RNA profiling of human testicular cells identifies syntenic lncRNAs associated with spermatogenesis

STUDY QUESTION

Is the noncoding transcriptional landscape during spermatogenesis conserved between human and rodents?

SUMMARY ANSWER

We identified a core group of 113 long noncoding RNAs (lncRNAs) and 20 novel genes dynamically and syntenically transcribed during spermatogenesis.

WHAT IS KNOWN ALREADY

Spermatogenesis is a complex differentiation process driven by a tightly regulated and highly specific gene expression program. Recently, several studies in various species have established that a large proportion of known lncRNAs are preferentially expressed during meiosis and spermiogenesis in a testis-specific manner.

STUDY DESIGN, SIZE, DURATION

To further investigate lncRNA expression in human spermatogenesis, we carried out a cross-species RNA profiling study using isolated testicular cells.

PARTICIPANTS/MATERIALS, SETTING, METHODS

Human testes were obtained from post-mortem donors (N = 8, 51 years old on average) or from prostate cancer patients with no hormonal treatment (N = 9, 80 years old on average) and only patients with full spermatogenesis were used to prepare enriched populations of spermatocytes, spermatids, Leydig cells, peritubular cells and Sertoli cells. To minimize potential biases linked to inter-patient variations, RNAs from two or three donors were pooled prior to RNA-sequencing (paired-end, strand-specific). Resulting reads were mapped to the human genome, allowing for assembly and quantification of corresponding transcripts.

MAIN RESULTS AND THE ROLE OF CHANCE

Our RNA-sequencing analysis of pools of isolated human testicular cells enabled us to reconstruct over 25 000 transcripts. Among them we identified thousands of lncRNAs, as well as many previously unidentified genes (novel unannotated transcripts) that share many properties of lncRNAs. Of note is that although noncoding genes showed much lower synteny than protein-coding ones, a significant fraction of syntenic lncRNAs displayed conserved expression during spermatogenesis.

LARGE SCALE DATA

Raw data files (fastq) and a searchable table (.xlss) containing information on genomic features and expression data for all refined transcripts have been submitted to the NCBI Gene Expression Omnibus under accession number GSE74896.

LIMITATIONS, REASONS FOR CAUTION

Isolation procedures may alter the physiological state of testicular cells, especially for somatic cells, leading to substantial changes at the transcriptome level. We therefore cross-validated our findings with three previously published transcriptomic analyses of human spermatogenesis. Despite the use of stringent filtration criteria, i.e. expression cut-off of at least three fragments per kilobase of exon model per million reads mapped, fold-change of at least three and false discovery rate adjusted P-values of less than <1%, the possibility of assembly artifacts and false-positive transcripts cannot be fully ruled out.

WIDER IMPLICATIONS OF THE FINDINGS

For the first time, this study has led to the identification of a large number of conserved germline-associated lncRNAs that are potentially important for spermatogenesis and sexual reproduction. In addition to further substantiating the basis of the human testicular physiology, our study provides new candidate genes for male infertility of genetic origin. This is likely to be relevant for identifying interesting diagnostic and prognostic biomarkers and also potential novel therapeutic targets for male contraception.

STUDY FUNDING/COMPETING INTEREST(S)

This work was supported by l'Institut national de la santé et de la recherche médicale (Inserm); l'Université de Rennes 1; l'Ecole des hautes études en santé publique (EHESP); INERIS-STORM to B.J. [N 10028NN]; Rennes Métropole 'Défis scientifiques émergents' to F.C (2011) and A.D.R (2013). The authors have no competing financial interests.

Genome-wide identification of Sox8-, and Sox9-dependent genes during early post-natal testis development in the mouse

