Annotation and analysis of the mitochondrial genome of Coniothyrium glycines, causal agent of red leaf blotch of soybean, reveals an abundance of homing endonucleases

Coniothyrium glycines, the causal agent of soybean red leaf blotch, is a USDA APHIS-listed Plant Pathogen Select Agent and potential threat to US agriculture. Sequencing of the C. glycines mt genome revealed a circular 98,533-bp molecule with a mean GC content of 29.01%. It contains twelve of the mitochondrial genes typically involved in oxidative phosphorylation (atp6, cob, cox1-3, nad1-6, and nad4L), one for a ribosomal protein (rps3), four for hypothetical proteins, one for each of the small and large subunit ribosomal RNAs (rns and rnl) and a set of 30 tRNAs. Genes were encoded on both DNA strands with cox1 and cox2 occurring as adjacent genes having no intergenic spacers. Likewise, nad2 and nad3 are adjacent with no intergenic spacers and nad5 is immediately followed by nad4L with an overlap of one base. Thirty-two introns, comprising 54.1% of the total mt genome, were identified within eight protein-coding genes and the rnl. Eighteen of the introns contained putative intronic ORFs with either LAGLIDADG or GIY-YIG homing endonuclease motifs, and an additional eleven introns showed evidence of truncated or degenerate endonuclease motifs. One intron possessed a degenerate N-acetyl-transferase domain. C. glycines shares some conservation of gene order with other members of the Pleosporales, most notably nad6-rnl-atp6 and associated conserved tRNA clusters. Phylogenetic analysis of the twelve shared protein coding genes agrees with commonly accepted fungal taxonomy. C. glycines represents the second largest mt genome from a member of the Pleosporales sequenced to date. This research provides the first genomic information on C. glycines, which may provide targets for rapid diagnostic assays and population studies.

Introduction Coniothyrium glycines (R.B. Stewart) Verkely & Gruyter is a soilborne pathogen that infects soybeans and the perennial soybean, Neonotonia wightii, causing lesions on foliage, petioles, pods and stems and eventual defoliation and premature senescence [1]. C. glycines produces melanized sclerotia that can germinate to either form infectious mycelia or produce pycnidia that in turn produce infectious conidia. The pathogen is spread locally via rain/water splash and human or animal movement, which scatter sclerotia and conidia onto neighboring plants. Leaf drop of infected leaves delivers sclerotia and pycnidia to the soil where they serve as sources of secondary inoculum. Sclerotia may also remain in the soil and restart the cycle of infection in the next growing season. There is no evidence that the fungus is seed-borne, but spread might occur from infected plant debris mixed in with untreated seed or through movement of contaminated soil.
The disease red leaf blotch (RLB) occurs predominantly in central and southern Africa [2] and the incidence of the disease has increased concomitantly with increased soybean production in regions where the pathogen is found. Yield losses of up to 50% have been reported in Zambia and Zimbabwe [3] [4]. While it does not currently occur within the United States, the ability of sclerotia to survive high temperatures and dry conditions suggest it could survive in soybean growing regions of the southern United States [5]. As a result, the Secretary of Agriculture has determined that C. glycines poses a significant risk to U.S. agriculture, and the pathogen is listed by USDA-APHIS as a Plant Pathogen Select Agent under 7 CFR, part 331 [6] [7]. Additionally, while C. glycines has been found to naturally infect only soybean and N. wightii, there is no evidence as to the pathogen's potential ability to infect other leguminous species, such as cultivated peanut and native, wild legumes that occur in the USA.
In the early stages of disease development, RLB may not be readily distinguished from other foliar soybean diseases such as Alternaria leaf spot, brown spot, or target spot. Current methods to identify C. glycines require time-consuming examination of morphological characteristics and temperature requirements. No molecular diagnostic assay currently exists to identify C. glycines. The examination of genomic sequences such as the mtDNA may provide targets for the development of diagnostic tools and also may provide insight into the mechanisms of disease resistance.
