements. D, The full length of 3'-UTR, fragment of S1, S2 or S3 of PFN2 were cloned to the downstream the re y luciferase reporter plasmids. E, Dual-luciferase reporter assay Page 24/29 indicated that the full length 3'-UTR and S3 fragment in PFN2 mRNA signi cantly increased the luciferase activity when HuD was co-expressed in 293T cells. Three independent experiments were performed, and data were presented as mean ± SD. *, P < 0.05, **, P < 0.01.

Figure 4
HuD stabilizes LYPLAL1-DT and induces it translocation, which enhances the proliferation, migration and invasion of SCLC cells. A and B, RT-PCR was used to compare the expression level of LYPLAL1-DT in the