Effects of thioredoxin reductase-1 deletion on embryogenesis and transcriptome

Abstract

Thioredoxin reductases (Txnrd)¹ maintain intracellular redox homeostasis in most organisms. Metazoans Txnrds also participate in signal transduction. Mouse embryos homozygous for a targeted null mutation of the txnrd1 gene, encoding the cytosolic thioredoxin reductase, were viable at embryonic day 8.5 (E8.5) but not at E9.5. Histology revealed that txnrd1^−/− cells were capable of proliferation and differentiation; however, mutant embryos were smaller than wild-type littermates and failed to gastrulate. In situ marker gene analyses indicated primitive streak mesoderm did not form. Microarray analyses on E7.5 txnrd^−/− and txnrd^+/+ littermates showed similar mRNA levels for peroxiredoxins, glutathione reductases, mitochondrial Txnrd2, and most markers of cell proliferation. Conversely, mRNAs encoding sulfiredoxin, IGF-binding protein 1, carbonyl reductase 3, glutamate cysteine ligase, glutathione S-transferases, and metallothioneins were more abundant in mutants. Many gene expression responses mirrored those in thioredoxin reductase 1-null yeast; however mice exhibited a novel response within the peroxiredoxin catalytic cycle. Thus, whereas yeast induce peroxiredoxin mRNAs in response to thioredoxin reductase disruption, mice induced sulfiredoxin mRNA. In summary, Txnrd1 was required for correct patterning of the early embryo and progression to later development. Conserved responses to Txnrd1 disruption likely allowed proliferation and limited differentiation of the mutant embryo cells.

Introduction

Thioredoxin reductase (Txnrd) enzymes are flavin-containing NADPH-dependent oxidoreductases that restore oxidized thioredoxin (Txn) to the reduced dithiol state [1–3]. Txn was first discovered as the provider of electrons to ribonucleotide reductase [4] and has subsequently been implicated in numerous enzymatic and regulatory processes [1, 5–7]. In mice, disruption of both copies of either the txn1 gene, encoding the ubiquitous cytosolic Txn1, or the txn2 gene, encoding the mitochondrial Txn2, results in early embryonic lethality [8, 9]. However, yeast and bacteria lacking Txn are viable as long as they retain an intact glutathione pathway [10–12].

Txnrd enzymes have been found in representatives of all phyla [1, 13, 14]. In eukaryotes, the txnrd1 and txnrd2 genes encode cytosolic and mitochondrial Txnrd enzymes, respectively. In mammals, a third gene, txnrd3, encodes a testis-specific isoform [15].

Metazoan Txnrd enzymes generally contain the unusual amino acid selenocysteine (Sec) [16], which, together with the cysteine in the conserved C-terminal sequence Gly-Cys-Sec-Gly, participates in redox activity [17–19]. Exceptions, for example in Drosophila, where two Cys residues, rather than a Cys and a Sec, are found at the C-terminal redox center [18], demonstrate that the Sec residue is not essential. Indeed, recombinant human Txnrd1 in which the Sec is replaced with Cys, retains some reductase activity [20]. Conversely, recombinant truncated human Txnrd that simply ends at the residue preceding Sec has none [20].

In addition to its general roles, Txnrd1-dependent reduction reactions are important in specific cells or situations for limiting signal transduction from several cell surface receptors, including the T cell receptor, the insulin receptor, and the epidermal growth factor receptor [21–27]. Txn1 is also required for DNA-binding activity of the transcription factors NFκB, AP1, steroid hormone receptors, and p53 [1, 28–30]. In yeast exposed to oxidants, dithiol-disulfide exchange reactions between Txn, glutathione peroxidase, and the AP-1 family transcription factor YAP1, results in oxidative activation of YAP1 and induction of downstream genes [31].

To study the functions of mammalian Txnrd1 in vivo, we generated a line of mice bearing a conditional-null txnrd1 allele, converted this to a null allele, and measured the effects on embryonic development of homozygous null embryos. Results showed the mutation was not acutely cytotoxic, as evidenced by the formation of embryos composed of several thousand cells. However, embryonic patterning and gastrulation were disrupted. Microarrays revealed gene expression responses that were similar to those previously observed in txnrd1-null yeast [32], indicating that related compensatory pathways are induced in these two evolutionarily distant organisms. Our results suggest a role for Txnrd1 in allowing correct developmental patterning and indicate that general cell physiological processes are not precluded by disruption of Txnrd1.

Materials and Methods

Mouse lines²,³

Isolation of genomic clone, construction of targeting vector, and production of the mouse lines for this study are described in Supplemental section. The targeting strategy and allele structures are shown in Fig. 1a². All renewable resources developed in this study, including phage, plasmids, and mouse lines, unless restricted by other parties, are freely available for unrestricted non-profit research use.

Fig. 1.

