Graphical abstract

   graphic file with name fx1.jpg
   [29]Open in a new tab

Highlights

     * •
       Trophoblastic cells produce GDF15 which works through autocrine and
       paracrine modes
     * •
       GDF15 phosphorylates Smad1/5 in trophoblast cells through binding
       to its receptor
     * •
       Phosphorylated Smad1/5 activates target genes, such as SNAI1/2,
       E-Cadherin, Vimentin
     __________________________________________________________________

   Reproductive medicine; Molecular physiology; Developmental biology

Introduction

   The developing of trophoblast cells from implanting blastocyst is the
   key event in the establishment of pregnancy.[30]^1 Miscarriage usually
   refers spontaneous pregnancy loss which involves many pathological
   events including disorder of placentation. TGF-βs, GDFs, BMPs,
   activins, inhibins, and anti-Mullerian hormone (AMH) belong to the
   transforming growth factor (TGF-β) family. The TGF-β superfamily
   members exert divergent functions in regulating extravillous
   trophoblasts (EVTs) invasion, which contribute to the balance of
   maternal-fetal interface during the pregnancy.[31]^2 TGF-β1 was proved
   to inhibit trophoblast cell invasion by inducing Snail-mediated
   downregulation of VE-cadherin.[32]^3 GDF8 promotes trophoblast cell
   invasion by inducing ALK5-SMAD2/3-mediated MMP2 upregulation.[33]^4
   GDF11 stimulates trophoblast cell invasion by upregulated ID2/MMP2
   pathway signaling.[34]^5 BMP2 promotes human trophoblast cell invasion
   and endothelial-like tube formation through ID1-regulated IGFBP3
   expression.[35]^6 However, the mechanism of GDF15 in trophoblast cells
   invasion is not completely understood.

   GDF15 is widely distributed in mammal and plays many pivotal roles in
   the physiological status such as pregnancy, inflammation, cancer,
   cardiovascular disease and obesity.[36]^7 GDF15 is detected in the sera
   of pregnant women and the level rises substantially with progress of
   gestation trimester. GDF15 can also be detected both in amniotic fluid
   and placental extracts in high levels. According to these findings, the
   placental trophoblast is a major source of the GDF15 present in
   maternal serum and amniotic fluid.[37]^8^,[38]^9

   In the present study, we examine the expression of GDF15 in the
   trophoblast tissues and cells as well as its involvement in the
   pro-invasive effects on HTR-8/SVneo and JEG-3 cell lines. The decrease
   of GDF15 level in villus indicated the disorder of trophoblast invasion
   and a risk of miscarriage. Our results indicated that GDF15 induced the
   phosphorylation of Smad1/5, which subsequently trans-regulated target
   genes such as SNAI1/2, SERPINE1 and TIMP3 which promoted the invasion
   of trophoblast cells.

Results

GDF15 expression in trophoblast and serum of pregnant women with and without
miscarriage

   To explore the role of GDF15 in the trophoblast and serum of pregnant
   women, GDF15 protein expression was detected in the villus from
   curettage of women without medical reasons (Normal, N = 18) and women
   with spontaneous miscarriage (Miscarriage, N = 16) samples using
   immunohistochemistry ([39]Figure 1A). The immunoreactivity score (IRS)
   of GDF15 staining decreased from 8.22 ± 0.76 in normal samples to
   4.31 ± 0.37 in miscarriage samples ([40]Figure 1B, p < 0.001).
   Moreover, GDF15 staining was graded into three categories: negative,
   weekly positive and strongly positive. Negative GDF15 staining was
   found in 0.00% of normal tissue (0/18) and 18.8% in miscarriage tissue
   (3/16). Week positive GDF15 staining was found in 27.8% of normal
   tissue (5/18) and 68.8% of miscarriage tissue (11/16). The percentage
   of strong positive GDF15 staining in normal tissue (13/18) was 72.2%
   while it was 12.5% in miscarriage tissue (2/16). The rate of positive
   GDF15 staining in normal tissue was significantly higher than that in
   miscarriage tissue ([41]Figure 1C, p < 0.05). Consequently, GDF15
   expression was detected by western blotting in four normal samples and
   four miscarriage samples, the expression of GDF15 relative to GAPDH was
   significantly lower in miscarriage than in normal samples
   ([42]Figures 1D and 1E, p < 0.05). Furthermore, we randomly collected
   serum from pregnant women with a history of recurrent pregnancy loss
   during 4–10 gestational weeks. 36 serums from 15 patients who had a
   live birth while 40 serums from 20 patients who experienced miscarriage
   again in the follow-up. The levels of GDF15 in women with live births
   and with miscarriages were measured using ELISA and were shown in
   [43]Table S1. GDF15 concentrates in the serum of women who had live
   births were much higher than those who had miscarriage, especially
   during 6–10 weeks ([44]Figure 1F). The results of identification of
   trophoblastic cells and tissues using CK7 and HLA-G as markers are
   presented in [45]Figures S1A–S1C. Immunocytochemical assays and Western
   blot analyses in trophoblast cells revealed that GDF15 was abundant in
   the HTR-8/SVneo, JAR and JEG-3 cell lines ([46]Figures S1D–S1F). The
   overexpression and knockdown efficiency of GDF15 was shown in the
   HTR-8/SVneo and JEG-3 cell lines ([47]Figures S1G and S1H).

Figure 1.