The SOX8 and SOX9 transcription factors are involved in, among others, sex differentiation, male gonad development and adult maintenance of spermatogenesis. Sox8(-/-) mice lacking Sox9 in Sertoli cells fail to form testis cords and cannot establish spermatogenesis. Although genetic and histological data show an important role for these transcription factors in regulating spermatogenesis, it is not clear which genes depend upon them at a genome-wide level. To identify transcripts that respond to the absence of Sox8 in all cells and Sox9 in Sertoli cells we measured mRNA concentrations in testicular samples from mice at 0, 6 and 18 days post-partum. In total, 621 and 629 transcripts were found at decreased or increased levels, respectively, at different time points in the mutant as compared to the control samples. These mRNAs were categorized as preferentially expressed in Sertoli cells or germ cells using data obtained with male and female gonad samples and enriched testicular cell populations. Five candidate genes were validated at the protein level. Furthermore, we identified putative direct SOX8 and SOX9 target genes by integrating predicted SOX-binding sites present in potential regulatory regions upstream of the transcription start site. Finally, we used protein network data to gain insight into the effects on regulatory interactions that occur when Sox8 and Sox9 are absent in developing Sertoli cells. The integration of testicular samples with enriched Sertoli cells, germ cells and female gonads enabled us to broadly distinguish transcripts directly affected in Sertoli cells from others that respond to secondary events in testicular cell types. Thus, combined RNA profiling signals, motif predictions and network data identified putative SOX8/SOX9 target genes in Sertoli cells and yielded insight into regulatory interactions that depend upon these transcription factors. In addition, our results will facilitate the interpretation of genome-wide in vivo SOX8 and SOX9 DNA binding data.

Proteome analysis and genome-wide regulatory motif prediction identify novel potentially sex-hormone regulated proteins in rat efferent ducts

The efferent ducts are a series of tubules that conduct sperm from the rete testis to the epididymis. They absorb most fluid and proteins originating from the rete testis during concentration of spermatozoa prior to their entry into the epididymis. Proteome analysis of micro-dissected efferent duct samples from adult rats was combined with genome-wide computational prediction of conserved hormone response elements to identify factors likely regulated by oestrogens and androgens. We identified 165 proteins and found subsets of the promoters controlling their corresponding genes to contain androgen- and oestrogen response elements (ARE/EREs) at similar frequencies. Moreover, EREs were significantly enriched among the loci identified compared with their genome-wide occurrence. The expression and localization of Anxa6, Ckb, Krt19, Park7, Pdzk1 and Tpt1 in the efferent ducts and other related hormone controlled tissues was further validated at the RNA or protein level. This study identifies many novel proteins predicted to play roles in sperm maturation and male fertility and provides significant computational evidence that the efferent ducts express genes transcriptionally controlled by sex hormones.

Database model and specification of GermOnline Release 2.0, a cross-species community annotation knowledgebase on germ cell differentiation

GermOnline is a web-accessible relational database that enables life scientists to make a significant and sustained contribution to the annotation of genes relevant for the fields of mitosis, meiosis, germ line development and gametogenesis across species. This novel approach to genome annotation includes a platform for knowledge submission and curation as well as microarray data storage and visualization hosted by a global network of servers.

AVAILABILITY

The database is accessible at http://www.germonline.org/. For convenient world-wide access we have set up a network of servers in Europe (http://germonline.unibas.ch/; http://germonline.igh.cnrs.fr/), Japan (http://germonline.biochem.s.u-tokyo.ac.jp/) and USA (http://germonline.yeastgenome.org/).

SUPPLEMENTARY INFORMATION

Extended documentation of the database is available through the link 'About GermOnline' at the websites.

GermOnline, a cross-species community knowledgebase on germ cell differentiation

GermOnline provides information and microarray expression data for genes involved in mitosis and meiosis, gamete formation and germ line development across species. The database has been developed, and is being curated and updated, by life scientists in cooperation with bioinformaticists. Information is contributed through an online form using free text, images and the controlled vocabulary developed by the GeneOntology Consortium. Authors provide up to three references in support of their contribution. The database is governed by an international board of scientists to ensure a standardized data format and the highest quality of GermOnline's information content. Release 2.0 provides exclusive access to microarray expression data from Saccharomyces cerevisiae and Rattus norvegicus, as well as curated information on approximately 700 genes from various organisms. The locus report pages include links to external databases that contain relevant annotation, microarray expression and proteome data. Conversely, the Saccharomyces Genome Database (SGD), S.cerevisiae GeneDB and Swiss-Prot link to the budding yeast section of GermOnline from their respective locus pages. GermOnline, a fully operational prototype subject-oriented knowledgebase designed for community annotation and array data visualization, is accessible at http://www.germonline.org. The target audience includes researchers who work on mitotic cell division, meiosis, gametogenesis, germ line development, human reproductive health and comparative genomics.

A screen for genes required for meiosis and spore formation based on whole-genome expression

BACKGROUND

Meiosis is the process by which gametes are generated with half the ploidy of somatic cells. This reduction is achieved by three major differences in chromosome behavior during meiosis as compared to mitosis: the production of chiasmata by recombination, the protection of centromere-proximal sister chromatid cohesion, and the monoorientation of sister kinetochores during meiosis I. Mistakes in any of these processes lead to chromosome missegregation.

