Down-regulation of interferon tau gene transcription with a transcription factor, EOMES

IFNT, EOMES, CDX2, CREBBP, ETS2, or JUN Transcripts in Ovine Conceptuses

Transcripts for IFNT and the regulators of IFNT gene expression, including EOMES, CDX2, CREBBP, ETS2, and JUN, were assessed in ovine conceptuses by RT-PCR and real-time PCR, and are shown in Figure 1A,B, respectively. Expression of IFNT transcripts was relatively high in Days-15 and -17 conceptuses, but decreased on Day 21. mRNA expression of the T-box transcriptional factor EOMES in trophoblasts was lowest on Day 15, peaked on Day 17, and declined slightly to approximately 85% of maximum on Day 21. CDX2 mRNA was present at high levels in Days 15 and 17 trophoblasts, and decreased on Day 21. CREBBP mRNA gradually increased towards Day 21. While changes in IFNT, CDX2, CREBBP, and EOMES mRNA levels were seen, ETS2 and JUN mRNAs remained unchanged in Days-15, -17, and -21 ovine trophoblasts (Fig. 1A,B). Similar results were obtained from bovine conceptuses during the equivalent periods (Fig. 2A,B). In the bovine conceptus, EOMES mRNA level was lowest on Day 20 and highest on Day 22.

Negative Effects of EOMES on Ovine IFNT-Reporter Transactivation

To examine the effect of EOMES on ovine IFNT gene transcription, the ovine IFNT-reporter (−654 to +51 bp, wild type) construct with the pSG5 vector (Mock) or pSG5-Eomes was transfected into JEG3 cells (Fig. 1C). Luciferase activity decreased about 50% from the control (Mock) when pSG5-Eomes was present. The effect of EOMES was then examined in combination with Cdx2, Crebbp, ETS2, and Jun, which had previously been shown to be most effective in the transactivation of the ovine IFNT-reporter construct (Imakawa et al., 2006). EOMES reduced the effect of CREBBP, JUN, ETS2, and CDX2 on the up-regulation of ovine IFNT-reporter transactivation (Fig. 1D). Similar results were obtained from the bovine IFNT-reporter (−631 to +51 bp, wild type) construct (Fig. 2C,D). These findings indicate that EOMES could act negatively against IFNT transactivation in both species.

In deletion studies, EOMES exerted a negative effect on luciferase activity between −654 and −551 bases of the ovine IFNT construct (Fig. 3A). It has previously been demonstrated that there are JUN and CDX2 binding sites in this region of the ovine IFNT gene (Yamaguchi et al., 1999; Imakawa et al., 2006). Removal of this region or mutation of the AP-1 (JUN) binding site reduced transactivation in the bovine IFNT gene in a similar manner (unpublished observations). Luciferase activity was further decreased with deletions between −271 and −123 bases, but was unaffected by the presence of EOMES (Fig. 3A). In short, EOMES affected luciferase activity between −654 and −551 bases of the ovine IFNT gene, but exhibited no apparent effect when the serially deleted upstream region from −551 to −58 of the ovine IFNT gene was examined (Fig. 3A).

To further examine whether or not CDX2, CREBBP, ETS2, and/or JUN interaction with EOMES affected IFNT gene transcription, these transcription factors were evaluated in combination for effects in the ovine IFNT-reporter transactivation (Fig. 3B). Transfection with Cdx2, Crebbp, Jun, or ETS2 enhanced the luciferase activity of the wild-type ovine IFNT-reporter construct. When Eomes was co-transfected, however, the increase in luciferase activity brought forth by Crebbp, ETS2, or Jun was reduced (Fig. 3B). When Crebbp/Jun or Crebbp/ETS2 were co-transfected with Eomes, luciferase activity was also reduced (Fig. 3C).