Design and confirmation of mice bearing txnrd1^cond and txnrd1⁻ alleles. a, schematic of alleles, targeting vector, and general mutagenesis strategy. Top, map of 5′ region of txnrd1 gene. Exons, represented as open boxes, are labeled below. Uppercase letters and asterisks indicate PCR primers. Probe and expected sizes of Southern blot bands are indicated. Below is the txnrd1-containing region of the targeting vector. Next is the correctly targeted (txnrd1^cond) allele, with diagnostic PCR products and Southern blot bands indicated. Below this is the Cre-recombined null (txnrd1⁻) allele. b, confirmation of allele was determined by genetic analysis of mouse lines developed from ES cells. PCR-based genotyping reactions (primers B, C, E, G combined) were performed on genomic DNA isolated from tail snips at weaning from offspring of the indicated mating. This primer set gives the following allele-specific product sizes: txnrd1⁺, 338 bp; txnrd1^cond, 542 bp; txnrd1⁻, 258 bp. Four pups had no txnrd1⁺ allele (pups 7 & 11 were txnrd1^cond/cond; pups 5 & 8 were txnrd1^−/cond), confirming that the downstream loxP was targeted into the txnrd1 allele. c, genotypic analysis of E7.5 embryos for microarray and RT-PCR analyses. Total nucleic acid was isolated from individual embryos and 10% of the total was used for radioactive PCR with primers B, D, and E. This primer set gives the following allele-specific product sizes: txnrd1⁺, 204 bp; txnrd1⁻, 258 bp.

Polysome analyses

Livers were harvested from txnrd1^−/+ adult male mice, homogenized, and subjected to sucrose gradient velocity sedimentation, fractionation, and RNA purification, as described previously [33]. An equal proportion (5%) of each fraction was used as template for an oligo(dT)-primed reverse transcriptase reaction and 1% of each cDNA was used as template in PCR reactions as described previously [34] using the primers described in the figure legend and in Table S1.

Embryo harvests and histology

Matings were set up between txnrd1^−/+ parents, plugs were checked each morning, and plug-date was defined as E0.5. For DAPI-staining, individual embryos were dissected into phosphate-buffered saline (PBS) containing 0.1% Tween-20 (PBST), fixed in buffered formalin, and stained with 0.01 mg/ml DAPI in PBST prior to evaluation by fluorescent microscopy. After taking photo-micrographs, each embryo was digested with 100μg/ml proteinase K (Sigma) in 1X TES (1% SDS, 10 mM Tris, pH 7.5, 5 mM EDTA) at 52°C overnight. DNA was precipitated with ethanol, resuspended in TE, and evaluated by radioactive PCR amplification using primers Txnrd1-B, -C, and -E (Table S1, Figs 1a, 1c), resolution on a sequencing gel, and autoradiography. For histology, whole pregnant uteri were harvested and fixed in buffered formalin. Uteri were embedded in paraffin and serial-sectioned longitudinally through the entire uterus. All sections were stained with hematoxylin and eosin and evaluated to ensure that differences between embryos were due to developmental progression and not to the plane-of-section.

In situ hybridizations

Detailed protocols are in Supplemental section; probes and methods were from the following sources: brachyury (t) [35]; Cripto [36]; Lim1 [37]; Fgf8 [38]; Snail1 [39].

Single-embryo cDNA production, RT-PCR, real-time PCR, and oligonucleotide array analyses

Pregnant uteri were harvested into ice-cold PBS; embryos were dissected and placed into individual 0.5 ml tubes containing 10 μl 1X TES (10 mM Tris, pH 7.5, 5 mM EDTA, 1% SDS) with 100 μg/ml proteinase K. Tubes were sealed and incubated 2–20 hours at 52°C. Samples received 90 μl of 0.1X TES, 5 μl of 5M NaCl, were extracted with an equal volume of phenol/choloroform/isoamyl alcohol (25:24:1) followed by pure chloroform, and were precipitated with 300 μl ethanol at −20°C overnight. Nucleic acid pellets were resuspended in 10 μl water. A portion (1–3 μl) was used for radioactive PCR-based genotype analyses, as above (Fig. 1c), and the remainder was used for cDNA production. Detailed methods for single embryo genotyping, cDNA production, and array analyses are in the Supplemental section.

Results

Design and production of txnrd1-mutant mice

The mouse txnrd1 gene encompasses 16 major exons, including a nontranslated first exon (“exon 0”) and 15 protein-coding exons, [40–42]. Txnrd1 catalytic activity requires a pair of cysteines near the N terminus of the protein (Cys⁵⁹ and Cys⁶⁴) and a C-terminal cysteine and selenocysteine, encoded in exon 15 [16]. Alternatively initiated or spliced Txnrd1 mRNAs have been described [41–43], in particular from testis [40], a tissue that exhibits particularly promiscuous patterns of transcription [44]. Some of these mRNAs could encode versions of Txnrd1 having additional N-terminal extensions; however all isoforms retain exons 1 and 2 [40]. The allele was designed to allow excision of these first two protein-coding exons (Fig. 1a), which encode conserved amino acids 1–30 that contact flavin-adenine dinucleotide (FAD), are an integral part of the Rossman fold common to NADPH-binding proteins [45, 46], and are found in all Txnrd1 mRNA isoforms [40]. Furthermore, when exons 1 and 2 are excised, the first in-frame AUG in the resultant mRNA (Met⁷⁰) is in a poor context for translation initiation and any polypeptide initiated here would lack both N-terminal active site cysteines (Cys⁵⁹ and Cys⁶⁴) (see below).

