In 1976, the cytogenetic analysis of six azoospermic individuals mapped the AZF (azoospermia factor) locus, which is necessary for normal spermatogenesis, to the distal long arm euchromatin of the human Y chromosome.¹ It was not until 1992 that molecular analysis confirmed these findings with the discovery of two interstitial deletions situated in this region in azoospermic men.² In 1996, a third, independent, deletion interval was identified and the three intervals were named AZFa, AZFb, and AZFc.³ Subsequent studies have demonstrated that deletions affecting the AZFc interval are by far the most common, affecting approximately 1 in 10 men with idiopathic azoospermia.^4–6

The first AZFc candidate gene to be isolated from the AZFc interval was DAZ (deleted in azoospermia).⁵DAZ originated on the Y chromosome as a transposition of a DNA segment containing the autosomal germ-cell specific gene DAZL, and comparative analyses in primates show that the DAZ transposition occurred ∼35 million years ago in a common ancestor of old world monkeys and humans.⁷ There is at least one other functional gene within the AZFc interval: CDY1, a retroposon that codes a protein with proven histone acetyltransferase activity.⁸ Although found only in primates, CDY1 is believed to have been retroposed directly into MSY (male specific region of the Y chromosome) from a transcript of the autosomal gene CDYL more than 150 million years ago, making it one of the oldest genes on the Y chromosome.⁹ At present, DAZ and CDY1 are the only transcription units in the AZFc interval for which there is any evidence that function has been selectively maintained on the Y chromosome.

In almost all cases, deletions of the entire AZFc interval are associated with infertility characterised by a drastic reduction in sperm count due to the loss of active spermatogenesis from most testicular tubules. Spermatogenesis is completed, but on a very small scale. Most AZFc deleted men are severely oligozoospermic or azoospermic, but reported sperm counts range from 0 to 7 million per millilitre. There are, nevertheless, rare AZFc deleted men who have conceived multiple children naturally.^10–12 All of the sons of these men are infertile. The few spermatozoa produced by most AZFc deleted men can be successfully used for in vitro fertilisation by ICSI (intracytoplasmic sperm injection). Although AZFc deleted children conceived by ICSI are generally healthy, it is important to identify the AZFc genes involved, as this will improve the diagnosis of genetic causes of male infertility, and, thus, allow an informed evaluation of the health risks faced by an ICSI conceived child.

Despite the high frequency of the AZFc microdeletion, the key genes have not yet been identified. The detection and characterisation of smaller deletions within the AZFc interval are required, in order to understand how the AZFc interval contributes to human fertility. Initial evidence that partial AZFc deletions are present in a small percentage of infertile men came from several groups focusing on either sequence family variant (SFV) or fluorescent in situ hybridisation (FISH) analysis of the DAZ gene.^13–15 Recently, using locus specific PCR assays and dual colour FISH, a common class of partial deletion was described and was named the gr/gr deletion (g1/g2,r1/r3,r2/r4).¹⁶ The gr/gr deletion is predicted to result in the loss of some, but not all, copies of all AZFc genes.¹⁷ In that study, gr/gr deletion was found in 3.8% of infertile men with a sperm count of <5×10⁶/ml, and 2.2% of fertile men. However, its absence from a group of 148 normospermic men led to the conclusion that the gr/gr deletion is associated with decreased sperm production resulting from the reduced dose of AZFc genes.¹⁶ Here, we present our characterisation of the AZFc interval in 300 infertile and 399 control men. Using quantitative and qualitative assays, we identified 34 cases of partial AZFc deletion: 32 lack some copies of both DAZ and CDY1, while two lack some copies of either DAZ or CDY1.