Reduction of IFNT-Reporter Transactivation by EOMES Through Transcription Factor Interaction

It was demonstrated that CREBBP/JUN and CREBBP/ETS2 complexes bind to the upstream region of the ovine IFNT gene (Xu et al., 2003). To investigate the protein-protein interactions among EOMES, CREBBP, JUN, and ETS2, an immunoprecipitation study was conducted with lysates from JEG3 cells that had been transfected with Crebbp/Eomes, Crebbp/ETS2, or Crebbp/Jun expression plasmids (Fig. 4). When the cell lysates were initially immunoprecipitated with mouse monoclonal antibody against CREBBP and then blotted with rabbit polyclonal antibody against EOMES, ETS2, or JUN, bands corresponding to EOMES, ETS2, or JUN were detected (Fig. 4; #1, #2, and #3). Immunoprecipitation was then conducted with lysates from JEG3 cells that had been transfected with Crebbp, Eomes, ETS2, and Jun expression plasmids. EOMES, ETS2, and JUN were detected in the lysates from JEG3 cells that had been transfected with Crebbp, Eomes, ETS2, and Jun plasmids (Fig. 4; #4).

Coculture Experiments and Changes in IFNT, CDX2, and EOMES Transcripts

Next, we investigated how EOMES expression is regulated in trophoblast cells using the previously established coculture system of bovine trophoblast CT-1 cells and uterine endometrial cells (Sakurai et al., 2012). In monoculture, increased IFNT gene expression was observed after adding uterine flushing from Day-17 (P17) pregnant animals (Fig. 5). CDX2 transcripts were also increased after adding P17 uterine flushing (Fig. 5). Changes in EOMES mRNA, however, were not observed in CT-1 cells (Fig. 5). In coculture, decreased IFNT and CDX2 gene expression in CT-1 cells was detected when P17 uterine flushing was added (Fig. 5). Under these conditions, however, the same treatment resulted in four- to eight-fold increase in EOMES transcripts (Fig. 5). Regardless of culture conditions, no changes in IFNT, CDX2, or EOMES mRNA were seen when the uterine flushing from cyclic animals (C15) was used. Results from the coculture experiments with CT-1 and endometrial epithelial cells demonstrated that both the uterine flushing from pregnant animals and physical contact with endometrial cells was essential for the up-regulation of EOMES in CT-1 cells.

Changes in the Transcriptional Regulation of Bovine IFNT in Cocultured CT-1 Cells

To determine how EOMES reduces the degree of IFNT-reporter transactivation, changes in the binding of CREBBP or JUN to the bovine IFNT loci was examined by chromatin immunoprecipitation from CT-1 cocultures with uterine endometrial cells (Fig. 6). JUN and CREBBP bound to upstream regions of IFNT gene in monocultured CT-1 cells, but not in cocultured cells. The data showed that the binding of JUN or CREBBP to the upstream region of the IFNT gene in CT-1 cells disappeared after the attachment of CT-1 to uterine endometrial epithelial cells.

Animal Experimentation and Uterine Flushing Preparation

Whiteface crossbred ewes were maintained at the farm of the University of Tennessee, Knoxville, TN, USA, and the protocol for sheep experimentation had previously been reviewed and approved by the animal care committee at the University of Tennessee. Animal care, estrous synchronization procedures, and tissue collections were done as previously described (Sakurai et al., 2010). Hysterectomy was performed on designated days. After hysterectomy, uteri from Day-15, -17, or -21 pregnant animals (P15, P17, or P21, respectively, n = 3 each day) were flushed with 20 ml of sterile phosphate-buffered saline (PBS, pH 7.2), and conceptuses were removed from the medium. In addition to flushings from pregnant animals, the uterine flushings from Day 15 cyclic animals (n = 3) were collected. Recovered PBS (18–19 ml) was centrifuged at 4,000g for 5 min to remove cell debris, supernatants were filtered through 0.22 µm membrane, and then samples were stored at −80°C until use. After thawing, the samples (2 ml from 18 ml uterine flushings) were concentrated and desalinated through the use of Microcon filter device (Ultracel YM-3, Millipore, Billerica, MA). Protein concentrations were determined with the Bradford reagent (BioRad Laboratories, Hercules, CA), and the concentrations were adjusted to 1 µg/µl with distilled water.