   [48]Figure 1
   [49]Open in a new tab

   Expression of GDF15 in trophoblast tissues and serum of pregnant women

   (A) Immunohistochemical (IHC) detection of GDF15 in the villus from
   curettage of women without medical reasons (Normal) and women with
   spontaneous miscarriage (Miscarriage) samples; original magnification,
   400×.

   (B) The scatterplot presents the immunoreactivity scores obtained for
   GDF15 staining in Normal and Miscarriage samples.

   (C) The bar chart illustrates the percentage of negative, week positive
   and strong positive groups classified by GDF15 staining.

   (D and E) (D) The expression of GDF15 in the normal and miscarriage
   villus samples was detected by western blotting and the quantitative
   analysis were shown (E).

   (F) The serum concentrations of GDF15 in the pregnant women during
   4–10 weeks with miscarriage and with live birth were detected by ELISA.
   Data were presented as Violin diagram. ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗,
   p < 0.001 (unpaired two-tailed Student’s t test). See also [50]Table S1
   for further information.

The single-cell RNA-seq analysis of GDF15 expression in The Human Cell
Landscape (HCL) database

   We analyzed GDF15 expression in The Human Cell Landscape (HCL) database
   using single-cell RNA-seq ([51]bis.zju.edu.cn/HCL/). Over 700,000
   single cells from more than 50 human tissue were classified into 102
   major clusters in the HCL database. The Landscape provides global view
   on single cell level （[52]Figure S2A） and placental tissue clusters
   were highlighted ([53]Figure S2B). In the Gallery, GDF15 was selected
   from single-cell digital gene expression matrix and the results of
   clustering analysis were displayed in [54]Figure S2C. Sorting by GDF15
   expression in clusters and tissues, GDF15 was found to be abundantly
   expressed in the prostate, kidney, placenta tissues ([55]Figures S2D
   and S2E). Consistent results of scRNA-seq analysis from three different
   placental origins revealed that GDF15 was specifically found in
   extravillous trophoblast and syncytiotrophoblast cells ([56]Figures S2F
   and S2G: Placenta1; [57]Figures S2H and S2I: Placenta_Tsang;
   [58]Figures S1J and S1K: Placenta_VentoTormo).

A decreased in GDF15 levels contributes to fetal loss in mice in vivo

   Six placentas from pregnant mice with and without absorbed fetus were
   collected randomly using the abortion mouse model (CBA/J♀×DBA/2♂,
   [59]Figures 2A and 2B). GDF15 expression in the trophoblasts from mouse
   placenta was detected by immunochemistry ([60]Figures 2C and 2D). In
   comparison to the trophoblast with a normal fetus, the trophoblast with
   an absorbed fetus had a considerably lower GDF15 staining score
   ([61]Figure 2E). GDF15 transgenic mice (wild type: GDF15^+/+,
   heterozygote type: GDF15^+/−, homozygote type: GDF15^−/−) were used to
   assess the impact of GDF15 on embryo survival. 31 offspring with wild
   type, 62 with heterozygote type and 20 with homozygote type were born
   from 16 heterozygote pregnant mice (GDF15^+/−♀×GDF15^+/−♂) included.
   The expected and observed actual distribution of offspring gene types
   are shown in [62]Figure 2F. In order to compare fertility, three groups
   were made: wild type group (GDF15^+/+♀×GDF15^+/+♂), heterozygote group
   (GDF15^+/−♀×GDF15^+/−♂) and homozygote group (GDF15^−/−♀×GDF15^−/−♂).
   All female mice were housed with male mice for 15 days in each group.
   The distribution of fertile and infertile female mice was depicted in
   [63]Figure 2G. Both homozygote and heterozygote groups had higher rates
   of infertility than the wild type group(p = 0.0427). The period between
   cohabitation and delivery was not significantly different between the
   three groups ([64]Figure 2H). Notably, the embryo resorption rate in
   the homozygote group was remarkedly higher than that of the
   heterozygote and wild type groups ([65]Figure 2I). The findings
   indicated that GDF15 deficiency contributed to embryo resorption and
   fetal loss in mice.

Figure 2.

   [66]Figure 2
   [67]Open in a new tab

   The decreased level of GDF15 participates in the process of fetal loss
   in mice in vivo

   (A and B) (A) The uterus and placentas (B) with normal and absorbed
   fetus from pregnant CBA mice.

   (C and D) (C) Immunohistochemical staining for GDF15 are shown in the
   placentas and villus (D) with normal and absorbed fetus.

   (E) The quantitative analysis of GDF15 staining is shown in the violin
   plot.

   (F) The pie chart presents the excepted and observed percentages of
   genotypes of C57 transgenic mice offspring from heterozygous parents
   (GDF15 KO +/−); wild type: +/+, heterozygote: +/−; homozygote: −/−.

   (G) The bar chart illustrates the percentage of fertile and infertile
   mice in the three genotype groups after continuous cohabitation for
   15 days.

   (H and I) (H) The time to deliver and embryo resorption rate at day
   12.5 (I) in the three genotype groups. ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗,
   p < 0.001 (unpaired two-tailed Student’s t test).