METHODS

Infertile men and controls

We studied five different groups: (i) 254 individuals with unexplained infertility and reduced or absent sperm counts of 0–5×10⁶ spermatozoa (spz)/ml (composed of 94 individuals with 0–0.1×10⁶ spz/ml, 73 with >0.1–1×10⁶ spz/ml, and 87 with >1–5×10⁶ spz/ml); (ii) 46 individuals with unexplained infertility and sperm counts of 5–131×10⁶ spz/ml; (iii) 210 unselected men of unknown fertility status; (iv) 185 fertile men with one or more children; and (v) four normospermic men. Samples were collected using approved protocols and the informed consent of all individuals was obtained. The required approval for this study was obtained from the French committee for the Protection of Persons in Biomedical Research (CCPPRB). The infertile men were drawn from a larger group of 361 individuals recruited from two infertility clinics in Marseilles and one in Toulon, France. For all patients, urogenital examination, karyotype Y microdeletion status, and sperm parameters, according to the criteria of the WHO (World Health Organization), were determined. These analyses allowed us to exclude 61 individuals: 28 obstructive azoospermic cases (including 11 cases of bilateral agenesis of the vas deferens associated with mutations in both alleles of CFTR), 13 47,XXY and one 47,XXY/46,XX/46,XY Klinefelter cases, three 47,XYY men, 10 unrelated AZFc deletions, four AZFb+c deletions, one partial AZFb deletion, and one AZFa deletion.

The control samples are mostly from local sources, but include 57 unrelated fathers from CEPH (Centre d’étude du polymorphisme humain) reference families. The unselected men of unknown fertility were anonymous blood donors from the local blood transfusion service (Etablissement Français du Sang Alpes Mediterranée). The proportion of Y haplogroups in the CEPH fathers is similar to our local populations, except for Hg-DE, which is more frequent in local controls (9.5%–19% local v 5% CEPH). The CEPH father, 142001, with the gr/gr deletion has seven children.

Sequence family variant (SFV) analysis

We screened for SFVs in DAZ and CDY1. For DAZ, we chose the SFV at the sequence tagged site sY587 in intron 10,¹⁸ also referred to as “DAZ-SNV V”.¹⁵ The sY587 SFV discriminates DAZ1/2 from DAZ3/4, and its position outside the intragenic duplications that affect exons 2–8 means that sY587 is present in only one copy per DAZ gene. To confirm the sY587 results, we also sequenced all men with a partial AZFc deletion for an SFV 93 bp upstream from exon 2 of the DAZ gene. This SFV also differentiates DAZ1/2 (A) and DAZ3/4 (G). The results agree with those obtained for sY587. Since the Y chromosome reference sequence shows no SFV within the transcribed portion of the CDY1 gene, we used a C/A SFV situated 7750 bp 5′ of the CDY1 translation start codon (CDY1-7750). We also used an SFV that we discovered in the 5′ UTR of CDY1 when sequencing CDY1 in our patients with partial AZFc deletions, and it was the only polymorphism that we discovered (MJM and EC, unpublished data). SFVs were scored by PCR followed by restriction enzyme digestion (2 units for 2 h): DAZ sY587, DraI (DAZ1/2 cut); CDY1-7750, PvuII (CDY1b cut); and CDY1-84, AluI (CDY1a cut). To prevent cross-amplification of CDY2, the CDY1-84 SFV was scored by two rounds of PCR, the first being specific for CDY1, followed by digestion. The SFV DAZ exon2–93 was scored by sequencing. Primer pairs: sY587, o912/o913; CDY1-7750, o1025/o1026; DAZ exon2-93, o1226/o1227; CDY1-84, 1° o1187/o1188 (30 cycles), 2° o1187/o1195 (10 cycles).

Cases of SFV absence were confirmed by sequencing, and no additional SNPs were detected.

Fluorescent in situ hybridisation

We performed FISH using the DAZ cosmid 18E8¹⁸ to probe interphase nuclei from lymphoblastoid lines. In the majority of interphase nuclei, which are at G1, 18E8 gives a single hybridisation signal for each DAZ gene pair present. In most men, where there are two DAZ gene pairs, cosmid 18E8 gives two signals on the majority of interphase nuclei.¹⁴

FISH was performed on interphase nuclei obtained using standard cytogenetic techniques on non-synchronised lymphoblastoid cell lines. The DAZ cosmid 18E8 was labelled with biotin-14dCTP (Bioprime DNA labelling system, Invitrogen, France), hybridised using a standard protocol and revealed by means of fluorescein isothiocyanate-conjugated avidin. Over 100 nuclei were screened for each patient. Nuclei were counterstained with propidium iodide diluted in anti-fade solution. Preparations were observed using an axioplan-2 Zeiss fluorescent microscope and the images were captured with a SenSys CCD camera (Photometrics, Tucson, AZ, USA).