RESULTS

To identify genes involved in meiotic chromosome behavior in Saccharomyces cerevisiae, we deleted 301 open reading frames (ORFs) which are preferentially expressed in meiotic cells according to microarray gene expression data. To facilitate the detection of chromosome missegregation mutants, chromosome V of the parental strain was marked by GFP. Thirty-three ORFs were required for the formation of wild-type asci, eight of which were needed for proper chromosome segregation. One of these (MAM1) is essential for the monoorientation of sister kinetochores during meiosis I. Two genes (MND1 and MND2) are implicated in the recombination process and another two (SMA1 and SMA2) in prospore membrane formation.

CONCLUSIONS

Reverse genetics using gene expression data is an effective method for identifying new genes involved in specific cellular processes.

The core meiotic transcriptome in budding yeasts

We used high-density oligonucleotide microarrays to analyse the genomes and meiotic expression patterns of two yeast strains, SK1 and W303, that display distinct kinetics and efficiencies of sporulation. Hybridization of genomic DNA to arrays revealed numerous gene deletions and polymorphisms in both backgrounds. The expression analysis yielded approximately 1,600 meiotically regulated genes in each strain, with a core set of approximately 60% displaying similar patterns in both strains. Most of these (95%) are MATa/MATalpha-dependent and are not similarly expressed in near-isogenic meiosis-deficient controls. The transcript profiles correlate with the distribution of defined meiotic promoter elements and with the time of known gene function.

Modular long-range regulation of Myf5 reveals unexpected heterogeneity between skeletal muscles in the mouse embryo

The myogenic factor Myf5 plays a key role in muscle cell determination, in response to signalling cascades that lead to the specification of muscle progenitor cells. We have adopted a YAC transgenic approach to identify regulatory sequences that direct the complex spatiotemporal expression of this gene during myogenesis in the mouse embryo. Important regulatory regions with distinct properties are distributed over 96 kb upstream of the Myf5 gene. The proximal 23 kb region directs early expression in the branchial arches, epaxial dermomyotome and in a central part of the myotome, the epaxial intercalated domain. Robust expression at most sites in the embryo where skeletal muscle forms depends on an enhancer-like sequence located between −58 and −48 kb from the Myf5 gene. This element is active in the epaxial and hypaxial myotome, in limb muscles, in the hypoglossal chord and also at the sites of Myf5 transcription in prosomeres p1 and p4 of the brain. However later expression of Myf5 depends on a more distal region between −96 and −63 kb, which does not behave as an enhancer. This element is necessary for expression in head muscles but strikingly only plays a role in a subset of trunk muscles, notably the hypaxially derived ventral body muscles and also those of the diaphragm and tongue. Transgene expression in limb muscle masses is not affected by removal of the −96/-63 region. Epaxially derived muscles and some hypaxial muscles, such as the intercostals and those of the limb girdles, are also unaffected. This region therefore reveals unexpected heterogeneity between muscle masses, which may be related to different facets of myogenesis at these sites. Such regulatory heterogeneity may underlie the observed restriction of myopathies to particular muscle subgroups.

Critical activities of Rac1 and Cdc42Hs in skeletal myogenesis: antagonistic effects of JNK and p38 pathways

The Rho family of GTP-binding proteins plays a critical role in a variety of cellular processes, including cytoskeletal reorganization and activation of kinases such as p38 and C-jun N-terminal kinase (JNK) MAPKs. We report here that dominant negative forms of Rac1 and Cdc42Hs inhibit the expression of the muscle-specific genes myogenin, troponin T, and myosin heavy chain in L6 and C2 myoblasts. Such inhibition correlates with decreased p38 activity. Active RhoA, RhoG, Rac1, and Cdc42Hs also prevent myoblast-to-myotube transition but affect distinct stages: RhoG, Rac1, and Cdc42Hs inhibit the expression of all muscle-specific genes analyzed, whereas active RhoA potentiates their expression but prevents the myoblast fusion process. We further show by two different approaches that the inhibitory effects of active Rac1 and Cdc42Hs are independent of their morphogenic activities. Rather, myogenesis inhibition is mediated by the JNK pathway, which also leads to a cytoplasmic redistribution of Myf5. We propose that although Rho proteins are required for the commitment of myogenesis, they differentially influence this process, positively for RhoA and Rac1/Cdc42Hs through the activation of the SRF and p38 pathways, respectively, and negatively for Rac1/Cdc42Hs through the activation of the JNK pathway.