GDF15 promotes the migration and invasion of human trophoblast cells in vitro

   A Transwell assay and a wounding-healing assay were applied to
   determine the effect of GDF15 on the motility of HTR8/Svneo and JEG-3
   cells in vitro. After 48 h of incubation in a Transwell assay, the
   numbers of HTR8-shGDF15 and JEG-3-shGDF15 cells invading across the
   Matrigel-coated membrane were significantly lower than those of
   HTR8-shCON and JEG-3-shCON respectively. Recombinant GDF15 protein
   (HTR8-shGDF15+rhGDF15 and JEG-3-shGDF15+rhGDF15) added to the cells
   rescued this decrease in invasion ([68]Figures 3A–3D, p < 0.001).
   Furthermore, the effect of GDF15 on the migratory capability of HTR8
   and JEG-3 cells was detected in a Transwell migratory assay by uncoated
   inserts. The number of HTR8-shGDF15 and JEG-3-shGDF15 cells migrating
   across the uncoated membrane was markedly lower than that of HTR8-shCON
   and JEG-3-shCON cells respectively. Consistently, the addition of
   rhGDF15 (HTR8-shGDF15+rhGDF15 and JEG-3-shGDF15+rhGDF15) to cells
   reversed this decrease in migration ([69]Figures 3A–3D, p < 0.001).
   Similarly, in the wounding-healing assay, HTR8-shGDF15 and
   JEG-3-shGDF15 cells significantly decreased would closure compared to
   HTR8-shCON and JEG-3-shCON cells after 48 h, while the decrease in
   wound closure was rescued with rhGDF15 in HTR8-shGDF15+rhGDF15 and
   JEG-3-shGDF15+rhGDF15 cells ([70]Figures 3E and 3F, p < 0.05). All of
   these results showed that GDF15 attenuation reduces the capacities of
   HTR8 and JEG-3 cells to invade and migrate in vitro.

Figure 3.

   [71]Figure 3
   [72]Open in a new tab

   GDF15 promotes the migration and invasion of human trophoblast cells
   in vitro

   (A and C) (A) The invasive and migratory potential of HTR8/Svneo and
   JEG-3(C) cells after silencing GDF15 and rhGDF15 rescue.

   (B and D) (B) Numbers of invasive and migratory (D) cells were shown as
   mean ± SD using triplicate measurements.

   (E) The migratory potential of HTR8/Svneo cells after silencing and
   rhGDF15 rescue was analyzed in a wound-healing assay for 0 and 48 h.

   (F) Scratch area was shown as mean ± SD using triplicate measurements.
   ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001 (unpaired two-tailed
   Student’s t test).

GDF15 activates the TGF-β pathway and regulates epithelial-mesenchymal
transition (EMT)-related genes in HTR8/SVneo cells

   A transcriptome sequencing analysis of HTR8-shGDF15 and HTR8-shCon cell
   lines was performed to investigate the mechanism by which GDF15
   promotes the motility of trophoblast cells. The differential expression
   genes including up-regulation and down-regulation were depicted as
   volcano map ([73]Figure 4A). Kyoto Encyclopedia of Genes and Genomes
   (KEGG) pathway enrichment analysis suggested that TGF-β signaling
   pathway involved in the mechanism of GDF15 in trophoblast cells
   ([74]Figure 4B). The expression of 15 genes regulated by TGF-β
   signaling pathway such as SERPINE1, TIMP3, ID1 and ID4 significantly
   differed between HTR8-shCon and HTR8-shGDF15 cells ([75]Figure 4C).
   Gene Ontology (GO) enrichment analysis identified 13
   epithelial-mesenchymal transition (EMT)-related genes, including
   SNAI1/2, MMP1, VIMENTIN and E-CADHERIN which exhibit the capacity of
   extracellular matrix, cell adhesion, cell migration, cell invasion, and
   whose expression remarkedly differed between the two groups
   ([76]Figure 4D). Furthermore, eight EMT-related genes and nine TGF-β
   signaling pathway-related genes were chosen for validation via
   real-time PCR assay. The mRNA of CDH1 (E-CADHERIN) was significantly
   increased, while levels of SNAI1/2, Vimentin, ZEB1, MMP1 and MMP2 were
   significantly decreased in the HTR8-shGDF15 compared to those in the
   control cells. The mRNA of ZEB2 was not altered in the two groups
   ([77]Figure 4E, p < 0.05). In addition, the mRNA of SERPINE1, TIMP3,
   TGM2, PMEPA1, HMGA1 and IGFBP5 were significantly lower in the
   HTR8-shGDF15 compared to those in the control cells, whereas the mRNA
   of LTBP1 and LTBP2 were not altered in the two groups ([78]Figure 4F,
   p < 0.05). β-actin was used as a reference gene in the above real-time
   PCR assay and related primers were listed in [79]Table S2. All these
   findings indicated that GDF15 participated the migration and invasion
   of trophoblast cells, possibly by regulating the TGF-β signaling
   pathway and EMT-related genes.

Figure 4.

   [80]Figure 4
   [81]Open in a new tab

   GDF15 activates the TGF-β pathway and regulates epithelial-mesenchymal
   transition (EMT)-related genes in HTR8/SVneo cells

   (A) The volcano plot illustrates the differential expressed genes
   between GDF15-silencing cells and their control cells in HRT8/Svneo
   cell line.

   (B) The bubble chart presents the main enrichment pathways through GO
   analysis of differential gene expression.

   (C and D) (C) The TGF-β pathway related genes and
   epithelial-mesenchymal transition (EMT)-related genes (D) expressed
   differentially shown by heatmap visualization.

   (E and F) (E) The mRNA levels of genes related EMT and TGF-β pathway(F)
   were validated by real-time PCR assay. mRNA was quantified using
   qRT-PCR and normalized to b-actin. ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗,
   p < 0.001 (unpaired two-tailed Student’s t test). See also [82]Table S2
   for further information.