PCR based quantification of copy number

In order to quantify copy number using PCR, we simultaneously amplified the AZFc locus to be quantified and a homologous locus outside the AZFc interval, as an internal standard, using a single primer pair in a standard PCR reaction of 28–32 cycles. The primers flank an insertion/deletion difference of 3–5 bp, which allowed the products amplified from the AZFc loci and the control loci to be separated by polyacrylamide gel electrophoresis. One of the primers was labelled at its 5′ end with a fluorochrome (IRDye 800, LI-COR, Lincoln, NE, USA) and the other carried a GTTTCTT tag at its 5′ end. This short tag has been reported to stimulate the Taq polymerase to add an adenosine residue to the end of the complementary (labelled) strand, resulting in the adenylation of almost all labelled fragments.¹⁹ This overcame the problem of doublet bands caused by heterogeneous adenosine addition, which had initially made quantification difficult. Following PCR, the reaction was mixed with 95% formamide, denatured at 95°C for 5 min, and the different sized loci separated on an automatic sequencer Gene ReadIR 4200 (LI-COR). Quantification was performed by ONE-DScan (Scanalytics, Fairfax, VA, USA).

We developed five quantitative tests of this type. The positions of the loci screened are shown in figs 1 and 2.

CDY1-cds v CDY2 (primers oMJ953a/o1023)

There are two identical copies each of CDY1 and CDY2, which share 98% nucleotide identity.²⁰CDY2 is not affected by the complete AZFc deletion. We amplified CDY1 and CDY2 across a 3 bp indel difference in the coding region, to give fragments of 134 bp for CDY1 and 137 bp for CDY2. All men were screened with this test.

CDY1-9825 v CDY2 (primers o1207/o1208)

A further CDY test that was used to screen all cases of SFV loss at CDY1-7750.

DAZ v DAZL (primers o1130/o1313)

In order to dose DAZ, we co-amplified a fragment of intron 10 from DAZ and DAZL. This intron contains the SFV at sY587 and is present in one copy per DAZ gene. Thus, from a 46,XY DNA with an intact AZFc interval, we would expect to amplify four copies of DAZ for every two copies of DAZL (66%:33%). The expected fragment sizes are 184 bp for DAZ and 187 bp for DAZL. When we performed the test, an extra band at 230 bp was observed in 144 out of 575 individuals. We determined, by sequencing, that this fragment results from a 40 bp insertion polymorphism in the DAZL intron 10, since in a given individual it has the same intensity as the 187 bp DAZL fragment, which has half its expected intensity. Furthermore, there are eight individuals with the 230 bp band who lack the 187 bp DAZL band altogether, and their observed frequency (0.014) is close to the expected homozygote frequency (0.018) for a biallelic locus with a minor allele frequency of 0.134 (154 out of 1152 chromosomes tested).

MYPT2Y v MYPT2 (primers o796/o1083)

We derived a test for the ends of the palindrome P1. This locus is present in two copies on the Y chromosome, where the proximal copy is situated between the amplicons b3 and r2 (DAZ2) and the distal copy is situated distal to amplicon b4 and, thus, outside the AZFc interval. This locus is part of a block with 98% identity to the MYPT2 gene on chromosome 1. The gene on the Y chromosome is a pseudogene that we refer to as MYPT2Y-ps. This test amplifies MYPT2Y-ps (173 bp), together with MYPT2, on chromosome 1 (168 bp), and was used to screen patients with a suspected deletion.

Yp v Xq (primers o955/o1024)

This test is derived from the PCDHX and PCDHY genes on Yp and Xq. It controls for high levels of 46,XY/45,X mosaicism which, if present, could lead to false deletion detection by the other dose tests. This possibility was excluded for all individuals except two males of unknown fertility who do not have a partial AZFc deletion.

Locus specific PCR flanking DAZ1/2

We performed two locus specific PCR assays. We derived primers (oMJ1084/oMJ1085) from either side of the P2/P1 junction fragment situated 70 kb distal to the 3′ end of DAZ2. The presence of this site specific junction fragment was tested for in all groups. Only one fertile Hg-J man who had not previously shown evidence of deletion failed to amplify this fragment, indicating that there is a low frequency of Y chromosomes (<0.2%) which lack this evolutionary junction fragment. This test corresponds to sY1291.¹⁶

The junction situated proximal to DAZ1 is defined by a sequence block that is duplicated on Yp. There is, however, a deletion/insertion difference between the Yp and the AZFc homologue that allowed us to develop a specific PCR test for AZFc with primers oMJ1276/oMJ1277. This region is part of u3 on the published AZFc map⁴ and is situated close to the 50f2/C locus.²¹ It corresponds to sY1191¹⁶ or sY1192.⁴ To determine the frequency of deletion polymorphisms at this locus in our population, we tested 242 non-deleted men (132 Hg-P, 61 Pg-Y*{xD,E,J,P}, 20 Hg-J, and 29 Hg-DE). No deletions were found (data not shown).

