Transcriptional profiling of fetal hypothalamic TRH neurons

Enrichment of embryonic hypothalamic TRH neurons

To obtain information about the transcriptome of developing TRH expressing cells, we induced GFP expression in TRH neurons using transfected primary hypothalamic cultures derived from rat embryos of 17 days of gestation. This stage corresponds to the terminal phase of differentiation of the TRH phenotype in the hypothalamus [18]. TRH neurons were enriched by FACS. The transcriptome of the TRH neurons and hypothalamic cells was determined by DNA microarray technology (Figure 1A).

We have previously reported the conditions to efficiently transfect TRH neurons in serum-supplemented cultures; control experiments suggested that most GFP+ cells were TRH neurons [17]. Taking advantage of these conditions, we transfected E17 hypothalamic cultures with a GFP expression vector under the control of the minimal Trh promoter region (phrTRH-GFP) and determined the transfection efficiency by FACS. After 48 h of transfection, 0.4% of cells were GFP+ (Figure 1B, middle panel). Preparative cell sorting followed by FACS analysis of the GFP+ cell population demonstrated a strong enrichment with approximately 94% of cells being GFP+ (Figure 1B, lower panel). In general, cell viability was higher than 90% in all conditions examined as determined by propidium iodide (PI) staining (data not shown).

To corroborate the neuronal identity of the sorted GFP+ cell population, the expression of Trh together with cell-type specific markers was examined by RT-PCR assays. GFP+ cells were separated from the GFP- cells by FACS 48 h after transfection. As a control, a mixed cell population consisting of GFP+ and GFP- cells was obtained from sorted transfected cultures without selection (GFP+/-), whereas non-transfected cells (NT) (at 3 days in vitro) were used to establish the basal levels of mRNA expression. An increase in Trh mRNA levels was observed in the GFP+ cells compared with NT cells; this was also evident with respect to GFP+/- cells (Figure 1C). The increased Trh expression in the isolated GFP+ cells correlated with an increase in the expression of the neuronal marker tau, whereas the expression of the glial cell marker gfap diminished (Figure 1C). As expected, Gfp mRNA was present only in the GFP+ and GFP+/- samples (Figure 1C). These data, together with our previous report [17], indicate that the sorting conditions permitted the enrichment of TRH neurons from a mixed primary cell culture using GFP expression as a marker of Trh proximal promoter activity.

DNA microarray analysis of TRH neurons and hypothalamic cells

Once we corroborated that TRH cells were enriched in the GFP+ cell population, total RNA from GFP+ cells was isolated from a pool of six independent experiments. A pool from three other independent experiments was used to isolate total RNA from GFP+/- and NT cells. These RNA preparations were used to synthesize biotinylated cRNA targets and to screen U34A DNA microarrays.

In the array-screening experiment, two separate aliquots of GFP+ and GFP+/-RNAs, and a single one for NT RNA were used. Target generated from each aliquot was hybridized to an array, generating single (NT) and replicate (GFP+ and GFP+/-) datasets. Within the rat U34A oligonucleotide microarray, we distinguished genes based on whether or not they had a well-characterized gene name in the GenBank database. 7699 probe sets were known genes and 1130 probe sets did not have an assigned gene symbol (expressed sequence tags, ESTs); the total number of transcripts analyzed was 8829 (see http://www.ncbi.nlm.nih.gov/geo/ array).

Analysis of the signal intensity with the Microarray Suite (MAS) 5.0 software (Affymetrix) provides a statistical mean to determine the presence or absence of a gene in a sample. This test is based on the comparison of the hybridization efficiency of the target to its complementary sequence with the cross-hybridization of the target to a mismatch sequence identical to the complementary sequence except for one base. Each gene is represented by 16 probe pairs, with one of the members of each pair containing one mismatch, selected from distinct regions of the gene. The signal values from these probes were used to determine the presence of a gene in the target and a P value calculated from these data. A P value of less than 0.05 was used as a cutoff to consider that a transcript was present and a P value above this threshold indicated that it was absent. Transcripts were considered significantly present or absent based on the following criteria: a) the normalized value from mean differences was higher than the microarray background and, b) the fold change between match and mismatch signals was higher than 2.

On the basis of this analysis, the percentage of transcripts present in the GFP+ and GFP+/- populations was very similar. In the GFP+ cell population, 41% (2929 annotated transcripts and 583 ESTs) of the genes represented on the microarray were present, whereas 59% (4664 annotated transcripts and 398 ESTs) were absent (Additional file 1, Table S1). In the GFP+/- cell population, 41% of the genes (2941 annotated transcripts and 639 ESTs) were present and 59% (4756 annotated transcripts and 434 ESTs) were absent. In the NT cell population, 42% (3014 annotated transcripts and 571 ESTs) of the genes were present and 58% (4583 annotated transcripts and 410 ESTs) were absent.