The txnrd1⁻ allele is functionally null

Animals homozygous for the conditional allele (txnrd1^cond/cond) survived and were phenotypically indistinguishable from wild-type littermates at all ages (Fig. 1b and not shown). The txnrd1⁻ allele lacks exons 1 and 2, but retains the promoter region and 5′ UTR-encoding exon 0 at a position roughly 15 kb upstream of the deletion mutation [41]. We predicted that pre-mRNA from the txnrd1⁻ allele would be spliced to join exons 0 and 3, creating an RNA lacking the native start codon and sequences encoding N-terminal FAD-binding and oxidoreductase active site amino acids (see above). RT-PCR analyses using total RNA harvested from adult heterozygous (txnrd1^−/+) kidney revealed only a single PCR product, which matched the size predicted for an mRNA in which exon 0 was spliced to exon 3 (not shown). Cloning and sequencing confirmed that this region of the cDNA contained an exon 0–3 splice structure (Fig. 2a). This cDNA contained two out-of-frame ATGs (reading frames of 13 and 8 codons) upstream of the first in-frame ATG (Met⁷⁰) and no in-frame ATGs in a favorable context for translation initiation [47]. Insertion of the mutant exon 0-3-4-5 cDNA region upstream and in-frame with green fluorescent protein in a cytomegalovirus promoter-driven expression vector inhibited accumulation of green fluorescent protein in transiently transfected cells (not shown). Polysome analyses on heterozygous liver cytoplasm preparations revealed that, while Txnrd1⁺ mRNA was almost entirely polysomal, Txnrd1⁻ RNA lacking exons 1–2 was not detected in the actively translated polysome fractions, but instead, was in the slowly sedimenting messenger ribonucleoprotein fractions (Fig. 2b). Thus, although RNA lacking exons 1–2 was produced from the mutated allele, this RNA likely was not translated. If residual translation occurred, the resulting predicted polypeptide would be incapable of catalyzing redox reactions. Thus, the txnrd1⁻ allele is functionally null.

Fig. 2.

The recombined txnrd1⁻ allele issues a non-translated mRNA. a, schematic. Predicted mRNAs from the wild-type and null alleles are shown. Diagonostic RT-PCR primer pairs and the sizes of expected products are indicated. RT-PCR-based cloning of Txnrd1 cDNAs from heterozygous mouse kidney RNA samples indicated that exon 0 spliced precisely to exon 3 (data not shown). The first two ATGs initiate short out-of-Txnrd1-frame ORFs that would issue predicted oligopeptides of 13 (-1 frame) and 8 (+1 frame) amino acids, respectively (first ATG generated by the exon 0/3 splice junction). The third ATG is in the Txnrd1 reading frame, at Met⁷⁰ in exon 3. Due to context (AAGCTGATGC), this ATG is predicted to not function as an initiator. b, translational efficiency of Txnrd1⁺ and Txnrd1⁻ mRNAs in heterozygous mouse liver. At top is shown an absorbance scan of a representative liver polysome gradient with the positions of ribonucleoprotein particles (RNP) and polysomes indicated above and the positions of gradient fractions indicated below. Below are RT-PCR analyses of the positions of Txnrd1 mRNAs in a heterozygous mouse liver polysome gradient using the PCR primer sets indicated in panel a.

Development of txnrd1^−/− embryos

At E7.5 and 8.5, the frequencies of txnrd1^−/− embryos from matings between txnrd1^−/+ parents matched that expected for a randomly segregating allele (Table 1). No txnrd1^−/− embryos were recovered at later times.

Table 1.

Genotype-specific survival of embryos from txnrd1^−/+ x txnrd1^−/+ matings¹.

stage	txnrd1+/+	txnrd1−/+	txnrd1−/−	n	litters
E7.5	16	27	18	61	5
E8.5	30	43	22	95	9
E9.5	8	13	0*	21	3
E10.5	10	14	0**	24	3

¹Data is only included from txnrd1^−/+ x txnrd1^−/+ matings. A larger analysis including matings in which one or both parents were txnrd1^−/cond also exhibited near-Mendelian survival of txnrd1−/− embryos until E8.5 with 100% loss of txnrd1−/− embryos by E9.5 (not shown). Where indicated, Chi squared analysis indicates genotype distribution differs significantly from that expected for a randomly segregating allele,

^*P<0.05;

^**P<0.01.

Embryos harvested from heterozygous matings at E7.5 were stained with DAPI, photographed using fluorescence microscopy, and then subjected to PCR-based genotypic analysis. Results showed that mutant embryos, although smaller than their wild-type and heterozygous littermates, contained several thousand cells, and that the internuclear-distance (an indicator of cell size) in txnrd1^−/− embryos was similar to that in wild-type littermates (not shown). Histological examination of whole pregnant uteri harvested at E7.5 and E8.5 from txnrd1^−/+ x txnrd1^−/+ matings revealed that mutant embryos had a strikingly defective morphology (Fig. 3). At E7.5, all embryos showed evidence of extensive cell-type differentiation. Mutants showed normal trophectoderm derivatives, including ectoplacental cone (EC), trophoblast giant cells (TGCs), and extraembryonic ectoderm. Primitive endoderm also formed; however the morphology of visceral endoderm (VE) differed from that seen in wild-type embryos. Normally, VE overlaying embryonic tissue (VE(em)) is thin and squamous; VE overlaying extraembyronic tissues (VE(ex)) is composed of cuboidal cells with apical vacuoles and a microvillous brush border (Fig. 3a) [48]. In txnrd1^−/− embryos, VE morphology was biased toward the morphology typical of the extraembryonic region (Fig. 3b, c, d). The embryonic ectoderm (epiblast) did not appear to differentiate and no morphologically distinct primitive streak was visible (Figs 3 b, c, d). However, the embryos continued to grow without further differentiation. By E8.5, mutant embryos were 3- to 8-fold larger than they were at E7.5, and the core of the embryo, which appeared to be undifferentiated ectoderm, expanded into a highly disorganized structure (Figs 3f, g, h). TGCs were abundant and the VE with extraembryonic morphology was further expanded, vacuolated, and folded upon itself into a thick layer (Fig. 3f, g, h).