PCR primer pairs

The 5′ to 3′ sequences of the unpublished oligonucleotide primers used in this study are: o796: ctcactacatgacattcagg; o912: tgtatttaaaatgtgcacttcactgt; o913: ccagtcacaaaatgccacat; o953a: tattgagacccttgcacctg; o955: aatttcttcacagccagacagc; o1023: gtttcttggagtttcccttctgtcacc; o1024: gtttcttgcactattcattgcacagac; o1025: gaaatgccataatgtgctaacactg; o1026: aaggagagtgttaatacataccctg; o1083: gtttcttcccagtatctagtacagtgc; o1084: agaactgccaggtctgtgtc; o1085: ttgccacaaagagatagctctgg; o1130: ttaagtactactgtagacac; o1187: gttcaggcacatggcatagc; o1188: ccactcacagttttgtggtc; o1195: cttcaacctcaaactcctgg; o1207: gtttcttccactgtagaaattcacctcc; o1208: gaagtttgcatagtggacagc; o1226: aactgagctgcactggtgac; o1227: ccccgaatgaccagcagcc; o1276: agtctgagtggctagtgcac; o1277: gaagcaaagttcagctgtgc; o1313: gtttcttgtataatgtagaagagtagagc.

The following annealing temperatures were used: o796/o1083: 56°C; o912/o913: 54°C; o953a/o1023: 58°C; o955/o1024: 54°C; o1025/o1026: 58°C; o1084/o1085: 60°C; o1130/o1313: 54°C; o1187/o1188: 60°C; o1187/o1195: 58°C; o1207/o1208: 56°C; o1226/o1227: 60°C; o1276/o1277: 60°C.

Y haplotyping

Y haplotyping was performed as previously published for the YAP,²² M9,^23,24 SRY4064,^25,26 and 92R7^25,26 polymorphisms. The Tat polymorphism²⁷ was assayed by PCR digestion with primers o1489: atgtatatagtacgtctgtagg and o1490: gtaagcataattgagaaggtgcc (annealing at 54°C). The 12f2 assay was performed as a duplex with SRY primers 3′SRY15 and 3′SRY16 and 12f2 primers 12f2D and 12f2F.²⁶ These primer pairs were used at a concentration of 300 nM for each SRY primer and 600 nM for each 12f2 primer and the annealing temperature used was 52°C. Polymorphisms were visualised by restriction enzyme digest for M9 (HinfI), SRY4064 (BsrBI), Tat (HpyCH4IV), and 92R7 (HindIII).

All men were haplotyped for YAP, 12f2, and 92R7 polymorphisms (table 1), and individuals with partial AZF deletions were further genotyped with the Tat, M9, and SRY-4064 SNPs, defining five haplogroups, E, J, K(xN3,P), N3, and P, and one paragroup Y*(xD,E,J,K).

Statistical analysis

We tested the significance of the observed difference in the incidence of gr/gr deletion or SFV absence between our infertile and control populations by means of Fisher’s exact test, using the program developed by Ø Langsrud (http://www.matforsk.no/ola/fisher.htm). Our null hypothesis was that incidence is the same in infertile and control populations. Given the exploratory nature of this study, all associations with p<0.05 are considered as significant, and worth further study. In this study, however, we have performed 11 tests of association with infertility: four DAZ+CDY1 gr/gr deletion types, loss of four SFV sites at DAZ and CDY1, loss of a CDY1-7750 SFV (a or b) with, and without, copy loss, and presence of a gr/gr deletion. Applying the Bonferroni correction for multiple testing (α = 0.05/number of tests performed), an α of 0.0045 is obtained. Thus, p<0.0045 is considered the true statistical significance for this study.