Phylogenetic analysis of the mtDNA will also be useful to clarify the taxonomy of this fungus. RLB was first observed on soybean in Ethiopia in 1955 and, based on the morphology of the pycnidial state, the causal fungus was identified as Pyrenochaeta glycines [8]. In 1964, Dactuliophora glycines was described as the cause of a leaf spot disease [9], and was subsequently identified as the sclerotial state of P. glycines [10]. Hartman and Sinclair [1] established the genus Pyrenochaeta to accommodate these synanamorphs. The fungus was re-classified as Phoma glycinicola in 2002 based on morphological characteristics [11] [12], and most recently was again re-classified as Coniothyrium glycines (R.B. Stewart) Verkely & Gruyter based on sequence analysis of regions of the ITS, SSU, LSU [13]. The mt genomes of only eight other members of the class Dothidiomycete, which includes several economically important plant pathogens such as the wheat pathogen, Stagonospora nodorum, and wheat leaf blotch, Zymoseptoria tritici (M. graminicola), can currently be found in GenBank. Six of these also share membership in the order Pleosporales with C. glycines. Comparison of the mt genome of C. glycines with the mt genome of these other eight fungi may help support or clarify the recent re-classification of C. glycines, as mitochondrial genomes are considered to be effective tools for evolutionary studies because they evolve independently of and at an accelerated rate from nuclear genomes [14][15] [16].
This study provides the complete mitochondrial genome of a pathogenic fungus identified as a USDA-APHIS Plant Pathogen Select Agent due to its potential impact on soybean production. Previously, the only genomic data available were specific sequences used in phylogenetic analysis of Phoma and Septoria spp [13] [17]. This sequence data may provide targets for the development of a rapid diagnostic assay and will help further clarify the evolving fungal taxonomy of the genus.

Materials and methods
Fungal isolate, library construction, and sequence assembly C. glycines-infected leaves were collected from soybean at the Rattray Arnold Research Station, Harare, Zimbabwe in March 2005 and shipped to the USDA-ARS Foreign Disease-Weed Science Research Unit at Fort Detrick, MD under Animal and Plant Health Inspection Service permit. Isolate Pg-21 was recovered from the leaves and maintained on 20% V8-juice agar at 20˚C in the dark. A 10% V8-juice broth was seeded with agar plugs containing mycelium of Pg-21 and grown for several weeks in the dark at 20˚C without shaking. Tissue was collected through vacuum filtration onto Whatman No. 1 filter paper in a Buchner funnel. Total DNA was extracted using the DNeasy Plant Mini kit (Qiagen, Germantown, MD). Culture identification was confirmed through sequencing of ITS fragments.
The mt genome was sequenced as part of a whole genome sequencing project with Illumina sequence libraries prepared using Nextera XT. Whole genome 2×300 paired-end sequencing was performed using Illumina MiSeq instrument. Reads were filtered and trimmed using Trimmomatic v.0.32 [18]. The iMetAMOS pipeline v. 1.5 [19] was used to optimize de novo assembly and perform quality checks. Elements of the pipeline include FastQC v. 0. 10 [24]. Resulting assemblies were polished using Pilon v. 1.8 [25]. Samtools v. 1.1 [26] and BLAST were used to remove low coverage and contaminating contigs. Initial shotgun assembly produced 1431 contigs greater than 1kb in size, with a median size of 11kb and median depth of coverage of 274X. Contig 76 was identified as an outlier with a size of 98,482 bp and average depth of coverage of 1542X. Discontinuous MegaBLAST searches revealed homology with fungal mt genome sequences. Finishing of the mt sequence was performed using CLC Genomics Workbench Genome Finishing Module (Qiagen, Germantown, MD), mapping raw Illumina reads back to contig 76, correcting assembly errors, and extending the contig ends.

Sequence annotation
The MFannot tool (http://megasun.bch.umontreal.ca/cgi-bin/mfannot/mfannotInterface.pl) was used to annotate the mt genome using genetic code 4 [27]. Annotation of open reading frames (ORFs) was reviewed and revised by BLAST homology searches against the NCBI protein database [28]. tRNAs were further evaluated against output from tRNAscan-SE [29], Dogma (Dual Organellar GenoMe Annotator) [30], and ARAGORN [31]. RNAweasel was used to classify identified introns as group I or group II introns [32]. Repeats were identified and analyzed with the Tandem Repeats Finder [33] and Palindrome and Einverted EMBOSS programs [34]. Codon usage for concatenated ORFs of twelve protein-coding genes was determined using the codon usage tool at http://www.bioinformatics.org/sms2/codon_usage.html with genetic code 4 [35]. The physical map of the Coniothyrium mtDNA was constructing using SnapGene Viewer (GSL Biotech; available at snapgene.com). The complete mt sequence of C. glycines isolate Pg-21 has been deposited in GenBank under the accession number MH337273.