Myf5 is a novel early axonal marker in the mouse brain and is subjected to post-transcriptional regulation in neurons

Myf5 is a key basic Helix-Loop-Helix transcription factor capable of converting many non-muscle cells into muscle. Together with MyoD it is essential for initiating the skeletal muscle programme in the embryo. We previously identified unexpected restricted domains of Myf5 transcription in the embryonic mouse brain, first revealed by Myf5-nlacZ(+/)(−) embryos (Tajbakhsh, S. and Buckingham, M. (1995) Development 121, 4077–4083). We have now further characterized these Myf5 expressing neurons. Retrograde labeling with diI, and the use of a transgenic mouse line expressing lacZ under the control of Myf5 regulatory sequences, show that Myf5 transcription provides a novel axonal marker of the medial longitudinal fasciculus (mlf) and the mammillotegmental tract (mtt), the earliest longitudinal tracts to be established in the embryonic mouse brain. Tracts projecting caudally from the developing olfactory system are also labelled. nlacZ and lacZ expression persist in the adult brain, in a few ventral domains such as the mammillary bodies of the hypothalamus and the interpeduncular nucleus, potentially derived from the embryonic structures where the Myf5 gene is transcribed. To investigate the role of Myf5 in the brain, we monitored Myf5 protein accumulation by immunofluorescence and immunoblotting in neurons transcribing the gene. Although Myf5 was detected in muscle myotomal cells, it was absent in neurons. This would account for the lack of myogenic conversion in brain structures and the absence of a neural phenotype in homozygous null mutants. RT-PCR experiments show that the splicing of Myf5 primary transcripts occurs correctly in neurons, suggesting that the lack of Myf5 protein accumulation is due to regulation at the level of mRNA translation or protein stability. In the embryonic neuroepithelium, Myf5 is transcribed in differentiated neurons after the expression of neural basic Helix-Loop-Helix transcription factors. The signalling molecules Wnt1 and Sonic hedgehog, implicated in the activation of Myf5 in myogenic progenitor cells in the somite, are also produced in the viscinity of the Myf5 expression domain in the mesencephalon. We show that cells expressing Wnt1 can activate neuronal Myf5-nlacZ gene expression in dissected head explants isolated from E9.5 embryos. Furthermore, the gene encoding the basic Helix-Loop-Helix transcription factor mSim1 is expressed in adjacent cells in both the somite and the brain, suggesting that signalling molecules necessary for the activation of mSim1 as well as Myf5 are present at these different sites in the embryo. This phenomenon may be widespread and it remains to be seen how many other potentially potent regulatory genes, in addition to Myf5, when activated do not accumulate protein at inappropriate sites in the embryo.

The muscle regulatory factors MyoD and myf-5 undergo distinct cell cycle-specific expression in muscle cells

The muscle regulators MyoD and Myf-5 control cell cycle withdrawal and induction of differentiation in skeletal muscle cells. By immunofluorescence analysis, we show that MyoD and Myf-5 expression patterns become mutually exclusive when C2 cells are induced to differentiate with Myf-5 staining present in cells which fail to differentiate. Isolation of these undifferentiated cells reveals that upon serum stimulation they reenter the cell cycle, express MyoD and downregulate Myf-5. Similar regulations of MyoD and Myf-5 were observed using cultured primary myoblasts derived from satellite cells. To further analyze these regulations of MyoD and Myf-5 expression, we synchronized proliferating myoblasts. Analysis of MyoD and Myf-5 expression during cell cycle progression revealed distinct and contrasting profiles of expression. MyoD is absent in G0, peaks in mid-G1, falls to its minimum level at G1/S and reaugments from S to M. In contrast, Myf-5 protein is high in G0, decreases during G1 and reappears at the end of G1 to remain stable until mitosis. These data demonstrate that the two myogenic factors MyoD and Myf-5 undergo specific and distinct cell cycle-dependent regulation, thus establishing a correlation between the cell cycle-specific ratios of MyoD and Myf-5 and the capacity of cells to differentiate: (a) in G1, when cells express high levels of MyoD and enter differentiation; (b) in G0, when cells express high levels of Myf-5 and fail to differentiate.