Experimental procedures for Japanese black cattle were approved by the Committee for Experimental Animals at Zen-noh Embryo Transfer Center and the University of Tokyo. At the Embryo Transfer Center, conceptuses were collected non-surgically from superovulated Holstein heifers by uterine flushing with lactated Ringer's solution on Days 17, 20, or 22 (n = 3 each), as described previously (Ideta et al., 2009). Briefly, all animals were crossbred beef heifers (n = 9), and were fed the same feed; water was supplied ad libitum. These heifers between 13 and 15 months old were synchronized by insertion of a progesterone-releasing intravaginal device (PRID, ASKA Pharmaceutical Co., Ltd., Tokyo, Japan) placed into the vagina on randomly selected days of the estrous cycle for 9 days. Two days before removal of the PRID, prostaglandin F_2α (PGF_2α, 2 ml, Resipron-C containing 0.25 mg/ml Cloprostenol, ASKA Pharmaceutical) was administered to the heifers by intramuscular (i.m.) injection. The heifers were also given 500 IU equine chorionic gonadotropin by i.m. injection (Serotropin, ASKA Pharmaceutical) at the time of PRID removal. The superovulatory treatments began on Day 9 of the estrous cycle, and consisted of eight decreasing doses of follicle-stimulating hormone by i.m. injection (total of 28 Armour units, Antrin R-10, Kanagawa, Japan) for 4 days. All heifers were observed for standing estrus within 48 hr after the first PGF_2α treatment, and were randomly assigned to three groups.

Cell Propagation

Human choriocarcinoma JEG3 cells (HTB36, American Type Culture Collection) were grown in Dullbecco's modified Eagle's medium (DMEM, Wako, Tokyo, Japan) supplemented with 10% (v/v) fetal bovine serum (FBS; JRH Biosciences, Lenexa, KS) and antibiotic/antimycotic solution (Invitrogen, Carlsbad, CA) at 37°C in air with 5% CO₂.

Bovine trophoblast CT-1 cells, a generous gift from Dr. Alan Ealy, University of Florida, were derived from 10- to 11-day in vitro-produced blastocysts, and propagated as described previously (Talbot et al., 2000). In this study, CT-1 cells were cultured on plastic plates coated with Matrigel at 37°C in air with 5% CO₂ in DMEM containing 10% (v/v) FBS (JRH Biosciences) supplemented with 4.5 g/L D-glucose (Invitrogen, Carlsbad, CA), nonessential amino acids (Invitrogen), 2 mM glutamine (Invitrogen), 2 mM sodium pyruvate (Invitrogen), 55 µM β-mercaptoethanol (Invitrogen), and antibiotic/antimycotic solution (Invitrogen). Cells have since been propagated through more than 50 passages on Matrigel-coated plates, and continue to grow as a monolayer in distinct colonies while retaining characteristic features of trophectoderm cells.

Bovine endometrial epithelial cells were cultured in DMEM/F12 containing 10% (v/v) FBS, 40 units/mL of penicillin, and 40 µg/mL of streptomycin at 37°C in air with 5% CO₂, and propagated as described previously (Skarzynski et al., 2000; Sakurai et al., 2012). The cells were used within three passages to avoid changes in their epithelial cell characteristics.

RNA Extraction and RNA Analysis

Total RNA was extracted from Day-15, -17, and -21 ovine conceptuses; Day-17, -20, and -22 bovine conceptuses; and monocultured or cocultured CT-1 cells as described below with ISOGEN (Nippon Gene, Tokyo, Japan), according to the manufacturer's recommendations. For PCR analyses, isolated RNA (total 1 µg) was reverse-transcribed to cDNA using ReverTra Ace qPCR RT Kit (Toyobo, Tokyo, Japan) including 1× RT buffer, Enzyme Mix, and primer Mix in a 10 µl reaction volume, and the resulting cDNA (RT template) was stored at 4°C until use. The cDNA reaction mixture was diluted 1:10 using DNase- and RNase-free, molecular-biology grade water and 2 µl were taken for each amplification reaction. PCR was carried out with 0.5 units of ExTaq HS polymerase (Takara Biomedicals, Tokyo, Japan), 1× ExTaq HS buffer, 0.2 µM of the oligonucleotide primers (Table 1), and 0.2 mM of dNTP in a final volume of 20 µl. The thermal profile for PCR was at 94°C for 10 min, followed by 30 cycles of 94°C for 30 sec, 60°C for 30 sec, and 72°C for 30 sec. The products were analyzed by 1.5% agarose gel electrophoresis, and the bands were then visualized with ethidium bromide staining, followed by photography with ATTO ImageSaver (AE-6905C) under UV light. The expression levels of beta-actin (ACTB) were used to confirm equal loading (Sakurai et al., 2010).