Phylogenetic analysis
Amino acid sequences of the twelve protein-coding genes shared in common among 25 fungal mt genomes were each aligned with MUSCLE from EMBL-EBI [36], and amino acids sharing low similarity were removed by Gblocks [37]. Sequences were concatenated using Seaview [38]. A maximum likelihood tree of aligned sequences was constructed with PhyML 3.0 using LG as the evolutionary model [39]. Branch support was assessed using the PhyML default of aLRT test (SH-Like).

Gene content and genome organization
The mt genome of C. glycines is a circular molecule with a length of 98,533 bp (Fig 1). The sequence is AT-rich with an overall G + C content of 29.01%, and 28.9% in the coding regions of the protein-coding genes. The RNA genes had a higher GC content of 35.1% while the intergenic spacers had a lower GC content of 24.8%.
Protein-coding genes of the mt genome included one gene encoding for ATP-synthase complex F0 subunit (atp6), three cytochrome oxidase subunits (cox1, cox2, cox3), seven nicotinamide adenine dinucleotide ubiquinone oxireductase subunits (nad1-6, nad4L), cytochrome b (cob), one ribosomal protein (rps3), and four hypothetical proteins (orf208, orf284, orf929, and orf1407) (Fig 1 and Table 1). The mt genome also encodes for small and large subunit ribosomal RNAs (rns and rnl) and 30 tRNAs (Fig 1 and Table 1). Genes were transcribed from both DNA strands. The cox1 and cox2 genes were adjacent to each other with no intergenic spacers. Similarly, nad2 and nad3 were adjacent with no intergenic spacers and nad5 is immediately followed by nad4L with an overlap of one base (Fig 1 and Table 1).
Within the intergenic spacers, four open reading frames (orf208, orf284, orf929, and orf1407) were found (Fig 1 and Table 1). Putative functions could be assigned to three of the ORFs: orf1407 encodes a putative DNA polymerase type B, orf929 encodes a putative DNAdependent RNA polymerase, and orf208 encodes a putative GIY-YIG endonuclease protein. All three showed similarity to relevant sequences in other fungi and possessed conserved domain motifs. Only orf284 contained no conserved motifs and could not be assigned a putative function, but showed similarity to hypothetical proteins from whole genome shotgun sequencing of Bipolaris maydis and B. zeicola. An additional GIY-YIG endonuclease motif was identified in the intergenic spacer between the rnl and atp6. This region showed similarity to endonucleases from other fungi, however no clear ORF could be identified suggesting that this may represent a degenerate endonuclease. Only 14.4% of the mt sequence is comprised of intergenic spacers.
Within the intergenic spacers, 10 perfect or near identical tandem repeats were identified ranging in size from 12-62 bp and with 2-5 copies (S1 Table). In addition, fifteen palindromes were identified ranging in size from 10-15 bp. A single inverted repeat of 30 bp was found.

Introns
Introns made up 54.1% of the mt genome with a total of 32 introns identified within 8 of the protein-coding genes and the rnl (Fig 1 and Table 2). Thirty of the introns were classified as group I introns. One intron was classified as a group II intron (intron3 of the rnl) and one intron could not be definitively classified (intron2 of cox2). Eighteen of the identified introns were determined to contain putative intronic ORFs with either GIY-YIG or LAGLIDADG homing endonuclease (HE) motifs. An additional eleven introns showed evidence of truncated or degenerate HE motifs and one possessed degenerate N-acetyl-transferase domains. Only two introns had no identifiable ORFs and BLAST analysis revealed no homology in the NCBI  protein database. All putative HEs showed significant similarity to those found in the mt genomes of other fungi and most were identified in other members of the Pezizomycotina subphylum. However, each was unique within C. glycines, showing no similarity to other intronic ORFs within the mt genome.