RESULTS

Screening for partial AZFc deletions

In order to investigate the incidence, extent, and impact on fertility of partial AZFc deletions in a large number of fertile and infertile individuals, we characterised the apparently intact Y chromosomes in five groups of men in four different ways: (i) quantification of five AZFc loci, including DAZ and CDY1, using quantitative PCR and FISH; (ii) detection of nucleotides (SFVs) that discriminate between members of the DAZ and CDY1 gene families; (iii) the specific detection of unique fragments flanking the DAZ1/2 gene pair at the u3 segment (proximal) and the P2/P1 palindrome junction (distal), corresponding, respectively, to sY1191/sY1192 and sY1291 used in previous studies^16,28; and (iv) haplotyping with MSY binary polymorphisms. Some of these results, and a summary of the Y chromosome variants defined in this study, are presented in figs 1–3.

Most partial AZFc deletions of DAZ or CDY1 are gr/gr deletions

This combination of tests was initially used to screen 487 men. All 487 men were screened with three quantitative tests: DAZvDAZL, CDY1-cds vCDY2, and Yp v Xq. This latter test controls for cases of 46,XY/45,X mosaicism. Some 18 men showed copy loss at both DAZ and CDY1, but not Yp. These men all showed SFV loss at both DAZ (sY587-[18]) and CDY1-7750 and were deleted for either one of the unique markers flanking the DAZ1/2 locus. We conclude that these 18 men carry partial AZFc deletions of the Y chromosome that correspond to the gr/gr deletions recently described.¹⁶

The original definition of the gr/gr deletion did not predict the deletion of the u3 segment.¹⁷ However a more recent study has shown that the Y chromosome of haplogroup Hg-N3 probably carries a gr/gr deletion.²⁸ The deletion of u3 was explained in this case by invoking the existence of an inversion around DAZ1/2, mediated by recombination between elements b2 and b3. Our results confirm this finding and, moreover, show that u3 deletion is not limited to Hg-N3 but is found on other types of Y chromosome (fig 2). We conclude that deletion of either u3 or the P2/P1 junction, but not both, is diagnostic of gr/gr deletion.

Only three men who did not show copy loss for DAZ or CDY1 showed SFV loss at both these loci. These men were, however, deleted for the u3 locus specific marker indicating that their Y chromosome carries a gr/gr deletion, but that the DAZ and CDY1 copy number has been restored by a subsequent b2/b4 duplication.¹⁶ We did not find any cases with SFV loss at sY587 in DAZ and CDY1 that were not also deleted for one of the unique sites, u3 or P2/P1, that flank DAZ1/2, suggesting that in most cases either of these approaches can be used to detect gr/gr deletions. Indeed, the loss of the DAZ3/4 variant at sY587 is diagnostic of gr/gr deletion, while the loss of the DAZ1/2 variant was found in only three undeleted men. Loss of the CDY1 SFV is a much less reliable indicator of deletion, since, of the 49 cases detected, only 22 have a partial AZFc deletion. Copy number status was verified at CDY1-7750 using a further quantitative test, CDY1-9825 v CDY2, and the results obtained at these two loci are in complete accord. This effectively excludes the possibility that small deletions, specific to the 5′ end of CDY1, are the cause of SFV absence at this site.

Two exceptional individuals showed deletions that affect either DAZ or CDY1, but not both: the normospermic man, 13-2968, (26×10⁶ spz/ml) is deleted for CDY1b only, while the fertile control, O/03, is deleted for the DAZ1/2 gene pair only. It is likely that O/03 is a second case of the b1/b3 deletion reported recently,¹⁶ but confirmation of this will require the mapping of his proximal boundary. The deletion carried by 13-2968 has not been previously reported, and shows that two copies of CDY1 are not required for normospermia.

Our results show that the vast majority of partial AZFc deletions that result in the loss of some copies of DAZ and CDY1 are gr/gr deletions, mediated by unequal crossing over between two large (∼540 kb) directly repeated gr amplicons (fig 2). We also show that our combination of SFV tests and locus specific tests provides a stringent method for their detection. We therefore screened a further 212 individuals with SFV and locus specific tests only. Quantitative tests were performed in gr/gr deletion cases, as a means of detecting the b2/b4 duplication. In this group we found a further 11 gr/gr deletions, one of which had a b2/b4 duplication.