Enriched transcripts in the TRH neurons

This analysis revealed that 30 transcripts were enriched in the GFP+ cells (Table 1). As expected, of the 30 transcripts enriched, we identified some transcripts previously associated with a neuronal phenotype, like the neurofilament heavy chain polypeptide (Nefh) [19], a voltage-dependent calcium channel [20], and a nicotinic alpha-receptor subunit [21]. Three of the enriched transcripts corresponded to novel transcripts within the developing hypothalamus, the Krüppel-like 4 transcription factor (Klf4), the TGFβ-inducible early growth response transcription factor (Klf10), also known as Tieg1, and the activator of transcription factor 3 (Atf3). In addition, a transcript up-regulated by vitamin D3 (Vdup1) was enriched in the TRH/GFP+ cell population, suggesting a potential physiological role of this vitamin within the hypothalamic TRH neurons, in agreement with previously reported data [22].

On the other hand, we found that some of the transcripts diminished in TRH/GFP+ cells (i.e. were probably not expressed in TRH cells) were associated with the glial cell phenotype. Among them are the collagen type I and type III (Col3a) [23, 24], and the follistatin-like gene, highly expressed in astroglial cells [25]. We also identified transcripts associated with cell cycle regulation, like annexin I, which negatively regulates cyclin D1 gene expression [26] (Additional file 1, Table S2).

We then decreased the microarray threshold to 1.5 fold change to determine if any missing classes of genes can be identified in the different cell populations. We used a heat map presentation and the gene expression profile to establish a hierarchical map based on the similarity of the gene expression values. The first scale, which is associated with a coloured strap, refers to genes with up-regulated (red) or down-regulated (green) expression levels in each cell population (Figure 2). The second scale represents the degree of regulation similarity among the genes. A value of zero indicates that the transcripts have the same regulation profile. Figure 2 shows part of the transcripts identified in each cell population. This analysis confirmed the enrichment of various transcripts (Klf4, Atf3, Nefh) in the GFP+ cells.

To validate the microarray data, we performed RT-PCR analyses for some of the transcripts presumably enriched in the GFP+ cell population. The levels of mRNAs for Nefh, Vdup1, and Klf4 were increased in the purified population when compared to the NT or GFP+/- cell populations (Figure 1D). On the contrary, the glial phenotype associated transcripts, Gafp and Col3a, were absent in the GFP+ cell population (Figure 1C and 1D). These results confirmed our microarray data.

Transcripts unique to the TRH neurons

To further breakdown the microarray data, a second method of analysis of the original signal intensities derived from the MAS software analysis was performed using a stringent P-value. This approach allowed us to identify transcripts present in the different cell populations with a high degree of certainty. Using a P value of < 0.001, we identified a total of 1864 and 1701 transcripts whose presence in the two GFP+/-replicates was significant. Similarly, we identified 1776 and 1714 transcripts in the replicate samples for the GFP+ cell population. In the NT cell population, we identified 1925 transcripts.

In order to identify the transcripts that were present in both replicates and to reduce the false positive rate in the detection of expressed transcripts, we defined the representative set of each sample as that containing transcripts significantly expressed in both replicates according to the P < 0.001 threshold. This resulted in 1600 [of which 288 were ESTs (Additional file 1, Table S4)] transcripts as representative of the GFP+ cell population (Figure 3 and Additional file 1, Table S3) and 1630 transcripts for the GFP+/- cell population. As shown in Figure 3, the overlap between the three cell populations indicates that 1361 transcripts were common to the three populations, whereas 112 transcripts were common to the GFP+ and GFP+/- cell populations but not expressed in the NT cell population. This comparison also shows that 51 transcripts were unique to the GFP+/- cell population, while 50 transcripts were unique to the GFP+ cell population at these thresholds. It should be noted that in this context unique transcripts refer to those transcripts that are uniquely detected in one or more populations shown in the Venn diagram, as they are likely to be expressed in undetectable levels at these thresholds in other compared populations. According to their GenBank annotations, several of these 50 transcripts (unique to GFP+ cells) are related to neuronal phenotype, e.g. synaptojanin 1 (Synj1); to translation machinery, e.g. eukaryotic translation initiation factor 3 subunit 9 (Eif3s9), ribosomal protein L27 (Rpl27); to basal metabolic machinery, e.g. acyl-CoA synthetase long-chain family member 5 (Acsl5), solute carrier family 37 member 4 (Slc37a4); to cell signaling, e.g. the serine-threonine kinase 38 (Stk38); in addition, transcripts encoding proteins with RNA-processing properties were also observed, i.e. the nuclear transcription factor Y gamma (Nfyc), the splicing factor arginine/serine-rich 10 (Sfrs10) and the Y box protein 1 (Ybx1). Transcripts related to CNS development were also identified, i.e. neurofilament heavy chain polypeptide (Nefh), and the nuclear factor I/B (Nifb). A transcript with chromatin remodeling properties, the transformation/transcription domain-associated protein (Trapp) was also identified (Table 2). These transcripts may play a critical role in the fetal development of hypothalamic TRH neurons.