The cox1 gene was the most common site for intron insertion, possessing ten of the 32 identified introns. Each of the ten introns also possessed either complete or degenerative putative HEs. Of these ten, only five were found to have high sequence identity to annotated introns found in the same location in the cox1 gene of the other Pleosporales. However, no other member of the Pleosporales possessed all five introns in common. The GIY-YIG HE of intron1 of cox1 showed 87% and 88% nucleotide identity to the corresponding introns of D. pinodes and P. chartarum, respectively. However, there was not a corresponding HE in the mt genomes of the other four Pleosporales species. Likewise, cox1 intron4, containing a LAGLIDADG HE, showed 88% nucleotide identity to the corresponding intron in B. cookei, but was found in no other Pleosporales species. The remaining five introns showed varying degrees of identity with introns from the mt genomes of more distantly related fungi, such as intron8 which showed 85% nucleotide identity with an intron from the corresponding location in Sclerotinia sclerotiorum (S2 Table).
The 2041-bp intron2 of cox1 has two regions with partial LAGLIDADG HE domains that showed 95-97% nucleotide identity with the 1208bp intron that occurs in the same position in the cox1 gene of D. pinodes. However, the central 1200 bp region of cox1 intron2 possessed a truncated GIY-YIG HE domain with no significant nucleotide similarity to any other fungus (S2 Table). This central region does show amino acid identity with a GIY-YIG HE located within an intron from the cob gene of the more distantly-related Chrysoporthe deutercubensis (Table 2).
While most introns showed nucleotide identity with introns inserted into the same gene in other fungi, nad4L intron1 shared identity with free standing orfs in S. sclerotiorum and P. nodorum. One intron, nad1 intron2, showed no nucleotide identity with other species from the Ascomycota, but rather showed identity with introns from two members of the Basidiomycota. This intron showed identity with an intron from the nad1 gene of Moniliophthora roreri and an intron from the cox1 gene of Fomitopsis palustris.

Codon usage and tRNA genes
Codon usage, summarized in S3 Table, shows a bias towards AT-rich codons, which reflects the high AT content of the C. glycines mt genome. Most protein coding genes start with the canonical translation initiation codon ATG with the exception of cox2 and orf1407, which appear to utilize UUA and AUA start codons, respectively. The preferred stop codon in the mt genome was TAA, occurring in 12 genes. The alternative stop codon TAG occurs in 3 genes. A traditional stop codon could not be identified for cox1. This absence, combined with the location of cox1 adjacent to cox2 with no intergenic spacers, suggested the possibility of a fused cox1-cox2 polyprotein rather than two separate proteins. Thirty tRNAs were identified and twenty of them occurred in two large clusters around the rnl, while five occurred singly between mt genes (Fig 1). The tRNAs occurred on both DNA strands.

Comparative genomics and phylogenetic analysis
Comparison of the mt genome of C. glycines with those from eight other members of the Dothidiomycetes revealed that in all nine species genes are encoded on both mtDNA strands. Comparison also found some conservation of gene order, most notably within the Order Pleosporales (Fig 2). In all nine species, nad4L and nad5 were adjacent, and in all but P. nodorum there are no intergenic spacers but rather a one base pair overlap between the two genes. Within C. glycines and the six members of the Pleosporales, cox1 and cox2 were also adjacent with no intergenic spacers. Three members of the Pleosporales possess a conserved gene block of nad5, nad4L, nad3, and nad2. C. glycines shows the same gene order, however the block is disrupted by insertion of cob, nad1, and nad4 between nad4L and nad3. C. glycines and the other Pleosporales species also lack the atp8 and atp9 genes which are typically found in fungal mt genomes, while both Capnodiales species possess both genes. All nine species also exhibit large clusters of tRNA genes around the rnl, and within the Pleosporales tRNA order is maintained as well. The conservation of gene and tRNA order is expanded among the Pleosporales, with six of the seven possessing a nad6-rnl-atp6 gene block with associated conserved tRNA cluster patterns (Table 3). P. chartarum possesses a similar gene block and tRNA cluster pattern, but the atp6 is displaced relative to the other Pleosporales. This conservation of tRNA gene order is carried to a lesser extent to the Capnodiales. Comparison of the mt genome of C. glycines and the other Dothidiomycetes with those of an additional 19 ascomycetous fungal species revealed several potentially distinguishing characteristics of this class. Of the 25 mt genomes compared, fifteen carry all genes on the same strand of DNA and an additional four mt genomes show the core coding genes encoded on the same strand with only tRNAs or hypothetical proteins encoded in the opposite direction (S4 Table). However, all nine members of the Dothidiomycetes contain genes distributed on both mtDNA strands. Also, while ribosomal protein S3 or S5 occurs within an intron of the rnl in 17 of the 25 species examined, among the Pleosporales rps3/rps5 occurs as a free standing ORF and the gene appears to be absent from the two Capnodiales species (Table 4). b Refers to the 14 conserved protein coding genes typical of fungal mitochondrial genomes: 11 genes encoding subunits of respiratory chain complexes (cob cox1, cox2, cox3, nad1, nad2, nad3, nad4, nad4L, nad5, and nad6) and 3 ATP synthase subunits (atp6, atp8 and atp9). c Ribosomal protein S3 or S5, when present, occurs as an intronic orf within the rnl of all above mt genomes with the exception of C. glycines, D. pinodes, P. nodorum, P.