All men in whom we detected a partial AZFc deletion were tested with a further quantitative test: MYPT2Y v MYPT2 situated at the end of palindrome P1. Copy loss at MYPT2Y was observed in men with a gr/gr deletion who are deleted for the P2/P1 junction, but not in cases of u3 deletion. Indeed, in the three men predicted to have a b2/b4 duplication, whose deletions must have arisen on a b2/b3 inverted Y chromosome, since they include u3 and not the P2/P1 junction, we saw increased copy number of the MYPT2Y-ps locus (shown for 13-560 and 13-230; fig 1B). The proximal copy of MYPT2Y-ps would not be deleted from a b2/b3 inverted Y chromosome, and would therefore be duplicated by a b2/b4 duplication, to give three copies instead of two. This provides further evidence for the existence of the b2/b3 mediated inversion polymorphism²⁸ and b2/b4 duplications.¹⁶

Our cumulative results are summarised in fig 2 and tables 1 and 2. Overall we detected 32 gr/gr deletions, four of which carry b2/b4 duplications. Based on SFV loss and Y haplogroup, we are able to discern 17 independent gr/gr deletions.

Quantification of DAZ gene pairs by FISH

FISH analysis with the DAZ cosmid 18E8 was performed on four infertile individuals without SFV loss at either sY587 or CDY1-7750 and on one fertile and six infertile individuals (** in fig 1B) with SFV absence at both these loci. This confirmed our PCR based analyses. Representative results are shown in fig 3.

One infertile man without SFV loss, 13-1877, showed 62% of nuclei with three or more signals, and only 35% of nuclei with two signals. We conclude that 13-1877 has three pairs of DAZ genes, and this was corroborated by PCR based dose tests (data not shown). A similar DAZ gene pair duplication has recently been described in a fertile sperm donor.²⁹ This duplication is most likely the result of the event reciprocal to the gr/gr deletion.

Loss of DAZ1/2 or DAZ3/4

Using sets of SFVs that discriminate between each DAZ gene, other studies have concluded that partial AZFc deletions can result in the loss of one, two, or three DAZ genes.^14,15 Our combined quantification and SFV analysis of the DAZ gene family in 487 men allowed us to address this question.

Out of 487 men, we identified 19 individuals with reduced DAZ copy number, who all lack the SFV at sY587. Since the SFV at sY587 differentiates DAZ1 and DAZ2 from DAZ3 and DAZ4, this shows that the vast majority of partial AZFc deletions remove either the DAZ1/2 gene pair or the DAZ3/4 gene pair, but not combinations of two DAZ genes from each gene pair. Since there are normally four DAZ genes and two DAZL genes, DAZ:DAZL ratios of 66%:33%, 50%:50%, or 33%:66% are expected, respectively, in cases where no, two, or three DAZ genes have been deleted. Since we never observed the 33%:66% ratio, we conclude that deletions of three DAZ genes are extremely rare. Although there is no obvious mechanism by which single DAZ gene deletions could arise, we cannot formally exclude their existence, since the expected difference in DAZ:DAZL dose ratios, 60%:40% as compared to 66%:33%, is too slight to measure reliably.

Inversion polymorphism in the DAZ3/4 palindrome P1

In the Y chromosome reference sequence,^20,30CDY1b lies outside the region that is predicted to be deleted by gr/gr recombination. We nevertheless detected 14 partial AZFc deletions that result in the loss of CDY1b, but that otherwise have all the characteristics of “classic” gr/gr deletions. These deletions involving CDY1b can, therefore, be explained either by invoking a second inversion polymorphism of the P1 palindrome, or the gene conversion of CDY1a by CDY1b prior to deletion. This latter possibility is very unlikely, as CDY1b deletions represent 14/32, or 48%, of the gr/gr deletions, and the observed frequency of this gene conversion event is <1%. It therefore seems likely that there is a further inversion polymorphism that may be mediated by crossing over between the homologous inverted segments b3 and b4. This suggests that there are at least four possible orders for sequences in this interval, and we find evidence for all four possibilities in the current study (fig 4). In light of these inversion variants, which place amplicon g3 within a gr amplicon (fig 4), we suggest that the original definition of gr/gr deletion¹⁶ should be extended to include any deletion that has arisen through unequal crossing over within g or r amplicons. The fact that even in only 32 men with a gr/gr deletion each haplogroup defined contains at least two inversion variants suggests that these inversions are relatively frequent events.