A novel GFPneo vector designed for the isolation and analysis of enhancer elements in transfected mammalian cells

We have designed a new approach to the direct cloning and rapid analysis of mammalian enhancer elements by fusing green fluorescent protein and neomycinphosphotransferase under the control of a thymidine kinase minimal promoter. DNA fragments containing known or potential enhancer elements can be inserted into a polylinker upstream of GFPneo and re-isolated from stably transfected cell lines by a direct transgene-specific polymerase chain reaction (PCR), for further analysis. C2C12 muscle cells were transfected with four vectors containing the GFPneo fusion gene regulated by the cytomegalovirus promoter, the myoD distal core enhancer and myoblast- and myotube-specific enhancers from the desmin gene. GFPneo shows robust epifluorescence by microscopy and flow cytometry and retains sufficient neo activity to permit selection of G418-resistant clones. The fluorescence signal pattern of GFPneo expressed under the control of the desmin enhancers mirrors their transcriptional profile during myogenic differentiation. This finding demonstrates the value of GFPneo as a tool to analyse differentiation stage-specific regulatory DNA elements in stably transfected mammalian cell lines. We were able to re-isolate the myoD enhancer mediating GFPneo expression from a stably transfected C2C12 clone by a transgene-specific PCR reaction, demonstrating the feasibility of using this new vector system for the isolation of regulatory sequences.

RhoA GTPase and serum response factor control selectively the expression of MyoD without affecting Myf5 in mouse myoblasts

MyoD and Myf5 belong to the family of basic helix-loop-helix transcription factors that are key operators in skeletal muscle differentiation. MyoD and Myf5 genes are selectively activated during development in a time and region-specific manner and in response to different stimuli. However, molecules that specifically regulate the expression of these two genes and the pathways involved remain to be determined. We have recently shown that the serum response factor (SRF), a transcription factor involved in activation of both mitogenic response and muscle differentiation, is required for MyoD gene expression. We have investigated here whether SRF is also involved in the control of Myf5 gene expression, and the potential role of upstream regulators of SRF activity, the Rho family G-proteins including Rho, Rac, and CDC42, in the regulation of MyoD and Myf5. We show that inactivation of SRF does not alter Myf5 gene expression, whereas it causes a rapid extinction of MyoD gene expression. Furthermore, we show that RhoA, but not Rac or CDC42, is also required for the expression of MyoD. Indeed, blocking the activity of G-proteins using the general inhibitor lovastatin, or more specific antagonists of Rho proteins such as C3-transferase or dominant negative RhoA protein, resulted in a dramatic decrease of MyoD protein levels and promoter activity without any effects on Myf5 expression. We further show that RhoA-dependent transcriptional activation required functional SRF in C2 muscle cells. These data illustrate that MyoD and Myf5 are regulated by different upstream activation pathways in which MyoD expression is specifically modulated by a RhoA/SRF signaling cascade. In addition, our results establish the first link between RhoA protein activity and the expression of a key muscle regulator.

The glucocorticoid receptor and AP-1 are involved in a positive regulation of the muscle regulatory gene myf5 in cultured myoblasts

The muscle regulatory factor, myf5, is involved in the establishment of skeletal muscle precursor cells. Little is known, however, about the control of the expression of the gene encoding this basic helix-loop-helix (bHLH) factor. We have addressed this question in the mouse myogenic cell line, C2, and in a derivative of this cell line where the myf5 gene is the only muscle-specific bHLH factor to be expressed at the myoblast stage. We present evidence that the synthetic glucocorticoid dexamethasone, and the pharmacological agent anisomycin, act synergistically to rapidly up-regulate the levels of myf5 transcript and protein. The glucocorticoid antagonist RU 486 abolishes this synergy, demonstrating the involvement of the glucocorticoid receptor. The expression of a dominant negative mutant of c-jun which interferes with the transactivating properties of all AP-1 family members also blocks the induction of myf5 by anisomycin and dexamethasone. An activator of protein kinase C (PKCs), 12-O-tetradecanoyl phorbol 13-acetate (TPA), abolishes the up-regulation of myf5 gene expression by dexamethasone and anisomycin, and its effect is counteracted by an inhibitor of PKCs, GF 109203X. These results point to the possible involvement of PKCs in the negative control of myf5. Evidence that both positive and negative regulation of myf5 transcripts, described here, does not require the fresh synthesis of transcription factors suggests that myf5 may behave like an immediate early gene.