subalpina, P. amphalodes, P. chartarum, S. bambusicola, and S. lycopersici. d The ribosomal proteins S3 or S5 were not annotated in the available sequences, but were putatively identified by blastx analysis against the non-redundant protein database. e Ribosomal RNAs were not annotated in the available sequences, but were putatively identified by blastn analysis against the rnl and rns of other fungal mt genomes. Additionally, while atp8 and atp9 are absent from the Pleosporales species, both are found in the other species with the exception of Pseudogymnoascus pannorum which lacks only atp9 ( Table 4). The proximity of cox1 and cox2, also characteristic of the Pleosporales examined to date, is not apparent among the other ascomycetous species. Several similarities across the species were revealed as well. The G+C content is consistent among all species, ranging from 23-32%, with the exception of Pyronema omphalodes with 43%, and all show some tRNA clustering around the rnl. In all but four species, nad4L and nad5 are adjacent with either no intergenic spacer or a single base pair overlap (S4 Table).
The size of the mt genome and the presence of introns varies across all species, ranging from 23743 bp in Z. cellare with no introns to 203051 bp in Sclerotinia borealis with 61 introns. In general, a larger number of introns is reflected in a larger genome size (Table 4). Among the Pleosporales, S. bambusicola has the smallest mt genome at 39030 bp, of which only 3.2% is comprised of the one intron identified. [40] P. nodorum (49761 bp) contains five introns, which make up 13% of the mtDNA [16], while D. pinodes (55973 bp) contains 14 introns, making up 26% of its mt genome size (NC_029396). Within C. glycines, the 32 identified introns comprised 54% of total mt genome size.
A phylogenetic tree was built with twelve protein-coding genes in common from 25 fungal species (Fig 3). This tree agrees with commonly accepted fungal taxonomy and supports the placement of C. glycines among the Pleosporales and recent reclassification to its own family, the Coniothyriaceae. [13] Discussion This research provides the first genomic information on the USDA APHIS-listed Plant Pathogen Select Agent C. glycines; data which may provide targets for rapid diagnostic assays and population studies. Additionally, C. glycines represents the second largest mt genome from a member of the Pleosporales sequenced to date. Mitochondrial genome size among fungi varies greatly from the smallest, Rozella allomyces, at 12055 bp [41] to the largest, Rhizoctonia solani, at 235849 bp [42]. At 98,533 bp, C. glycines is of larger than average size and only 23 other currently available fungal mt genomes are larger. Among the fungi there is no correlation between mtDNA size and gene content.
The gene content of fungal mt genomes is largely conserved. However, it is notable that C. glycines lacked two of the core set of genes typical of fungal mt genomes: atp8 and atp9. These two genes were also absent from the mt genomes of other Pleosporales species [16] [40]. While gene content may be conserved, gene order is not equally conserved and relative gene order varies both between and within major fungal phyla [43][44] [45]. Alignment of the C. glycines mt genome with other members of the Dothidiomycetes identified a lack of synteny in gene order and gene orientation. However, limited conserved gene blocks were observed. The uninterrupted gene pairs of nad2-nad3 and nad4L-nad5 occurred in all nine Dothidiomycetes species, while the pairing of cox1-cox2 occurred only within all seven Pleosporales species and not the two Capnodiales species. Additionally, nad1-nad4 remain coupled in only three species from the Pleosporales. A conserved gene block nad2-nad3 and nad4L-nad5 was identified among three of the Pleosporales, but within the C. glycines mt genome this block is interrupted by three other genes. However, six of the seven Pleosporales species showed an atp6-rnl-nad6 conserved gene block, which included two large clusters of tRNAs on either side of the rnl in a relatively conserved pattern. Additionally, protein-coding and tRNA genes of C. glycines and the eight other Dothidiomycetes are encoded on both mtDNA strands, while the majority of ascomycetes species examined here have genes encoded on a single DNA strand. The pattern of gene order in mt genomes may provide a road map to trace the evolutionary route of fungal taxonomy. As additional species from the Dothidiomycetes, and the Pleosporales specifically, are analyzed, the additional mt signals will indicate if conserved gene blocks identified to date are characteristic of the Order Pleosporales and further help elucidate fungal taxonomy. Comparative genomics and phylogenetic analysis presented here supports the placement of C. glycines within the Pleosporales and its recent reclassification to its own family, the Coniothyriaceae [13].