Fig. 3.

Development of txnrd1^−/− embryos. Whole pregnant uteri were harvested from matings between txnrd1^−/+ parents and were prepared for histology. Longitudinal serial paraffin-embeded sections were prepared through the entire uterus and were stained with hematoxylin and eosin. Representative E7.5 (a-d) and E8.5 (e-h) embryo-central sections are shown. Abbreviations: EC, ectoderm; PE, parietal endoderm; PS, primitive streak; TGC, trophoblast giant cells; VE, visceral endoderm; VE(em), non-vacuolated squamous visceral endoderm overlying embryonic tissue; VE(ex), vacuolated cuboidal visceral endoderm overlying extraembryonic tissue.

To more accurately determine which embryonic tissues and structures were forming, we harvested E7.5 embryos and looked at marker gene expression patterns by in situ hybridization (Fig. 4). Brachyury (t) and Lim1, which are normally expressed in mesodermal regions of the primitive streak [35, 37], showed little or no expression in mutants. Snail1, which is normally expressed in the ectoplacental cone and pre-streak mesoderm [39] was only expressed in the ectoplacental cone in mutants (Fig. 4). Cripto, which is normally expressed by presumptive mesodermal cells in the epiblast of pre-streak embryos [36], showed a reduced expression domain, consistent with under-representation of embryonic tissues. Fgf8, which is normally expressed in the cells of the primitive streak and to a lesser extent in anterior endoderm, was observed in two reduced domains in the mutants, one corresponding to proximal embryonic ectoderm and one corresponding to the distal VE (Fig. 4). A similar pattern of Fgf8 expression is found in embryos having defective axis formation, including nodal-, lim1-, Otx2-, and Cripto-mutants [49, 50]. These results suggested that development was arrested in a pre-streak stage, mesoderm did not form, embryonic tissues were reduced, and patterning of the embryonic tissues was either arrested or incorrect, resulting in a failure to gastrulate.

Fig. 4.

Embryonic gene expression patterning. Embryos were harvested at E7.5 from matings between txnrd1^−/+ parents and stained by whole-mount in situ hybridization for the indicated marker genes. After staining and photographing, embryos were digested with proteinase K, nucleic acids were isolated, and genotypes were determined by radioactive PCR (see Fig. 1c). Abbreviations as Fig. 3 except: m, mesoderm; n, node; t, brachyury.

Single-embryo transcriptome profiles

Histological and marker analyses showed that the txnrd1 gene was not required for cell survival or proliferation but was required for embryonic patterning and perhaps to allow differentiation of some tissue types. To better describe the processes that were affected by the mutation, we developed a protocol to compare global mRNA levels (transcriptomes) in individual homozygous mutant and wild-type E7.5 littermates by microarray and RT-PCR (see Supplemental section).

Transcriptome data for three pair of embryos (six arrays) was analyzed statistically using GeneSpring software (Agilent). Signals for annotated probe sets that differed by ≥ 3-fold between mutant and wild-type embryos and whose reproducibility within the three embryo pairs yielded reliable data (P ≤ 0.05) were considered “differentially abundant”. By these criteria, 27 of the 45,101 probe-sets, representing 21 different mRNAs, had a stronger signal in the mutants (Table 2), and 47 probe sets, representing 35 mRNAs, had a weaker signal in the mutants (Table 3).

Table 2.

mRNAs more abundant in txnrd1^−/− than in txnrd1^+/+ embryos at E7.5.

mRNA1	difference2	−/− signal3	P-value4	description	GenBank
Igfbp1	11.7	567	0.04	IGF-binding protein 1	NM_008341
Npn3 (Srxn1)	10.9	474	0.00	Neoplastic progression 3 (sulfiredoxin 1)	BM210600
Gstμ3	8.7	338	0.01	glutathione S-transferase μ3	J03953
Mash2	8.5	3061	0.03	Mash-2, achaete-scute homologue-like 2	AK010738
Mash2	7.2	2267	0.04	Mash-2, achaete-scute homologue-like 2	AK010738
Gstμ1	6.9	752	0.00	glutathione S-transferase μ1	NM_010358
Gstμ1	6.5	1511	0.00	glutathione S-transferase μ1	NM_010358
Nxf7	6.4	240	0.01	nuclear RNA export factor 7	AJ305317
Gstμ1	5.4	292	0.01	glutathione S-transferase μ1	J03952
Mt1	4.9	4410	0.01	metallothionine 1	NM_013602
Cbr3	4.9	315	0.01	carbonyl reductase 3	AK003232
Gclm	4.4	1396	0.01	glutamate-cysteine ligase	NM_008129
Mt2	4.3	7407	0.02	metallothionine 2	AA796766
Ltb4dh	4.3	2330	0.00	leukotriene B4 12-dehydrogenase	BC014865
Mgst2	3.6	233	0.01	microsomal glutathione S-transferase 2	AV066880
S100A11	3.6	1144	0.02	S100 calcium-binding protein A11	BC021916
Ugt1a2	3.5	926	0.02	UDP glycosyltransferase 1 A6	D87867
Ugt1a2	3.5	663	0.01	UDP glycosyltransferase 1 A6	BC019434
Blvrb	3.5	360	0.00	biliverdin reductase B	BC027279
S100A1	3.4	1144	0.02	S100 calcium-binding protein A1	AI266795
Grina	3.4	445	0.01	glutamate receptor-associated 1	AW212189
Gaa	3.4	760	0.01	glucosidase, alpha, acid	BB357227
Mod1	3.3	569	0.00	malic enzyme, supernatent	BC011081
Gaa	3.2	533	0.00	glucosidase, alpha, acid	NM_008064
Mod1	3.2	366	0.03	malic enzyme, supernatent	AK006387
Htatip2	3.1	636	0.05	HIV-1 tat interactive protein 2	AF061972
Cln6	3.1	391	0.00	ceroid-lipofuscinosis, neuronal 6	BB514333

¹mRNAs having multiple entries were represented by more than one oligonucleotide probe set on the arrays and thus provided independent data readouts. In some cases, probe sets were designed from different GenBank files; in other cases they were designed to a common GenBank file, as indicated in right column.