A DNA binding factor (UBF) interacts with a positive regulatory element in the promoters of genes expressed during meiosis and vegetative growth in yeast

We have studied the bipartite regulatory element UASH/URS1 in the promoter of HOP1, whose product is required for synapsis and correct pairing of homologous chromosomes during the first meiotic division. HOP1 is transcriptionally repressed by the URS1 motif during vegetative growth and induced during meiotic prophase by the UASH motif in cooperation with the bifunctional URS1 site, which is required for full induction of HOP1. While URS1 is bound in vitro by the Buf and Ume6 repressor proteins, we demonstrate for the first time by electrophoretic mobility shift assays and interference footprinting that the UASH site interacts in vitro with a novel factor called UBF (UASH binding factor) which is present in haploid and diploid cycling, as well as sporulating cells. Point mutations in the HOP1 UASH motif abolish UBF-dependent DNA binding activity in vitro and meiotic HOP1 gene expression in vivo. Furthermore, we show that UBF binds in vitro to UASH-like sequences in the promoter regions of several meiosis-specific and non-specific genes and propose that UBF mediates gene expression through its interaction with the UASH motif in both cycling and sporulating cells.

Anatomy of a transcription factor important for the start of the cell cycle in Saccharomyces cerevisiae

Entry of yeast cells into the mitotic cell cycle (Start) involves a form of the CDC28 kinase that associates with G1-specific cyclins encoded by CLN1 and CLN2 (ref. 1). The onset of Start may be triggered by the activation of CLN1 and CLN2 transcription in late G1 (ref. 2). SWI4 and SWI6 are components of a factor (SBF) that binds the CACGAAAA (SCB) promoter elements responsible for activation in late G1 of the HO endonuclease, CLN1 and CLN2 genes. A related factor (MBF) containing SWI6 and a 120K protein binds to the ACGCGTNA (MCB) promoter elements responsible for late G1-specific transcription of DNA replication genes. Nothing is known about how these heteromeric proteins bind DNA. We show here that SWI4 contains a novel DNA-binding domain at its N terminus that alone binds specifically to SCBs and a C-terminal domain that binds to SWI6. SWI4's DNA-binding domain is similar to an N-terminal domain of the cdc10 protein that is a component of an MBF-like factor from Schizosaccharomyces pombe and is required for Start. An involvement of this kind of DNA-binding domain in transcriptional controls at Start may therefore be a conserved feature of eukaryotic cells.

The DNA binding and oligomerization domain of MCM1 is sufficient for its interaction with other regulatory proteins

The MCM1 gene encodes an essential DNA binding protein that, in cooperation with the transactivators alpha 1 and STE12 and the repressor alpha 2, confers mating specificity to haploid yeast cells. We show that the amino-terminal third of the MCM1 protein is sufficient for the physical interaction with these factors. A strain expressing just 98 amino acids encompassing the oligomerization and DNA binding domains of MCM1 is viable and mating competent. This motif exhibits considerable similarity to a domain of the mammalian transcription factor SRF. A 98 amino acid hybrid gene coding for the MCM1 DNA binding domain and SRF dimerization domain is sufficient for viability but not for the expression of mating type specific genes. In vitro binding studies suggest that a region of approximately 50 amino acids of MCM1 is essential for providing contacts with alpha 1, alpha 2 and STE12.

Reversible methylation and inactivation of marker genes in sequentially transformed tobacco plants

Doubly transformed tobacco plants were obtained following sequential transformation steps using two T-DNAs encoding different selection and screening markers: T-DNA-I encoded kanamycin resistance and nopaline synthase; T-DNA-II encoded hygromycin resistance and octopine synthase. A genetic analysis of the inheritance of the selection and screening marker genes in progeny of the doubly tranformed plants revealed that the expression of T-DNA-I genes was often suppressed. This suppression could be correlated with methylation in the promoters of these genes. Surprisingly, both the methylation and inactivation of T-DNA-I genes occurred only in plants containing both T-DNAs: when self-fertilization or backcrossing produced progeny containing only T-DNA-I, expression of the genes on this T-DNA was restored and the corresponding promoters were partially or completely demethylated. These results indicated that the presence of one T-DNA could affect the state of methylation and expression of genes on a second, unlinked T-DNA in the same genome.