Genetic exchange within the non-recombining Y chromosome

A previous study concluded that SFV absence was the result of gene conversion between the arms of the large palindromes that make up most of the AZFc interval, but internal AZFc deletions in the fertile population were not considered as a possible alternative explanation, and no quantification was performed at the loci studied.³¹ Here, we establish that the majority of cases of SFV absence at CDY1-7750 are not associated with gr/gr deletion or copy loss. We therefore further investigated this question for two SFVs at the CDY1 locus, in our population of individuals known to carry two copies of CDY1. The SFVs used were CDY1-7750 and CDY1-84, from the 5′ UTR of CDY1. This latter polymorphism has been published independently.³¹

A comprehensive phylogeny of human Y chromosomes, based on the genotyping of MSY binary polymorphisms, is now available.²⁵ The genotyping of our study population allowed us to assign individuals to specific branches of the phylogeny and determine the relative position of SFV loss or acquisition.

The results for 465 undeleted men are summarised in fig 5. Our analysis shows that the CDY1 SFVs used are present on all four Y chromosomes defined, demonstrating that the presence of these two CDY1 SFVs is the ancestral state of all the derived Y chromosomes. Furthermore, SFV loss is not limited to the non-derived Y chromosomes, but is found in all four groups. We conclude that the majority of SFV loss at the CDY1 loci tested does not result from deletion, but that, as previously proposed,³¹ it is frequently a consequence of intra-palindrome gene conversion.

It is clear from our data that for an SFV to be a useful indicator of partial deletion, the incidence of its absence from intact Y chromosomes must be estimated for the Y haplotypes that compose the study group. This has not been done in other studies.^14,15

DAZ and CDY1 copy number on the gr/gr deleted Hg-N3 Y chromosome

It has been proposed that b2/b4 duplication may be a way in which full spermatogenic activity can be restored to an individual carrying a gr/gr deleted Y chromosome.¹⁶ If gr/gr deletions have an effect on spermatogenesis by reducing gene dosage, then a strong selection for these re-duplicated gr/gr deleted chromosomes would be expected, and consequently gr/gr deleted Y chromosomes present at a high percentage in a population would be the duplicated, rescued form. Two examples of gr/gr deleted Y chromosomes that have attained a high prevalence in certain populations have so far been described: the Hg-D2b Y chromosome present in 20–30% of Japanese men,^16,32 and the Hg-N3 Y chromosome present in 50% of Finnish men.^21,27,28 The gr/gr deletion on the Hg-D2b Y chromosome is associated with a sperm count of >40×10⁶ spz/ml in 66% of fertile carriers.³³ It has been shown by FISH that the D2b gr/gr deleted Y chromosome does not carry a b2/b4 duplication.¹⁶ This information is, however, not yet known for the Hg-N3 Y chromosome.

Using the Tat bi-allelic marker, we identified three control men and one infertile man with an Hg-N3 Y chromosome, all of whom carry the same gr/gr deletion, type 4b, deleted for CDY1b, DAZ3/4, and u3. Our results confirm that the Y chromosome on which the Hg-N3 deletion arose carried a b2/b3 inversion around DAZ1/2,²⁸ but show in addition that it carried a second inversion of the P1 palindrome around DAZ3/4. Dose tests revealed that only one fertile Hg-N3 man carries a b2/b4 duplication, and thus has four copies of DAZ and two copies of CDY1, while three others have only two copies of DAZ and one copy of CDY1. There would therefore appear to have been no strong selection for the b2/b4 duplication among men with either Hg-D2b or Hg-N3 Y chromosomes. Indeed, we and others¹⁶ have found that b2/b4 duplications are generally rare among the gr/gr deleted chromosomes, indicating that the majority of gr/gr deletions have little effect on male fertility.

DAZ and CDY1 SFVs define four deletion types

Partial AZFc deletions of both DAZ and CDY1 were not found with significantly different frequencies in infertile and control populations: 18 infertile men (6%) and 14 control men (3.5%) (p = 0.085) had such deletions. We therefore investigated the possibility that certain types of gr/gr deletion are associated with male infertility. The incidence of SFV loss without deletion at DAZ (sY587) or CDY1-7750 loci was low (table 2C), indicating that these SFVs can be used to define deleted loci with a low error rate. On the basis of the DAZ and CDY1 SFVs retained, we were able to define four gr/gr deletion types (fig 2 and table 2B).