Reverse-transcribed cDNA (2 µl) synthesized from ovine and bovine conceptus and cultured CT-1 cells RNAs (n = 3 from each day) were analyzed by quantitative real-time PCR with a 7900HT Fast Real-Time PCR System (Applied Biosystems, Tokyo, Japan). For each reverse-transcribed template, the PCR reaction was carried out with an ExTaq HS (Takara Biomedicals), SYBR Green I (Cambrex Bio Science, Rockland, ME), ROX Reference Dye (Invitrogen), and oligonucleotide primers for each target (Table 1). Amplification efficiencies of each target and the reference gene beta-actin (ACTB) were examined through their calibration curves and found to be comparable (Bustin et al., 2009). The PCR amplification consisted of 40 cycles at 95°C for 10 sec, annealing at 60°C for 20 sec, and extension at 72°C for 40 sec. The threshold cycle (Ct) value for each target was determined by Sequence Detection System software v1.2 (Applied Biosystems). Expression levels of each mRNA were normalized by calculating the Ct values based on subtracting the Ct value of targets mRNA from the Ct value of the internal control, ACTB mRNA. Each amplification ended with a melting curve analysis to confirm the specificity of amplification and absence of primer dimers (Sakurai et al., 2009).

Reporter Gene Constructs and Expression Plasmids

Preparation of 5′-deletion constructs representing various lengths of ovine and bovine IFNT promoter-enhancer regions (ovine TP-010, GenBank accession number: M88773 and bovine IFNTc1, GenBank accession number: AF238613, respectively) was performed as described previously (Yamaguchi et al., 1999; Xu et al., 2003). The nucleotide structure of a murine Jun expression vector, pRVSV Jun, has been described previously (Yamaguchi et al., 1999). A pRVSV plasmid, which contains only the Rous Sarcoma Virus (RSV) long terminal repeat (LTR) and a poly-(A) signal sequence, was used as the control for the determination of basal transcriptional activity. The human ETS2 expression vector was a pSG5-based construct containing the SV40 promoter and SV40 enhancer (Wakiya et al., 1996). The mouse Cdx2 and Eomes expression plasmids in pRC-cytomegalovirus (CMV), kindly provided by Drs. EunRan Suh and Steve Reiner, respectively (University of Pennsylvania School of Medicine), were subcloned into the pSG5 plasmid (Agilent Technologies Japan, Tokyo, Japan), and nucleotide structures were confirmed by DNA sequencing. The mammalian expression vector encoding the full-length mouse Crebbp cDNA (pRc/RSV-Crebbp), driven by RSV LTR, was a generous gift from Dr. D. Thanos, Columbia University (Merika et al., 1998). A pRc/RSV plasmid (The Institute of Physical & Chemical Research [RIKEN] Gene Bank, Tsukuba, Japan) that contains only RSV LTR and a poly-(A) signal was used as a control to determine the basal transcriptional activity. It should be noted that sequence similarity between transcription factor expression plasmids used in this study and those of the bovine were more than 80%.

Transient Transfection and Luciferase Assay

JEG3 cells were placed on 24-well plastic microplates, and grown to 50–60% confluence before transfection (Imakawa et al., 2006). Plasmid DNA was transfected into JEG3 cells with TransFast (cationic lipids, Promega, Madison, WI) according to the manufacturer's instructions. In brief, 0.8 µg of each DNA construct and 3 µl of TransFast were mixed in 200 µl DMEM. Amounts of total plasmids for each transfection were adjusted with the inclusion of pSG5 only (empty vector, Mock). After 10–15 min, plated cells were overlaid with the plasmid mixture and incubated at 37°C for 48 hr. To normalize transfection efficiency, pRL-TK vector (containing the herpes simplex virus thymidine kinase promoter and the Renilla luciferase gene, Promega) was transfected along with each pGL3 vector into JEG3 cells (Imakawa et al., 2006). The ratio of pGL3 to pRL-TK vector was 160:1 (800:5 ng).

After 48 hr of transfection, the cells were washed with PBS and lysed with 100 µl Passive Lysis Buffer (Promega) at room temperature for 20 min. Luciferase activity was determined by the Dual-Luciferase Reporter Assay System (Promega) on a Microlumat LB 96 P luminometer (Perkin–Elmer, Foster City, CA). Luciferase activity resulting from various ovine IFNT-reporter constructs was normalized through the determination of transcriptional efficiency with Renilla luciferase activity (Imakawa et al., 2006; Sakurai et al., 2009). Luciferase activity was expressed as fold-activation relative to an appropriate control within the experiment.