²Only mRNAs having an average abundance difference of three-fold or greater between txnrd1^−/− and txnrd1^+/+ embryos were included.

³Average signal from all three replicates given; only mRNAs in which the higher signal was ≥ 200 units were included.

⁴All data based on statistical analysis of individual arrays from three wild-type and mutant littermate pairs using GeneSpring; only mRNAs showing sufficient signal reproducibility between the three paired array sets to yield a P-value ≤ 0.05 were included.

Table 3.

mRNAs less abundant in txnrd1^−/− than in txnrd1^+/+ embryos at E7.5.

mRNA1	difference	+/+ signal	P-value	description	GenBank
Nkx1-2	−21.3	639	0.04	Nk1 transcription factor-related 2	NM_009123
Agtlr1	−20.1	823	0.03	angiotensin receptor-like 1	BB533323
Pcdh8	−15.4	293	0.02	protocadherin 8	BB076893
Sox11	−14.7	1580	0.01	SRY-box containing gene 11	BB656631
Pmp22	−14.1	743	0.02	peripheral myelin protein 22	NM_008885
Thsd2	−12.5	496	0.03	thrombospondin, type 1, domain 2	BG072958
Ms4a4d	−12.5	322	0.04	membrane-spanning 4, domain A4D	NM_025658
Ebaf	−11.5	257	0.04	endometrial bleeding associated factor	AV214969
Ccnd2	−11.4	944	0.03	cyclin D2	NM_009829
Ccnd2	−10.9	1217	0.01	cyclin D2	BQ175880
Pcdh18	−9.5	303	0.02	protocadherin 18	BM218630
Ccnd2	−8.2	383	0.02	cyclin D2	AK007904
Myl7	−7.8	246	0.00	myosin, light polypeptide 7	NM_022879
Ccnd2	−7.7	285	0.01	cyclin D2	NM_009829
Pbx1	−7.6	770	0.02	pre B-cell leukemia factor 1	BG070361
Fgf13	−7.5	367	0.04	fibroblast growth factor 13	AF020737
Sox11	−7.2	329	0.04	SRY-box containing gene 11	BM508495
Dll1	−6.3	340	0.01	delta-like 1	NM_007865
Gli3	−6.1	478	0.04	GLI-Kruppel family member 3	BB311687
Frzb	−6.1	388	0.01	frizzled-related protein	U91905
Lef1	−6.0	409	0.04	lymphoid enhancer binding factor 1	AV156352
Dpysl2	−5.6	1834	0.03	dihydropyrimidinase-like 2	BQ174209
Rhobtb3	−5.5	1090	0.02	Rho-related BTB domain 3	BM942043
Rhoe	−5.4	875	0.03	ras homologue gene family, member E	BC009002
Mpdz	−4.5	678	0.01	multiple PDZ domain protein	AK019164
Ror2	−4.5	230	0.02	receptor YK-like orphan receptor 2	AV324603
Ndn	−4.5	255	0.00	necdin	NM_010882
Tnfrsf19	−4.4	383	0.03	tumor necrosis factor receptor 19	NM_013869
Tgif	−4.1	605	0.01	TG-interacting factor	NM-009372
Zfp334	−4.0	413	0.04	zinc finger protein 334	AU022425
Ndn	−4.0	955	0.03	necdin	AW743020
Ccnd1	−4.0	1376	0.04	cyclin D1	NM_007631
Ndn	−4.0	866	0.04	necdin	AV124445
Ptbp2	−3.9	1174	0.03	polypyrimidine binding protein 2	BB076855
Ndn	−3.8	732	0.04	necdin	AW743020
Wnt5a	−3.8	404	0.05	wingless-related 5a	BC018425
Marks	−3.8	525	0.01	myristoylated ala-rich PKC substrate	AW546141
Prkcm	−3.6	489	0.04	protein kinase C μ	AV297026
Hmga2	−3.6	1093	0.03	high mobility group AT-hook 2	X58380
Prkcm	−3.6	489	0.04	protein kinase C μ	NM_008858
Cdh2	−3.6	2808	0.05	cadherin 2	BC022107
Ets1	−3.5	284	0.01	E26 avian leukemia oncogene 1	BB151715
Tcfl5	−3.5	2041	0.04	procollagen type IX, alpha 3	AV044715
Hmgb3	−3.4	4111	0.02	high mobility group box 3	NM_008253
Marks	−3.3	328	0.04	myristoylated ala-rich PKC substrate	AW546141
Wnt5a	−3.2	522	0.04	wingless-related 5a	BB067079
Marks	−3.0	1371	0.02	myristoylated ala-rich PKC substrate	AW546141

¹All footnotes from Table 2 apply to Table 3.