GDF15 stimulates the TGF-β pathway by increasing of Smad1/5 phosphorylation

   To determine whether GDF15 stimulates the TGF-β pathway through the
   transcription effect of Smad2/3 or Smad1/5, overexpression and
   silencing of GDF15 by lentivirus in HTR8/SVneo and JEG-3 cell lines
   were used in immunocytochemistry and western blotting.
   Immunocytochemistry analysis revealed that rhGDF15 treatment induced
   nucleus translocation of Smad5 but not Smad2 and Smad3 compared to
   those in PBS group in HTR8/SVneo cells ([83]Figure 5A). Silencing GDF15
   with lentivirus in JEG-3 cells also resulted in significantly lower
   Smad5 levels in the nucleus with no discernible difference of Smad2/3
   compared to control cells ([84]Figure 5B). Furthermore, western
   blotting assay showed that overexpression of GDF15 in both HTR8/SVneo
   and JEG-3 cells increased the protein levels of phosphorylated Smad1/5,
   VIMENTIN, and SNAI2 while decreasing the protein levels of E-CADHERIN
   compared to those in the respective control cells. Moreover, silencing
   GDF15 decreased the protein levels of phosphorylated Smad1/5, VIMENTIN,
   SNAI2 while increasing the protein levels of E-CADHERIN compared to the
   respective control cells. However, the protein levels of Smad5, Smad2,
   phosphorylated Smad2, Smad3 and phosphorylated Smad3 were not affected
   by the GDF15 change ([85]Figures 5C–5E). Furthermore, Silencing Smad5
   with shRNA attenuated the migration and invasion of HTR8/Svneo cells
   compared to the respective control cells by a transwell assay
   ([86]Figures 5F and 5G). Western blotting revealed that VIMENTIN and
   SNAI2 protein levels were lower and E-CADHERIN levels were higher in
   HTR8/SVneo and JEG-3 Smad5-silencing cells than those in control cells
   ([87]Figures 5H and 5I).

Figure 5.

   [88]Figure 5
   [89]Open in a new tab

   GDF15 activates the TGF-β pathway by increasing phosphorylating of
   Smad1/5

   (A and B) (A) Immunocytochemistry for Smad2, Smad3 and Smad5 staining
   in HTR8/SVneo cell line after using PBS or rhGDF15 and JEG-3 cell
   line(B) after GDF15-silencing.

   (C–E) (C) The expression of Smad2, p-Smad2, Smad3, p-Smad3, Smad5,
   p-Smad1/5, E-cadherin, Vimentin, SNAI2 and GAPDH protein in the
   GDF15-overexpression and GDF15 silencing HTR8/Sneo and JEG-3 cell lines
   and their control cell lines were detected by Western Blotting and the
   quantitative analysis were shown (D and E).

   (F and G) (F) The invasive and migratory potential（G）of Smad5-silencing
   HTR8/Svneo cells and their control cells were analyzed by the transwell
   cell invasion and migration assays.

   (H and I) (H) The expression of Smad5, E-cadherin, Vimentin, SNAI2 and
   GAPDH protein in the Smad5 silencing HTR8/Sneo and JEG-3 cell lines and
   their control cell lines were detected by Western Blotting and the
   quantitative analysis were shown and normalized to GAPDH (I). ∗,
   p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001 (unpaired two-tailed Student’s t
   test).

   To assess GDF15’s transcriptional effect on target genes, a dual
   luciferase reporter system was constructed in PGL3.0 vector with
   promoters of SNAI1/2, CDH1(E-CADHERIN), SERPINE1, TIMP3 and LTBP1,
   which contained the motifs of SBE (Smad binding element：GTCTGNCN) or
   GC-rich sites of SNAI1/2, SERPINE1, TIMP3, LTBP1 and E-box（CAGGTG） of
   CDH1 ([90]Figures 6A and 6B). The luciferase activity of SNAI1/2,
   SERPINE1 and TIMP3 promoters was significantly higher while CDH1
   promoter was significantly lower in GDF15-overexpression HTR8/SVneo
   cells than that in the respective control cells ([91]Figure 6C).
   Moreover, the luciferase activity of SNAI1/2, SERPINE1 and TIMP3
   promoter was significantly higher while CDH1 promoter was significantly
   lower in HTR8/SVneo GDF15-silencing cells than that in the respective
   control cells ([92]Figure 6D). No change of luciferase activity of
   LTBP1 promoter was observed in the HTR8/SVneo GDF15-overexpresssion and
   silencing cells. All these results indicated that GDF15 accelerated the
   phosphorylation and nucleus translocation of Smad5 and regulating the
   target genes including SNAI1/2, CDH1, SERPINE1 and TIMP3.

Figure 6.

   [93]Figure 6
   [94]Open in a new tab

   The alteration of GDF15 to the transcriptional regulation of
   Smad5-related genes

   (A and B) The promoter structures of SNAI1/2, SERPINE1, TIMP3 and LTBP1
   containing Smad-binding elements (SBE) or GC-rich motifs and CDH1
   containing E-box were shown in the pattern diagram.

   (C and D) The luciferase activity relative to Renilla control was
   measured in GDF15-overexpression and GDF15-silencing HTR8/Svneo cells
   and the respective control cells. p < 0.05; ∗∗, p < 0.01; ∗∗∗,
   p < 0.001 (unpaired two-tailed Student’s t test). See also [95]Table S3
   for further information.