With gene content being largely conserved, the size variation evident among fungal mt genomes is instead attributable to variations in the structure and size of intergenic spacers and the number and size of introns [46][47] [48]. The larger than average mt genome size of C. glycines was attributed to the relatively high number of introns identified, with 32 introns comprising over half of the total mt genome size. This abundance of introns, most of which possess complete or degenerate HEs, may also provide valuable tools for the evaluation of evolutionary history and intron mobility [49] [50][51] [52][53] [54]. While the cox1 gene is considered the most common insertion site for group I introns in fungal mt genomes, the number of introns inserted varies widely from zero in some fungi to the fourteen identified in Podospora anserina [55]. The present study of C. glycines found five of ten cox1 introns, which all possess either complete or truncated HE domains, shared high sequence identity with corresponding introns from the six other Pleosporales species annotated, suggesting common ancestral origin. However, it is notable that none of these five putative HEs occurred in all seven Pleosporales species. The remaining five cox1 introns showed varying degrees of identity with introns from the mt genomes of more distantly related fungi. For example, cox1 intron8 contained a GIY-YIG HE that showed 85% nucleotide identity with an intron from the corresponding location in S. sclerotiorum of the Helotiales and intron5, with its LAGLIDADG HE, shared 71% identity with an intron from Lachancea mirantina, a member of the Saccharomycotina subphylum (S2 Table). The similarity to HEs from more distantly related fungi suggest possible acquisition through horizontal transfer rather than retention from a common ancestor. Additional evidence of horizontal transfer comes from nad1 intron2 and its LAGLIDADG HE which showed no nucleotide identity with introns from other species of the Ascomycota, but rather showed identity with introns from two distantly related members of the Basidiomycota.
The examination of cox1 HEs also revealed evidence of multiple insertion events during the course of evolution. While cox1 intron2 possessed end regions with truncated LAGLIDADG domains and high nucleotide identity to a single orthologous intron from D. pinodes, the central region of this intron, with a truncated GIY-YIG domain, showed only amino acid similarity to an intron from the cob gene of the more distantly-related C. deutercubensis, suggesting the insertion of a new sequence into an already present HE.
It is difficult to determine the precise roles that intron retention, intron acquisition through horizontal transfer, and intron loss have played in constructing the C. glycines mt genome as it has been annotated here. The question remains if some fungal lineages possess a mechanism by which they accumulate and retain HEs while other fungal lineages appear to have lost all introns, and what that mechanism might be. However, this analysis of HEs does suggest that a complex pattern of insertions and horizontal transfers of introns are responsible for the relatively large mt genome size of C. glycines.
Supporting information S1  Table. Gene order of the fungal mt genomes used for comparative genomics and phylogenetic analysis.
(XLSX) Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture.
The U.S. Department of Agriculture (USDA) prohibits discrimination in all its programs and activities on the basis of race, color, national origin, age, disability, and where applicable, sex, marital status, familial status, parental status, religion, sexual orientation, genetic information, political beliefs, reprisal, or because all or a part of an individual's income is derived from any public assistance program. (Not all prohibited bases apply to all programs.) Persons with disabilities who require alternative means for communication of program information (Braille, large print, audiotape, etc.) should contact USDA's TARGET Center at (202) 720-2600 (voice and TDD). To file a complaint of discrimination, write to USDA, Director, Office of Civil Rights, 1400 Independence Avenue, SW, Washington, DC 20250-9410 or call (800) 795-3272 (voice) or (202) 720-6382 (TDD). USDA is an equal opportunity provider and employer.