Co-Immunoprecipitation Analysis

JEG3 cells were plated in 10-cm plastic plates and grown to 70–80% confluence before transfection. Plasmid DNAs, Eomes/Crebbp, ETS2/Crebbp, Jun/Crebbp, or Eomes/Crebbp/ETS2/Jun, were transfected into JEG3 cells using HilyMax reagents (Dojin Chemicals) according to the procedures recommended by the manufacturer, as described previously (Imakawa et al., 2006), and cultured for 48 hr. JEG3 cells were suspended in 1 ml Nonidet P-40 (NP-40) buffer (NP-40:10 mM Tris–HCl, pH 7.4, 1% NP-40, 150 mM NaCl, 0.5 mM EDTA, and 0.5 mM NaF) containing protease inhibitors (2 mM PMSF, 25 µg/ml Aprotinin, 5 µg/ml Leupeptin, and 1 µg/ml Pepstatin A), and lysed on ice for 1 hr. The protein concentrations of these samples were determined with the Bradford reagent (BioRad Laboratories) and the concentration adjusted to 1 mg/ml. One ml (1 mg/ml) of each sample was passed through chromatin immunoprecipitation (ChIP)-Grade Protein G Magnetic Beads (Cell Signaling, Danvers, MA) at 4°C for 1 hr. To these samples, 2 µg of mouse monoclonal antibody against the carboxyl terminus of human CREBBP (sc-7300, Santa Cruz Biotechnology, Santa Cruz, CA) were added at 4°C for 2 hr. Normal mouse IgG (Sigma–Aldrich) was used as a negative control. After incubation with each antibody, these samples were incubated with 30 µl of ChIP-Grade Protein G Magnetic Beads for 1 hr. Protein–antibody-magnetic bead complexes were washed three times with NP-40 buffer, and then protein was diluted with Sample buffer. Diluted samples were subjected to Immunoblot analysis.

Immunoblot Analysis

Immunoprecipitates with anti-CREBBP antibody was separated on 10% reducing SDS–PAGE gels, and transferred onto PVDF membrane (Millipore, Bedford, MA). After blocking with 2.5% (w/v) skim milk and 2.5% (w/v) bovine serum albumin (BSA, Sigma) at 4°C overnight, the membranes were incubated with rabbit anti-mouse EOMES antibody (1:1,000, ab23345, Abcam, Tokyo, Japan), rabbit anti-human ETS2 (1:1,000, sc-351, Santa Cruz Biotechnology), or rabbit anti-mouse JUN (1:1,000, #9162, Cell Signaling) antibody at room temperature for 1 hr.

To determine EOMES and IFNT protein expression in mono- or cocultured CT-1 cells, cell lysates were prepared in RIPA Lysis buffer (50 mM Tris–HCl, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% SDS, 1 mM Na₃VO₄, 50 mM NaF). After incubation, the CT-1 cells were lysed with RIPA lysis buffer, and protein concentrations were determined with the Bio-Rad protein assay dye reagent concentrate (Bio-Rad laboratories, Hercules, CA). Cell lysates (10 µg) were loaded into each lane and separated through 10% SDS–polyacrylamide gel electrophoresis (SDS–PAGE), and were then transferred onto PVDF membrane (Millipore). After blocking them with 2.5% (w/v) skim milk and 2.5% (w/v) BSA overnight, PVDF membranes were treated with rabbit monoclonal anti-IFNT antibody (Operon, Tokyo, Japan), rabbit polyclonal anti-TBR2/EOMES antibody (Abcam, Tokyo, Japan), or rabbit anti-human nucleoporin P62 (internal control) (NUP62, BD Biosciences, Pharmingen, NJ) diluted 1:1,000 with 2.5% (w/v) skim milk and 2.5% (w/v) BSA. After washing, horseradish peroxidase-conjugated anti-rabbit IgG (1:10,000 for each antibody) was applied as the secondary antibody and the membranes were further incubated at room temperature for 1 hr. For visualization, ECL Plus Western Blotting Detection Reagents (GE Healthcare, Buckinghamshire, UK) were used, and signals were detected on X-ray film (Kodak, New York, NY).