Correlation analysis between the expression of GDF15 and Smad5, SNAI1 and
SNAI2 in villus specimens

   To validate the correlation between GDF15 expression and TGF-β
   signaling pathway related proteins in villus specimens, IHC was used to
   detect the expression of GDF15, Smad5, SNAI1 and SNAI2 in 18 villus
   from artificial abortion without pathological factors(Normal) 16 villus
   from spontaneous miscarriage (Miscarriage) ([96]Figure 7A). Similar to
   GDF15 expression, the staining scores of Smad5, SNAI1 and SNAI2 in the
   villus of Miscarriage were significantly lower than those in the villus
   of Normal ([97]Figures 7B–7E). Notably, a Pearson correlation analysis
   revealed that GDF15 expression was positively correlated with Smad5,
   SNAI1 and SNAI2 expression in 34 villus samples (r^2 = 0.1377, p =
   0.0307; r^2 = 0.1339, p = 0.0333; r^2 = 0.2317, p = 0.0039
   respectively, [98]Figures 7F–7H). Therefore, those results indicated
   that GDF15 was linked to increased activity of the TGF-β/Smad5
   signaling pathway in human trophoblast tissue.

Figure 7.

   [99]Figure 7
   [100]Open in a new tab

   Correlation analysis between the expression of GDF15 and Smad5, SNAI1
   and SNAI2 in villus specimens

   (A–E) (A) 18 normal villus and 16 miscarriage villus samples were
   analyzed by IHC and the representative expression of GDF15, Smad5,
   SNAI1 and SNAI2 is shown and staining score of each sample was
   quantified(B–E). (unpaired two-tailed Student’s t test).

   (F–H) Correlation of GDF15 and SNAI1, SNAI2 and Smad5 expression
   respectively was analyzed in the villus samples by Pearson Chi-Square
   test. p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001.

Discussion

   Emerging evidence supports that TGF-β pathway plays a crucial role in
   trophoblast cell invasion and
   proliferation.[101]^2^,[102]^10^,[103]^11^,[104]^12 However, the
   precise role of GDF15 in the induction of invasion of trophoblast
   during pregnancy remains to be elucidated. Our results shed new light
   on the role of GDF15 in trophoblast invasion. Trophoblast cells,
   particularly syncytiotrophoblast were the primary source of GDF15 in
   our immunochemistry assay ([105]Figure 1A). Both HTR8/SVneo
   (trophoblast cell line) and JEG-3 (choriocarcinoma cell line) secreted
   GDF15 protein abundantly ([106]Figures S1A and S1B). GDF15 is produced
   by decidual stromal cells and trophoblasts in other
   studies.[107]^13^,[108]^14 The current study revealed that the
   expression of GDF15 in the trophoblast and serum significantly
   decreased in pregnant women with miscarriage. These results were
   consistent with the of Tong S’s early findings that GDF15 was a
   predictor of first-trimester miscarriage.[109]^15^,[110]^16
   Furthermore, GDF15 is reduced in the trophoblast tissue of the abortion
   mouse model and GDF15 deficiency is associated with embryo resorption
   in mice. A lack of GDF15 expression predicted a poor pregnancy
   prognostic in both human and mice. GDF15 could be a biomarker for
   pregnancy outcome in women who have had spontaneous abortions in the
   past, and our data showed that lower GDF15 levels was associated with
   risk of adverse pregnancy outcome. GDF15 has been validated as a
   predictor of miscarriage in pregnant women. In summary, GDF15 is likely
   to be an important serum biomarker for guiding pregnancy progression
   and prognosis evaluation.

   The invasive process of trophoblast following embryo implantation is
   essential for human placentation as well as pregnancy establishment and
   maintenance.[111]^17 Insufficient invasion of trophoblast cells of
   placenta into the maternal decidua directly results in a poor prognosis
   for pregnant patients.[112]^18^,[113]^19 In this study, transwell assay
   showed that silencing GDF15 reduced invasive and migratory ability in
   HTR8/SVneo and JEG-3 cell lines in vitro. Subsequently, we demonstrated
   that the addition of recombinant human GDF15 rescued cell invasion and
   migration in vitro. Using RNA sequencing and real-time PCR, we observed
   that trophoblast cells with GDF15 silencing expressed more E-CADHERIN
   and less Vimentin, SNAI1, SNAI2, ZEB1, MMP1 and MMP2. Silencing GDF15
   expression had also altered TGF-β pathway related genes expression,
   such as SERPINE1, TIMP3, TGM2, PMEPA1, HMGA1, and IGFBP5. These data
   support that GDF15 may be involved in invasion and migration of
   HTR8/SVneo cells through TGF-β pathway signaling. In addition, previous
   research has found a link between GDF15 and invasion in other
   cancers.[114]^20^,[115]^21^,[116]^22^,[117]^23^,[118]^24

