5HTT-/- mice were constructed as previously described8. Wild-type, 5-HTT heterozygote or knockout mice were allowed to reach adulthood at ambient pressure in Denver, Colorado (85 KPa; ambient pressure at sea level is 100 KPa). The breeding colony was maintained as 5HTT^+/- (heterozygote), so all mice used were littermates. Mice were genotyped by polymerase chain reaction using primers specific to the wild-type allele (ATATCCAATGGGTACTCCGCAG and TGGTGAATCTCAGCCACCAG) and the neomycin resistance gene used in the knock-out (TCGACGTTGTCACTGAAGCG and GGATACTTTCTCGGCAGGAGC).

Eight week old mice were anesthetized with ketamine 200 mg/kg and xylazine 10 mg/kg, the right jugular surgically exposed. Right ventricular (RV) pressure was measured using a 1.4 French Pressure Volume Conductance System SPR-839 (Millar Instruments, Houston, Texas), which was inserted into the RV through the surgically exposed right jugular vein. The hemodynamic phenotyping results were continuously recorded using a Millar MPVS-300 unit coupled to a Powerlab 8-SP A/D converter, acquired at 1000 Hz. All hemodynamic phenotyping results were captured to a Macintosh G4 computer utilizing Chart5.3 software. Heparin was injected into the circulation prior to sacrifice, the left atria removed, lungs flushed through the right ventricle, and lungs collected and flash frozen in liquid nitrogen. All animal studies were approved by the UCHSC intramural animal care and use committee.

All RNA for the gene arrays consisted of sex-matched pooled samples from three different animals, to reduce variability, with two Mouse Genome 430 2.0 arrays used for each condition (a total of six arrays from 18 pooled animals are thus reported). Samples were prepared for Affymetrix arrays using 2.5 μg of total lung RNA according to the Affymetrix GeneChip Expression Analysis Manual. All array results have been submitted to the NCBI gene expression and hybridization array data repository (GEO, http://www.ncbi.nlm.nih.gov/geo/), as series GSE11035.

Affymetrix cel files were imported into dChip 18, normalized to median overall signal intensity, and the perfect match/mismatch model used to determine per-gene hybridization strength. Pairwise comparisons between 5HTT wildtype and knockout mouse arrays were used to determine changed genes, with same-direction changes in all four pairwise comparisons needed for a gene to be considered changed. Gene ontology was determined using the Classify Genes tool within dChip18, with gene ontology files downloaded from the Gene Ontology Consortium http://www.geneontology.org19.

Analysis focused on gene ontology groups rather than highly changed individual genes for several reasons. First, different strains can use different genes to accomplish the same phenotypic end20; pathways are thus more important than individual genes. Next, arrays are actually rather poor at determining quantitative changes, as shown by results in this study. Finally, our goal was to determine overall pathways altered by loss of 5HTT, rather than individual genes; selection of a low required fold change while maintaining a low false-discovery rate is optimal for this process.

Western blots were performed as previously described21, using Klf4 rabbit polyclonal (1:500 dilution, CeMines, Golden, CO) and Kcne4 mouse monoclonal (1:300 dilution, Jackson ImmunoResearch Laboratories, West Grove PA). Equal protein loading was determined by Bradford assay; blots were stripped and reprobed with antibodies to β-actin to confirm this. A single strong band at the appropriate size was assumed to indicate specificity.

Primers were designed using sequences downloaded from Genbank with the Applied Biosystems ABI Primer Express program(AB, Foster City, CA). Each primer was searched against BLAST to ensure that it did not match any known gene aside from that for which it was designed (especially other family members). Upon receipt, each primer was tested and titered to ensure it gave a single, clean band of the appropriate size. Primer sequences are listed in Additional File 1. RNA was made from snap-frozen whole mouse lung using a Qiagen RNeasy mini kit (Valencia, CA), and first strand cDNA was created from 1 μg total RNA using a Qiagen QuantiTect^® Reverse Transcription Kit (Valencia, CA). Quantitative RT-PCR was carried out using SYBR Green chemistry in a 7300 Real Time PCR System (Applied Biosystems, Foster City, CA). Each measurement was made in triplicate and expressed relative to the detection of the housekeeping gene hypoxanthine guanine phosphoribosyl transferase (HPRT). Absence of contaminating genomic DNA was confirmed by PCR of no-RT control with cogenomic HPRT primers.

While 5HTT^-/- mice had never previously been found to have abnormal RVSP, we wished to verify this before proceeding. Mice at 8 weeks of age were weighed and anesthetized in preparation of closed-chested catheterization. Right ventricular systolic pressure (RSVP) and heart rate for each experimental mouse was obtained by inserting a catheter into the right ventricle through a surgically exposed right jugular vein. Under normoxic conditions, with ambient pressure and spontaneous ventilation, RVSP and heart rate were unchanged between 5-HTT^+/+, 5-HTT^+/- and 5-HTT^-/- mice (not shown).

RNA was extracted from homogenized whole lung from 5-HTT^+/+, 5-HTT^+/- and 5-HTT^-/- mice, with RNA pooled from 3 samples to reduce individual variability for each chip. Two chips, with pools from different animals, were used for wild-type, heterozygote, and homozygote 5-HTT knockouts; all animals were littermates. RNA was used to probe Affymetrix MOE430 v2 arrays. Affymetrix Cel files were further analyzed with dChip.

Using relatively low stringency for comparisons, including a minimum 1.2× fold change between any pair of samples with a minimum threshold for absolute change between any pair, we found 124 genes changed between from 5-HTT^+/+ and 5-HTT^-/- mice with a false discovery rate of 7%. Using the same criteria and swapping samples between groups reduced this to 6–10 genes, indicating that a large majority of the 124 genes likely represented genotype-dependent changes. These were sorted into gene ontology groups by dChip.

Half of these genes fell into one of two broad gene ontology groups; stress response and immune related genes, and genes related to muscle structure, vasoreactivity, or actin dynamics (Figure 1, Additional File 2). Other categories included cell cycle and apoptosis, developmental, energy metabolism, heme-oxygenes (HO-1), and matrix related genes (Additional file 3). Where a gene fell into more than one ontology group, for the purpose of these tables it was placed in the group for which references seemed to predominate. Changes in 5-HTT^+/- mice by array were similar to those in 5-HTT^-/- mice (Additional files 2, 3). Absolute changes shown are in arbitrary units of hybridization intensity.

To confirm results obtained by Affymetrix arrays, selected genes from different categories were analyzed by quantitative RT-PCR against cDNA derived from six individual mice per group. These included inflammation-related genes C/ebpβ and Il1β(Figure 2A), matrix related genes Mmp9 and S100a8 (Figure 2B), muscle related genes Klf4 and Kcne4, and the developmental gene Klf9 (Figure 2C). We found in each case that quantitative RT-PCR confirmed the change seen on the Affymetrix arrays, with two important differences (Additional file 4 and 5). First, in most cases the fold change was much greater by quantitative RT-PCR than by array, and second, there was an evident dose effect in six of seven genes by quantitative RT-PCR which was not apparent by array.

While most of the quantitative PCR results had very strong statistical values, scatter in values was greater and the corresponding statistical significance weaker for muscle developmental gene Klf4 and the voltage gated potassium channel Kcne4. Western blot analysis was therefore performed to confirm changes in protein level of these genes, with comparison to β-actin for protein loading. We found even protein loading with decreased Klf4 and Kcne4 protein across samples (Figure 3), with average decreases in protein levels of 3.5× and 2.0× for Klf4 and Kcne4 respectively by densitometry.