Animals

Wistar rats raised at our animal facility, maintained in standard environmental conditions (lights on between 0700-1900 h, temperature 21 ± 2°C) with rat chow and tap water ad libitum. Animal care and protocols followed the guidelines for the use of animals in neuroscience research of the Society for Neuroscience, USA, and were approved by the Animal Care and Ethics Committee of the Instituto de Biotecnología, UNAM.

Cell culture and transfection

Hypothalamic primary cultures were prepared from E17 rat embryos as previously described [17]. Briefly, pregnant Wistar rats were anesthetized with pentobarbital (33 mg/kg b.w.) and the embryos removed individually. The hypothalamus was then excised through an imaginary line between both eye and ear superior edge; the dissected area was limited by the optic chiasm and lateral sulcus including mammillary bodies to a 2-3 mm depth avoiding thalamic area. Hypothalami were dissociated with trypsin (Sigma) and viability monitored by trypan blue exclusion (95%). Cells (5 × 10⁶) were plated onto poly-D-lysine (Sigma; 10 μ g/ml) pre-coated 60 mm Petri dishes in DMEM supplemented with 10% fetal bovine serum (FBS, GIBCO), 0.25% glucose (Sigma), 2 mM glutamine (Sigma), 3.3 mg/ml insulin (Sigma), 1% antibiotic-antimycotic (GIBCO) and 1% vitamin solution (GIBCO). Cultures were maintained in a REVCO incubator at 37°C in humidified air/7% CO₂.

Twenty-four hours after seeding, cells were transfected essentially as described [17]. In general, 8 mg of branched polyethylenimine (PEI) (600-1000 KDa; Fluka) solution was diluted in 10 ml of water, pH adjusted to 6.9 with 0.2 N HCl and the solution filtered (Millipore, 0.22 μ m). PEI (30 μ l) and plasmid DNA (10 μ g) were separately diluted to adjust NaCl to 150 mM in a final volume of 50 μ l, vortexed and incubated for 10 min at room temperature; subsequently, the polymer solution was added to the DNA, vortexed-mixed, incubated for 10 min at room temperature followed by the addition of 900 μ l of serum-free DMEM. The supplemented DMEM was removed from the culture dishes and the transfection mixture was added. After 3 hours incubation, transfection mix was removed and fresh supplemented DMEM was added. Forty-eight hours after transfection, cells were trypsinized (0.25% trypsin-EDTA) and subjected to FACS.

Plasmid construct

The minimal Trh promoter (-776/+84 bp) conferring tissue-specific expression [16] was excised with EcoR1 and BamH1/EcoRV digestion from the pNASS-rTRH-Luc expression vector. The Trh promoter fragment sticky ends were filled with Klenow DNA polymerase and subsequently sub-cloned into the SacI/BamH1 sites present on the pACT2 vector. Finally, the Trh promoter fragment was cloned into the SacI/BamHI sites in the phrGFP promoterless expression vector (Stratagene) and sequenced using primers corresponding to the Trh promoter 3' region (5'-ATG CAT AGA TCT TCT AGA TA-3') and to the phrGFP 5'region (3'-TGC AGG CCG GTG TTC TTC AGG A-5'); the resulting plasmid was named phrTRH-GFP.

Fluorescence-activated cell sorting

For preparative cell sorting, 5 × 10⁶ hypothalamic cells plated on 60-mm dishes were transfected as described above. After 48 h, cells were trypsinized, washed, resuspended in PBS/1% FBS (10⁷ cells/ml) and filtered through a 40 μ m nylon mesh (BD Bioscience, San Jose, USA). Cells were purified from a pool of five 60-mm dishes using the FACS Vantage (Becton Dickinson, San Jose, CA) and the exclusion method at high speed (60 μ l/min). Cells were sorted using the settings previously described [17] and analyzed by analytical flow cytometry as described below. In general, 20,000 GFP+ cells were purified from 5 × 10⁶ cells.