   Next, the molecular mechanism of GDF15 in trophoblast invasion was
   explored. There were several signaling pathways responsible in
   regulating trophoblast invasion, including PI3K/Akt/GSK3β,[119]^25
   Notch,[120]^26 TGF-β[121]^27 and STAT3 pathways.[122]^28 As previous
   stated, GDF15 is a member of TGF-β superfamily, the roles of other
   members such as TGF-β1, BMPs, actins, and inhibins in trophoblast
   invasion have been discussed in detail in the review of Leung
   PCK.[123]^2 Our study revealed overexpression of GDF15 had the opposite
   effect to that of silencing, promoting phosphorylation of Smad1/5 and
   production of Snail1/2 and Vimentin, while inhibiting the expression of
   E-cadherin. The immunocytochemistry assay indicated that rhGDF15
   treatment increased Smad5 nucleus deposition, whereas silencing GDF15
   has the opposite effect. No effect of phosphorylation of Smad2 and
   Smad3 was observed after ectopic expression or silencing of GDF15. It
   has been reported that Smad1/5 impedes E-cadherin transcription, which
   results in the stabilization of Snail.[124]^29^,[125]^30 In our study,
   blockade of Smad5 by shRNA in HTR8/SVneo cells decreased invasion and
   migration, which was accompanied by Snail2 and Vimentin suppression. In
   summary, GDF15 promotes invasion and migration in trophoblast by
   increasing Snail1/2 and Vimentin expression through TGF-β/Smad1/5
   pathway. Mad homolog domain 1 (MH1 domain) of Smad5 is essential for
   specific DNA binding, such as SBE or GC-rich motif.[126]^31
   Consistently, overexpression and silencing of GDF15 in HTR8/SVneo cell
   altered the transcriptional activity of SNAI1/2, CDH1(E-Cadherin),
   SERPINE1 and TIMP3 via TGF-β/Smad1/5 pathway. Intriguingly,
   SERPINE1(PAI-1) and TIMP3 were reported to be transcriptionally
   activated by Smad2/3[127]^32^,[128]^33^,[129]^34^,[130]^35^,[131]^36
   and SERPINE1 could also be induced by TGF-β/Smad2/3 pathway.[132]^37
   Our study also suggests that SERPINE1 and TIMP3 are involved in
   GDF15-induced signal activation, implying that a crosstalk mechanism
   between GDF15 and non-Smad pathway exists. However, the exact mechanism
   has yet to be elucidated.

   E-cadherin could enhance Smad1/5 phosphorylation in intervertebral
   discs,[133]^38 whereas Smad5 also could downregulate E-cadherin
   expression and promote human glioma cells migration.[134]^39^,[135]^40
   Snail is a Smad-dependent transcriptional regulator,[136]^41^,[137]^42
   which was also verified to suppress the transcription level of
   E-cadherin.[138]^43 Here, we reported that GDF15 promoted the Smad5
   phosphorylation and nuclear translocation in trophoblast cells. Nuclear
   accumulation of Smad5 activated many transcriptional factors, including
   EMT-related genes. Expectedly, the levels of Vimentin and Snail2, two
   target genes of TGF-β pathway, were increased in response to GDF15
   ectopic expression, and this effect can be reversed by Smad5 depletion.
   Previous research indicated that Vimentin and Snail2 acts as a
   powerfully inducer of invasion by repressing E-cadherin transcription
   in trophoblast.[139]^44^,[140]^45^,[141]^46 In our study, an increment
   of Vimentin and Snail2 followed by GDF15 might retard E-cadherin in the
   GDF15 overexpression cells or rhGDF15 treatment. Collectively, our
   study demonstrated that GDF15 enhances invasion and migration of
   trophoblast cells by regulation of Vimentin and Snail expression as
   well as E-cadherin transcription through TGF-β/Smad1/5 signaling.

   In this study, we first demonstrated that GDF15 was a major enhancer of
   invasion in trophoblast cells using in vitro and in vivo assays as well
   as a group of pregnant women and animal models. Second, we showed that
   GDF15 could directly phosphorylate Smad1/5, activating TGF-β/Smad1/5
   signaling and increasing Vimentin and Snail2 expression while
   decreasing E-cadherin. Accumulated nuclear Smad5 induced the
   transcription of Snail1/2, ZEB1/2, Vimentin, MMP1 and MMP2 as well as
   the reduction of E-cadherin, which consequently enhancing trophoblast
   cell migration and invasion. Finally, we discovered that GDF15 was
   positively correlated with Smad5 and Snail1/2 expression in clinical
   villus samples. In summary, these findings revealed that GDF15
   stimulated invasion and migration in trophoblast. The critical
   signaling pathway involved in this process was identified, as were
   potential diagnostic and therapeutic targets in women with spontaneous
   miscarriage.

Limitations of the study

   A potential limitation of this study is that we have not been able to
   present EVT in the IHC results, because we only found a small portion
   of EVT shown in our IHC results from villus samples. However, we have
   showed that GDF15 was abundantly expressed in both syncytiotrophoblast
   and EVT. The experiments were conducted to confirm the function of
   GDF15 in HTR8 cell line from EVT source.

STAR★Methods

Key resources table

   REAGENT or RESOURCE SOURCE IDENTIFIER
   Antibodies
     __________________________________________________________________

   Rabbit polyclonal antibody to GDF15 Santa Cruze sc-66904
   Rabbit polyclonal antibody to Smad2 Cell Signaling #5339
   Rabbit polyclonal antibody to phospho Smad2(Ser465/Ser467) Cell
   Signaling #18338
   Rabbit polyclonal antibody to Smad3 Cell Signaling #9532
   Rabbit polyclonal antibody to phospho Smad3 (Ser423/425) Cell Signaling
   #9520
   Rabbit polyclonal antibody to Smad5 Cell Signaling #12534
   Rabbit polyclonal antibody to phospho Smad5(Ser463/465) Cell Signaling
   #9516
   Rabbit polyclonal antibody to SNAI1 Cell Signaling #3819
   Rabbit polyclonal antibody to SNAI2 Cell Signaling #9585
   Mouse monoclonal antibody VIMENTIN Santa Cruze sc-6260
   Mouse monoclonal antibody E-CADHERIN Santa Cruze sc-8426
   Mouse monoclonal antibody GAPDH Santa Cruze sc-47724
     __________________________________________________________________