Notably absent from the list of differentially abundant mRNAs were those encoding most cell proliferation markers (see below). Also absent was the gene encoding the mitochondrial Txnrd2 protein. mRNAs encoding only five other reductases showed evidence of mRNA overabundance in the mutants. These were sulfiredoxin 1 (Srxn1), carbonyl reductase 3 (Cbr3), leukotriene B4 12-dehydrogenase (Ltb4dh), malic enzyme 1 (Mod1), and biliverdin reductase b (Blvrb) (Table 2). Many mRNAs that were underrepresented in the mutants encoded proteins not expressed until during or after gastrulation (Table 3), consistent with the mutant embryos failing to gastrulate.

To further test whether mRNAs encoding components participating in proliferative or oxidoreductase pathways were affected by the mutation, we used GeneSpring software to evaluate relative levels of all mRNAs whose signal reproducibility (variance) in the arrays provided reliable data (P ≤ 0.05) for genes involved in these processes (Tables S2 and S3). Of the 88 annotated probe sets meeting these criteria in the cell proliferation set, only those representing cyclins D1 and D2 showed differential mRNA abundance (≥ 3-fold differential) between mutant and wild-type embryos. Probe sets having reproducible yet similar signals between mutant and wild-type embryos in the proliferation set included 24 sets for DNA polymerases or primases, 7 sets for histones, and 3 sets for deoxyribonucleotide synthesis (ribonucleotide reductase, dihydrofolate reductase, and thymidylate synthase)(Table S2). The 64 annotated probe sets giving reliable data in the oxidoreductase set showed no underrepresented mRNAs and only 5 mRNAs (listed above) that were ≥ 3-fold overrepresented in the mutants. Probe sets having reliable yet similar signals between mutant and wild-type embryos in the oxidoreductase set included peroxiredoxins (Prdx) 1, 2, and 4 (Table S3).

Representative mRNAs from the microarray data were chosen for RT-PCR or real-time RT-PCR confirmation analyses using littermate-pairs of single wild-type and mutant E7.5 embryos (Figs 5, S2). All primer pairs were designed to amplify 3′ cDNA sequences spanning an exon/exon junction. Real-time PCR analyses gave data that qualitatively matched the array data and thermal dissociation curves that matched predictions based on the length and G/C-content of the amplified region (Figs 5b, S2). Of particular importance, single-embryo RT-PCR analyses confirmed that, whereas Prdx1, 2, and 4 mRNA levels did not differ substantially between mutant and wild type embryos, Srxn1 mRNA levels were dramatically higher in the mutants (Fig. 5c).

Fig. 5.

Analyses of gene expression in littermate pairs of single E7.5 txnrd1^+/+ and txnrd1^−/− embryos. All assays were performed on single E7.5 embryos genotyped as in Fig. 1c. a, RT-PCR confirmation of relative mRNA abundances in paired sets of txnrd1^−/− and txnrd1^+/+ embryos, as indicated. b, real-time PCR validation of relative mRNA levels for selected genes of interest. Each data point represents the ratio of txnrd1^−/− to txnrd1^+/+ mRNA in a paired littermate set of single E7.5 embryos. c, single-embryo RT-PCR confirmation of expression differences of Srxn and Prdx mRNAs from E7.5 littermate pairs of two litters, as indicated. Abbreviations: Cbr3, carbonyl reductase 3; Cdh2, cadherin homologue 2; Fgf13, fibroblast growth factor family member 13; Gstμ1, glutathione S-transferase μ1; Gclm, glutamate-cysteine ligase; Igfbp1, insulin-like growth factor-binding protein 1; Mash2, achaete-scute homologue-like 2; Mod1, malic enzyme; Pcdh8, protocadherin 8; Pl1, placental lactogen 1; Prdx, peroxiredoxin; Sox11, SRY box-containing gene 11; Srxn1, sulfiredoxin 1; t, brachyury.

In summary, our data show that disruption of the txnrd1 gene in mice blocked mesoderm formation and patterning of the embryo, but had little effect on zygote-derived extraembryonic cell and tissue types. The mRNA abundance profiles in the mutant embryos were consistent with a failure to form advanced embryonic tissue types and inconsistent with the mutation disrupting proliferation. Many mRNA abundance differences matched those reported in txnrd1-null yeast [32], suggesting that compensatory pathways are generally shared between these evolutionarily distant species. However, notably absent from the list of up-regulated mRNAs in mutant mice were the Prdxs, which have been shown to be strongly up-regulated in thioredoxin reductase-deficient yeast [32]. The implications of these findings are discussed below.

Discussion

Survival, proliferation, and differentiation of cells in txnrd1^−/− embryos

Survival of txnrd1^−/− embryos to E8.5 indicated that cells lacking Txnrd1 were capable of synthesizing DNA, RNA, lipid, and protein, and performing all basic metabolic functions required for cell survival, growth, and proliferation within the context of early mammalian development. In addition, txnrd1^−/− embryos expressed early differentiation markers, including mRNA for Cripto, Fgf8, Snail1, Mash2, and Pl1, some of which were shown by in situ hybridization to be regionally restricted within the embryo (Figs 4, 5). These results indicate that some cell differentiation occurred in the mutants.