The percentage of GFP+ cells before and after purification by preparative cell sorting was determined by analytical flow cytometry using the FACS Vantage. All data acquisition and analyses were performed using the Cell Quest software (Becton Dickinson). To estimate the number of GFP+ cells, a FL1 histogram was generated (FL1, 530/30 nm short pass filter, detects green signal) and positive cells were defined as those cells in the region M1. The percentage of cells in M1 from the empty vector (mock)-transfected cells was subtracted from the percentage of plasmid-transfected cells in M1. Cell viability was determined using propidium iodide (PI) (Sigma) as reported [47]. PI fluorescence was detected with the FL2 emission channel (585/42 nm band pass filter). The percentage of dead cells was determined as the percentage of PI+ cells in a FL1 versus FL2 plot after subtracting the percentage of PI+ cells from the mock-transfected cells.

DNA microarray analysis

The microarray analysis was performed as described in the Affymetrix expression analysis technical manual http://www.affymetrix.com. Total RNA (10 μ g) was extracted from three different cell populations: i) sorted TRH-GFP+ cells (GFP+); ii) TRH-GFP+ and GFP- mixed cells (GFP+/-) passed through the FACS but not sorted, and iii) non-transfected cells (NT). To obtain a sufficient amount of RNA for each cell population, the number of independent experiments pooled for the GFP+ sample was higher than for the other samples. Therefore, a pool of six independent experiments was used to prepare total RNA from the GFP+ and three independent experiments for GFP+/- or NT cells.

The microarray target was synthesized from total RNA. RNA was reverse-transcribed into double-stranded cDNA with a T7 promoter-containing primer using SuperScript II reverse transcriptase, RNase H, and DNA polymerase (Invitrogen). After precipitation with 5M NH₄OAc and ethanol, the cDNA was used as a template in a biotin-labeled in vitro transcription reaction (Enzo BioArray, Affymetrix). Resulting target cRNA was collected on RNAeasy columns (QIAGEN, Valencia, CA) and then fragmented for hybridization on the microarrays.

The rat U34A microarray from Affymetrix was used. It contains probes for approximately 7699 well-annotated genes and around 1130 expressed sequence tags (ESTs) from Rattus norvegicus. Probes consist of 16 pairs of 25-mer oligonucleotides for each gene. One member of each pair contains a single base point mutation, and the signals of the pairs are compared to assess specificity of hybridization. Biotinylated target cRNA (15 μ g) was hybridized to the array and then processed using the Affymetrix GeneChip Fluidics Workstation 400. After binding with phycoerythrin-coupled avidin, microarrays were scanned on a Hewlett-Packard Gene Array Scanner (Hewlett-Packard Co., Palo Alto, CA). Data were deposited in the NCBI Gene Expression Omnibus repository http://www.ncbi.nlm.nih.gov/geo/ with the accession number GSE28441.

Results were analyzed using Affymetrix MAS 5.0 software. Individual microarrays were scaled to produce a mean signal intensity of 125. Iterative comparisons of different microarray datasets were done with MAS 5.0 comparison analysis as previously described with modifications [48]. To determine the expression difference between the GFP+ and GFP+/- cell populations, an additional approach was adopted. This involved grouping the two replicates of the control (i.e, GFP+/-) and the sorted sample (i.e, GFP+). Briefly, signal intensities of the two replicates of control and sorted datasets were averaged to represent the expression level of a transcript in the respective control and sorted populations. These averaged intensities were then used to calculate the fold enrichment in expression in sorted sample over control for each transcript. To identify transcripts that are enriched in one sample but underrepresented in the other, a threshold of more than 2 fold increase or decrease in expression was considered significant.

RT-PCR

RNA was extracted as previously described [18]. Reverse transcription (RT) was performed with 1 μg of RNA using the M-MLV reverse transcriptase (Invitrogen) in the presence of oligo-dT₁₅ primer. PCR was carried out in a total reaction volume of 50 μl. Primers (Additional file 1, Table S5) were designed using the PRIMER 3 software [49]. In general, PCRs were performed using 25 pmol of each of the specific forward and reverse primers, 1 μl of dNTP mix (20 mM each; Boehringer) and 1/5 of the RT reaction product.

Transcripts amplified by PCR included: Krüppel-like factor 4, collagen type III alpha 1 (Gene ID 84032), up-regulated by 1,25 dihydroxyvitamin D-3 (Gene ID 117514), neurofilament heavy chain[50], green fluorescent protein[51], Trh, glyceraldehyde 3-phosphate dehydrogenase (g3pdh)[18], Tau[52] and the glial fibrillary acidic protein[53]. Amplification was performed for 30 cycles except for g3pdh (24 cycles). PCR cycling conditions consisted of one cycle of melt temperature of 94°C for 1 min, a primer annealing step at 60°C or 64°C (Klf4) for 1 min, a polymerization step at 72°C for 1 min and a final extension at 72°C for 10 min. PCR products were electrophoresed in 2% agarose gel and bands stained with ethidium bromide.