   Bacterial and virus strains
     __________________________________________________________________

   shGDF15 lentivirus GeneChem biology company N/A
   shSmad5 vetors GeneChem biology company N/A
     __________________________________________________________________

   Chemicals, peptides, and recombinant proteins
     __________________________________________________________________

   rhGDF15 Raybiotech 228-12036-2
   TGF-β1 MCE HY-[142]P70543
   penicillin Invitrogen 15070063
   streptomycin Invitrogen 15070064
   Puromycin Sigma P8833
   Lipofetamine 2000 Invitrogen 11668019
   Matrigel BD 354234
     __________________________________________________________________

   Critical commercial assays
     __________________________________________________________________

   human GDF15 ELISA kit R&D system #DGD150
   Dual-luciferase Assay Kit Promega E1960
   RNAiso TaKaRa 9108
   PrimeScript RT Reagent Kit TaKaRa RR037A
   SYRB-Green fluorescence signal detection assay TaKaRa RR420L
     __________________________________________________________________

   Experimental models: Cell lines
     __________________________________________________________________

   HTR8/SVneo cells ATCC CRL-3271
   JAR cells ATCC HTB-36
   JEG-3 cells ATCC HTB -144
     __________________________________________________________________

   Experimental models: Organisms/strains
     __________________________________________________________________

   CBA/J Charles river N/A
   BALB/c Charles river N/A
   C57BL/6N GDF15+/- Cyagen N/A
     __________________________________________________________________

   Oligonucleotides
     __________________________________________________________________

   Primers for real time RCR, see [143]Table S2 This paper N/A
   Primers for LUC detection, see [144]Table S3 This paper N/A
     __________________________________________________________________

   Recombinant DNA
     __________________________________________________________________

   pIRES2-AcGFP1-Neo This paper N/A
   pIRES2-AcGFP1-Neo-GDF15 This paper N/A
     __________________________________________________________________

   Software and algorithms
     __________________________________________________________________

   Prism Graphpad 7.0 GraphPad Software [145]https://www.graphpad.com/
   ImageJ Schneider et al.[146]^7 [147]https://ImageJ.nih.gov/ij/
   [148]Open in a new tab

Resource availability

Lead contact

   Further information and requests for resources and reagents should be
   directed to and will be fulfilled by the Lead Contact, Shan Li
   ([149]lishan2.0521@stu.xjtu.edu.cn). Corresponding Email:
   [150]zpsheng@mail.xjtu.edu.cn and [151]lishan2.0521@stu.xjtu.edu.cn.

Materials availability

   GDF15 CRISPR/CAS9-based knockout C57 mice are available upon request.

Experimental model and study participant details

Human specimens

   A total of 18 villus samples from women who underwent induced abortions
   without medical reasons and 16 villus samples from women who
   experienced spontaneous abortion between 8 and 9 weeks of gestation
   were collect from the First Affiliated Hospital of Xi’an Jiaotong
   University from 2014 to 2016. A total of 79 serums from pregnant women
   had a live birth and 83 serums from pregnant women who spontaneous
   aborted were collected from 4 to 10 gestational weeks respectively. The
   study was approved by the institutional review board named as Ethics
   Committee of Medical School of Xi’an Jiaotong University in Shannxi,
   China. All the participates were informed the consent before
   collection.

Cell lines

   The human trophoblast cell lines HTR8/SVneo, JAR and JEG-3 was
   perchased from ATCC. The HTR8/SVneo cells were cultured in Dulbecco’s
   modifed Eagle’s medium (DMEM)/F-12 (supplemented with 10% fetal bovine
   serum (FBS)(BI) and 100unites/ml of penicillin and streptomycin
   (Invitrogen, Walthan, MA United States). JAR and JEG-3 was cultured in
   DMEN with 10%FBS at 37 in a 5% CO2 atmosphere. The medium was replaced
   every 2 to 3 days. Extrinsic rhGDF15（Raybiotech, Norcross, GA, United
   States, 50ng/ml） and TGF-β1 (MCE, United States, 10ng/ml) was purchased
   from recombinant proteins.

Spontaneous abortion animal model Construction and Transgene mice of GDF15

   Five CBA/J female mice and five BALB/c male mice were mated to
   constructed normal pregnancy model, five CBA/J female mice and five
   DBA/2 male mice were mated to constructed spontaneous abortion
   model[49]. The animals were obtained from (Vital River Laboratory
   Animal Technology Co., Ltd., Beijing, China). C57BL/6N wild type
   (GDF15+/+), GDF15 heterozygote (GDF15+/-) and GDF15 homozygote
   (GDF15-/-) were obtained from Cyagen (Guangzhou, China). At day 12.5,
   all female mice were sacrificed to examine embryo resorption. All
   experimental mice used were specific-pathogen free (SPF). This study
   was approved by the Institute of Animal Committee at the Medical
   College of Xi’an Jiaotong University. The male mice cohabited with the
   female ones (1:2) to test the quantity and the embryo resorption rate
   of their offspring. embryo resorption rate= absorbed fetus/（absorbed
   fetus+ normal fetus）.