Systems affected by Txnrd1 disruption

Yeast lacking a functional Txn system survive due to compensation by the glutathione pathway [12, 51]. In the current study, transcriptome profiling indicated twenty-one annotated mRNAs were substantially overrepresented in txnrd1^−/− mouse embryos (Table 1). Interestingly, several of these mRNAs are also upregulated in yeast lacking Txnrd1 [32]. These included mRNAs for glutathione-S-transferases (GSTs) metallothioneins (MTs), and carbonyl reductase 3 (Cbr3) (Table 1). However, unlike in yeast [32], levels of mRNAs encoding glutathione reductases and peroxiredoxins (Prdxs) were not higher in txnrd1^−/− as compared to wild-type embryos (Fig. 5c; Tables 2 and S3).

1) GSTs, MTs, and Cbr3

GSTs are detoxifying enzymes that conjugate glutathione to potentially toxic substrates [52]; they are not reductases. Similarly, although MTs contain iron-thiolate centers [53], they cannot sequester ferric or ferrous iron [54], the metals with the highest potential for generating reactive oxygen species in cells [55]. Thus, although both mouse embryos and yeast overaccumulate GST and MT mRNAs in response to disruption of txnrd1 (Table 1) [32], this probably does not directly help cells combat oxidative stress or compensate for a deficit in reducing activity. We suspect these are non-adaptive responses. For example, disruption of the Txn pathway in yeast [51], and perhaps also in mammals, shifts the thiol maintenance burden onto the glutathione pathway. This, in turn, might lead to reduced flux through glutathione-dependent enzymes, including GST, and reduced GST activity might stimulate feedback induction of GST mRNA accumulation [56]. Similarly, the iron-thiolate centers in MTs confer redox sensitivity onto heavy metal-binding by the protein [57–59]. As a result, txnrd1 disruption might antagonize heavy metal chelation by MTs, resulting in elevated levels of free intracellular zinc, a potent inducer of mt gene expression [60, 61], and causing the observed increase in MT mRNA levels. This would suggest that redox-dependent liberation of zinc by MTs might affect transcription of other genes in response to txnrd1 disruption. Interestingly, in Drosophila, GST mRNAs have been shown to be induced in response to zinc [62]. Thus, release of zinc from MTs might also contribute to the strong GST mRNA response observed following txnrd1 disruption in yeast [32] and mouse embryos (this study). Further investigations will be needed to test whether MTs participate in a redox sensor-and-response mechanism and to determine the basis of the GST mRNA response to a Txnrd1 deficiency.

Little is known about the functions of Cbr. Yeast lacking Cbr are viable [63]. Biochemical studies show the enzyme reduces lipid carbonyl groups, which arise from lipid peroxidation during oxidative stress, to alcohols [64], and this may protect neurons from oxidative damage [65]. Upregulation of Cbr3 mRNA in txnrd1-null yeast [32] and mice (Figs 5a, b; Tables 2, S3) might be a reaction to intracellular lipid peroxidation, and thus indicative of intracellular oxidative stress at membrane surfaces. Further studies will be required to determine the pathways by which Cbr3 mRNA levels respond to the loss of Txnrd1.

2) Prdx system

It is of interest that txnrd1^−/− mouse embryos did not exhibit increased levels of mRNAs encoding Prdxs (Fig. 5c; Table S3). In yeast, the Tsa1 gene, encoding the major cytosolic form of 2-Cys Prdx, is the most highly induced gene in txn1-null yeast; the Tsa2 gene, encoding a second cytosolic 2-Cys Prdx, is also highly induced [32]. None of the six mammalian 2-Cys Prdxs were substantially induced in txnrd1^−/− embryos (Fig. 5c; Table S3).

Prdxs are ubiquitous Txn-dependent peroxidases [66–68]. The 2-Cys family of Prdxs initiate catalysis when a cysteine at the active site reacts with H₂O₂ to form a sulfenic acid intermediate. The sulfenic acid reacts with a resolving cysteine to form a disulfide, which is subsequently reduced by Txn to return the enzyme to its initial state [66] (Fig. 6). An appendage unique to eukaryotic 2-Cys Prdxs slows the rate of formation of the disulfide [69]. As a result, even at moderate H₂O₂ concentrations, the sulfenic acid intermediate can react with a second H₂O₂ molecule to form a sulfinic acid before it can react with the resolving cytseine to form a disulfide (Fig. 6). The sulfinic form is inactive and cannot be reduced by Txn [69]. H₂O₂ has been implicated as an important signaling molecule in eukaryotic cells [70](see below). It has been hypothesized that eukaryotes evolved 2-Cys Prdxs to serve as “floodgates” that prevent low levels of H₂O₂ from triggering signaling but allow H₂O₂ to oxidize downstream substrates once a threshold is exceeded [68]. Srxns, a novel class of slow ATP-dependent enzymes, convert the sulfinic acid form of Prdx back to the active form [67–71] (Fig. 6).

Fig. 6.

Eukaryotic Prdx catalytic cycles and inferred compensation for Txnrd1 disruption. At top is shown the normal Prdx catalytic cycles [68, 71]. The cycle at left is Txnrd1-dependent and functions to hydrolyze H₂O₂ at low concentrations. The cycle at right is Txnrd1-independent, and is only active at high H₂O₂ concentrations, where the Prdx sulfenic acid is further oxidized to a sulfinic acid, which cannot be reduced by Txn [74]. The mechanisms of Srxn catalysis are still under investigation; most models agree it is an ATP-dependent reaction [71, 72]. Below is shown the compensatory shifts expected in yeast [32] (left) and mice (right) in response to Txn system disruption. In both species, disruption of Txnrd1 is expected to disrupt the low [H₂O₂] Prdx cycle. Yeast strongly upregulate Prdx mRNA levels [32], suggesting compensation involves increasing rates of Prdx protein synthesis. By contrast, mice did not alter Prdx mRNA levels, but strongly upregulated Srxn mRNA (this study). Either response might bolster the high [H₂O₂] Prdx cycle.