Coculture CT-1 Cells and Endometrium Cells

Coculture assay was performed as described previously (Sakurai et al., 2012). In short, CT-1 monocultures on collagen type I-coated 6-well plates were incubated with 10 µg proteins from uterine flushing (C15 or P17) for 48 hr. In cocultures, endometrial cells were initially incubated on 6-well plates coated with collagen type I to reach confluence. CT-1 cells and 10 µg protein from uterine flushing (C15 or P17) were then plated over the endometrial cells for a 48 hr incubation period. After incubation, cells were washed twice with PBS and incubated in PBS for 5 min, and then CT-1 cells were separated from substratum of collagen type I-coated dish (monocultured) or adhering bovine endometrial epithelial cells (cocultured) with gentle pipetting of PBS. To collect cocultured CT-1 cells without contamination of uterine cells, the cells in PBS were passed through a 70-µm cell strainer (BD Biosciences), and the cells that remained on the membrane were purified further using a Percoll gradient. Percoll was diluted with DMEM to obtain dilutions of 12.5%, 25%, and 50%. The gradient column was prepared in a 15-ml Falcon tube by gently depositing a 1 ml layer of each solution, starting from a 50% dilution at the bottom. One ml of the CT-1 cell suspension in PBS was deposited at the top layer, and the gradient was centrifuged at 800g for 25 min. CT-1 cells were located in the Percoll layers of 12.5% and 25%, from which CT-1 cells were collected. Trypan blue exclusion tests confirmed cell viability over 95%. Mono- or cocultured CT-1 cells were then subjected to RNA isolation or transcription factor binding analysis through the use of chromatin immunoprecipitation (ChIP).

Chromatin Immunoprecipitation Assay

ChIP was performed as described previously (Sakurai et al., 2012). Briefly, CT-1 cells were cultured with or without endometrial epithelial cells (coculture or monoculture, respectively), and then cultured for 48 hr. Potential protein–DNA complexes in cultured CT-1 cells were cross-linked through the addition of 1% formaldehyde (Wako, Osaka, Japan) for 15 min. DNA from cells was sonicated to less than 500 base pairs. The supernatant of sonicated cells was diluted 10-fold, and 1% of the diluted lysates was used for total genomic DNA as input DNA control. Immunoprecipitation was performed overnight at 4°C with rabbit monoclonal anti-JUN antibody (10 µg; ab32447, Abcam) or rabbit polyclonal anti-human CREBBP antibody (ab2832, Abcam). Normal rabbit IgG (Santa Cruz Biotechnology) was used in place of specific antibody as a negative control. For isolation of DNA, 30 µl protein-G magnetic beads (Cell Signaling Technology) were added, and incubated for 30 min. Beads were then washed with High-salt or Low-salt buffer. Histone complexes were eluted from the antibody by freshly prepared elution buffer (1% SDS, 0.1 M NaHCO₃, 10 mM DTT). Protein–DNA cross-links, including the input DNA control samples, were incubated with 5 M NaCl and proteinase K (Invitrogen) at 65°C for 4 h. DNA fragments were extracted with a column (QIAquick PCR purification kit; Qiagen) following the protocol provided by the manufacturer, and resuspended in 60 µl elution buffer, from which 2 µl were used for PCR (35 cycles) with primers for the upstream region of the bovine IFNT (IFN tau-c1, Genbank accession number; AF238613) gene. The sequences of primers for the region from −877 to −644 bp are 5′-GTGAAGAGTTGACTCATTGG (sense) and 5′-TACACCTGTGGGCTTAGTTG (anti-sense), and for the region from −184 to −14 bp are 5′-TTGACAAACCCAAATTTTATTGG (sense) and 5′-GCAAGGCTTTTAAATAGGGAAGA (anti-sense). The PCR products were separated on a 1.5% agarose gel containing ethidium bromide, and were visualized under UV light.

Statistical Analysis

The real-time PCR data for gene expressions or luciferase assay data were analyzed by one-way ANOVA followed by Tukey's test for multiple comparisons against the control for each experimental group with the StatView statistical analysis software (version 5; SAS Institute Inc.). Differences of P < 0.05 were considered to be significant. In these analyses, treatment and day were considered independent variables, with replicate as the dependent variable.