Method details

The Human Cell Landscape (HCL) single-cell RNAseq database

   All data was achieved from the website([152]bis.zju.edu.cn/HCL/).
   Single cell gene expression and cluster analysis were conducted in
   single cell level in the landscape.

Lentivirus infection

   Overexpression plasmid and small RNA interference (siRNA) encapsulated
   by stable knockdown (shRNA) lentivirus were obtained from GeneChem
   biology company (Shanghai, China). Puromycin (Invitrogen, United
   States, 5ug/ml) was used to selected stable infected cells. RT-qRCR
   and/or Western blotting assays were applied to verified the infection
   efficiency.

Immunohistochemistry and immunocytochemistry

   The immunohistochemical (IHC) and immunocytochemistry assays were
   performed as previously described[47]. The primary antibodies were
   GDF15(1:100, Santa Cruze, USA), Smad2(1:100, CST, USA), Smad3(1:100,
   CST, USA), Smad5(1:100, WanLei, China), SNAI1(1:100, Santa Cruze, USA),
   SNAI2 (1:100, CST, USA), VIMENTIN (1:100, Santa Cruze, USA) and
   E-CADHERIN (1:100, Santa Cruze, USA). The GDF15 staining was divided
   into groups based on the positive percentage and staining density.
   Positive percentage scores were categorized into five grades:0-5% (0),
   6-25% (1), 26-50% (2), 51-75% (3) and 76-100% (4). The staining
   intensity was graded as follows: unstained (0), weak (1), moderate (2),
   strong (3). By multiplying the “staining intensity grade” by the
   “positive percentage score”, the immune-reactivity score (IRS) was
   calculated. The standard is as follows: ≤2 (negative），3-9 (week
   positive), 9-12 (strong positive).

Quantitative Polymerase Chain Reaction analysis and RNA sequence

   RNAiso (TaKaRa, Osaka, Japan) was used to isolate total RNA and the
   PrimeScript RT Reagent Kit (TaKaRa, Osaka, Japan) was used to
   synthesize cDNA. SYRB-Green fluorescence signal detection assay
   (TaKaRa, Osaka, Japan) was conducted for real-time RCR. The gene ACTB
   was used to normalize the results quantified by the 2-ΔΔCT method on an
   Gentier 96E instrument (TianLong, China). The RNA of HTR8/SVneo-shCon
   and HTR8/SVneo-shGDF15 was used to conduct RNA-seq in the BGISEQ-500
   platform. The data was analyzed using the NOISeq method with the
   threshold log2 fold-change >2.

Western blotting

   Western blotting analyses were performed as previously described using
   50μg lysate from fresh tissue and cells[48]. Protein was visualized
   using enhanced enhanced chemiluminescence detection system (Tanon
   5100). The primary antibodies were GDF15(1:1000, Santa Cruze, USA),
   Smad2(1:500, CST, USA), phosphorylated Smad2 (Ser465/Ser467, 1:500,
   CST, USA), Smad3(1:500, CST, USA), phosphorylated Smad3 (Ser423/425,
   1:500, CST, USA), Smad5 (1:1000, WanLei, China), phosphorylated
   Smad1/5(Ser463/465, 1:1000, CST, USA), VIMENTIN (1:1000, Santa Cruze,
   USA), E-CADHERIN (1:500, Santa Cruze, USA), SNAI2 (1:500, CST, USA) ,
   GAPDH(1:10000, Proteintech, China).

ELISA assay

   GDF15 levels in human serum were determined using a human GDF15 ELISA
   kit (#DGD150, R&D system, Minneapolis, MN, USA). The serum from
   peripheral blood was collected and diluted 10 times. The procedures
   were carried out in accordance with the manufacturer’s instructions for
   the commercial kit. The optical density of each well was measured at
   450nm and the concentrations of each sample were calculated using
   standard curves.

Transwell assays and scratch wound-healing assay

   To asssess trophoblast cells invasion, Matrigel (40ul, 1:8 dilution
   rate; Sigma, St Louis, MO) precoated transwell inserts (Corning,
   Cambridge, MA) with 8.0 nm diameter were used. Matrigel-free precoated
   transwell inserts were used to evaluate trophoblast cell migration. For
   the scratch wound assay, HTR8/Svneo cells/well (1×10^6) were seeded
   into a 6-well plate and incubated until confluent. A tip was used to
   scratch each well and the FBS-free medium was used to wash the detached
   cells at 0 hour and 48 hours respectively.

Luciferase reporter assay

   For the luciferase reporter plasmid, fragments of the SNAI1/2, CDH1,
   SERPINE1, TIMP3 and LTBP1 promoters were cloned into pGL3.0 Basic
   Vector (Promega， Madison， WI， USA). The primers were listed in the
   [153]Table S3. GDF15-modified cells were seeded into 24-well-plates and
   Lipofetamine 2000 (Invitrogen) was used to co-transfected with
   reconstructed plamids and pRL-TK plasmids. Dual-luciferase Assay Kit
   (Promega， Madison， WI， USA) was used to detect luciferase activities by
   luminometer (Promega).

Quantification and statistical analysis

   Each sample was applied at least three times. To compare the
   significance of difference, data was calculated using the Student’s
   t-test or the Chi-square test. The correlation was investigated using
   Pearson linear-regression analysis. Statistical analyses were performed
   with Prism Graphpad 7.0 (GraphPad Software, San Diego, California USA,
   [154]www.graphpad.com.). P value less than 0.05 were deemed
   statistically significant.

Acknowledgments