Although the txnrd1^−/− embryos did not exhibit increased levels of mRNAs encoding Prdxs, mRNA from the npn3 gene, encoding the mouse homologue of yeast srxnI [72–74], was the second most highly overrepresented mRNA in txnrd1^−/− embryos (11-fold; Tables 2, S3; Fig. 5c). This suggests that, like upregulation of mRNAs encoding GSTs, MTs, and Cbr3, augmentation of the Prdx system is a conserved response to an impaired Trx system. However, whereas yeast respond by upregulating Prdx mRNA levels, mice responded by upregulating Srxn, and thereby might more effectively restore the sulfinic form of Prdx to the active form. Interestingly, in both species, this might provide an unexpected ATP-dependent means of H₂O₂ detoxification through the Txnrd1-independent Srxn-dependent high [H₂O₂] loop of the Prdx catalytic cycle (Fig. 6). Further studies will be required to test this.

Effects on signaling

Many developmental fate-determining and patterning signals in embryos involve growth factor binding to surface receptors whose activities involve tyrosine phosphorylation. These factors include BMPs, FGFs, IGFs, and others. An intact Txn system may be necessary to attenuate signaling by surface receptors. In response to ligand binding, NADPH-oxidase enzymes associated with the cytosolic aspect of growth factor receptors generate H₂O₂ [22–24, 26, 75–83]. A conserved catalytic cysteine at the active site of protein tyrosine phosphatases is oxidized by H₂O₂ to a sulfenic acid, which further reacts with a backbone nitrogen to form a sulfenylamide [68]. Protein tyrosine phosphatase inactivation by H₂O₂ may facilitate tyrosine kinase signaling by stabilizing phosphotyrosines [26, 70, 82]. Txn1 reduces the sulfenylamide of oxidized protein tyrosine phosphatases and restores enzyme activity [68]. Txn1 is also required for H₂O₂ decomposition by Prdx at low H₂O₂ concentrations (Fig. 6; see above). Thus, an intact Txn system may be required to limit H₂O₂ levels and prevent excessive tyrosine kinase signaling.

Excessive signaling from surface receptors could have diverse effects. For example, if two signaling pathways compete for fate determination in a particular cell, hyper-activation of a tyrosine phosphate-dependent pathway might shift the equilibrium toward having all cells respond only to this signal, thus excluding induction of other cell fates. Alternatively, hyper-activation of a phosphotyrosine-dependent pathway, especially in the absence of effective H₂O₂ clearance (see above) might kill the target cells, effectively excluding differentiation along these fates. Interestingly, the most dramatically overrepresented mRNA in mutant embryos was that encoding the insulin-like growth factor-binding protein, IGFBP1 (Table 1). In a recent study, Goretta et al. [84] showed that IGFBP3 is up-regulated in cytokine-induced HepG2 human hepatocarcinoma cell cultures that were treated with siRNA targeted to Txnrd1. Increased IGFBP mRNA levels in Txnrd1-deficient cells may be a feedback response to repress excessive signaling from the phosphotyrosine-dependent IGF receptor.

Phenotypic differences between txnrd1^−/− mouse lines

While the study described here was in progress, another mouse line bearing a disrupted txnrd1 gene was described that exhibits a substantially different phenotype [85]. Homozygous embryos in the Jakupoglu et al. [85] study did not have a gastrulation defect. Instead, these mice arrested at a later stage, just before turning, with a neural tube and head folds, somites, and a functional heart [85]. At this stage, all cell types except cardiac muscle underwent cataclysmic death.

Although both the allele described in the present study and that of Jakupoglu et al. [85] are predicted to be functionally null in the recombined state (each eliminates a different set of essential active site amino acids, for example), the phenotype we report here is far more severe. The allele developed in the current study eliminates exons 1 and 2, including all functional ATGs and the two N-terminal active site Cys⁵⁹ and Cys⁶⁴ residues (Fig. 1); the other eliminates exon 15, including the C-terminal Cys and Sec amino acids [85]. Jakupoglu et al’s [85] allele would be predicted to produce a functional translatable mRNA encoding a catalytically inactive protein. However they were able to detect only trace amounts of mRNA and no protein accumulation[85], suggesting that this C-terminal truncation destabilizes the entire mRNA. All other reported parameters, including mouse and ES cell strains and care conditions, are similar between the two studies.

Jakupoglu et al. [85] suggested that Txnrd2 compensated partially for absence of Txnrd1 in most cell types, allowing normal embryonic development to the neural fold/somite stage, and completely in cardiac cells, allowing full heart development and function. Cell cycle arrest of non-cardiac cells was implicated in embryonic failure at the neural fold/somite stage. Unlike our study, global gene expression was not measured. By contrast, our transcriptome analysis showed that Txnrd2 mRNA was not more abundant in mutant embryos (Table 2, Fig. 5a), and, like in Txnrd1-null yeast [32] and Txnrd1-knock-down HepG2 cells [84], most proliferation markers, including mRNAs encoding histones, DNA polymerases, and others, were similar in mutant and wild-type embryos (Table S2). Further analyses on both alleles will likely be required to resolve the complex physiological roles of Txnrd1 in mouse embryos and why such different phenotypes might be manifested.