Similar frequencies of gr/gr deletions of CDY1b with either DAZ1/2 (0.7% v 0.5%) or DAZ3/4 (1.3% v 1.5%) were observed for infertile versus control populations. Deletion of CDY1a with DAZ1/2 is slightly more frequent in infertile populations (2.3% v 1.0%; p = 0.14). However, a statistically significant, but weak, association with infertility was observed only with the type 4 gr/gr deletion characterised by DAZ3/4 and CDY1a SFV loss: 6/254 infertile men with <5×10⁶ spz/ml and 2/399 control men (p = 0.042) had this deletion. No fertile men carried this deletion (0/189). We did, however, find this deletion in a man with agenesis of the vas deferens who carries two mutations of the CFTR gene (13-4255; fig 2), further weakening this association between type 4 gr/gr deletion and infertility. Collectively, the incidence of gr/gr deletion of CDY1a shows a weak association with infertility: 12/300 infertile men v 6/399 control men (p = 0.035) had this deletion.

Loss of the CDY1-7750 SFV is associated with infertility

Comparing our infertile and control populations, we found similar frequencies of SFV absence for DAZ1/2, DAZ3/4, or CDY1b (distal), with and without deletion (table 2A). There was, however, a significantly higher frequency in the infertile compared to the control populations for the absence of the CDY1a SFV (proximal) at CDY1-7750 (27/300 v 14/399; p = 0.002). This result is significant even when the number of tests is taken into account (significance: p<0.0045; see Methods). This indicates that the absence of this SFV is associated with infertility. This increase is found for all four haplogroups defined (table 3A) and is at the limit of statistical significance in the largest, Hg-P (10/157 v 5/214; p = 0.047).

Table 2C shows the incidence of SFV absence without deletion. In men in whom there is no evidence of a partial AZFc deletion, absence of a CDY1-7750 variant (a or b) is significantly more frequent in the infertile than the control population, even when multiple testing is taken into account (19/282 v 8/383; p = 0.0026). Absence of each variant at CDY1-7750 is also significantly higher in infertile men compared to control men: 15/282 v 8/383 for CDY1a (p = 0.02) and 4/282 v 0/383 for CDY1b (p = 0.03). The higher incidence of CDY1-7750 absence is seen in all three derived haplogroups (table 3B), suggesting that it is not an artefact of population stratification. This indicates that loss of a segment of the AZFc interval defined by CDY1-7750 has a deleterious effect on male fertility. The mechanism for this loss is not clear and will require further study. Since SFV loss is not associated with copy loss in these cases, a deletion of a gr/gr duplicated chromosome, such as we describe in patient 13-1877, is one possibility. In this case, the absence of a locus flanking CDY1-7750 could underlie the infertility. If, on the other hand, gene conversion were responsible, the SFV at CDY1-7750 would be required for full male fertility.

DISCUSSION

Most gr/gr deletions have little effect on male fertility

Here, we show that the majority of partial AZFc deletions that affect copies of CDY1 or DAZ are probably mediated by gr/gr amplicon recombination. Based on the DAZ and CDY1 genes deleted, we sub-divided the gr/gr deletions found into four types. Deletions of CDY1a show a weak association with infertility, suggesting that some may cause infertility. On the other hand, deletions of CDY1b, representing 48% of the total number of gr/gr deletions, are present in 2% of both infertile and control populations, suggesting that they represent variants of the Y chromosome with little effect on fertility. Our results do not exclude the possibility that certain gr/gr deletions, occurring on particular Y chromosomes, could be the cause of male infertility, but they do strongly suggest that most gr/gr deletions do not have a significant effect on male fertility. Our study therefore challenges the conclusion of a recent study,¹⁶ that the majority of gr/gr deletions affect male fertility through a reduction in sperm production. It remains possible that all gr/gr deletions reduce sperm numbers. This question could best be addressed by comparing sperm counts in groups of fertile men carrying the same gr/gr deletion with and without b2/b4 duplications.

gr/gr recombination is not evenly distributed in the gr amplicon

It was originally assumed that the gr/gr unequal cross over frequency must be comparable to that of the b2/b4 event that generates the complete AZFc deletion because, at ∼540 kb, the gr amplicon is more than twice the length of the b2 or b4 amplicon.¹⁶ On this basis, it was estimated that gr/gr deletions must have a negative effect on infertility, otherwise they would be present on >40% of contemporary Y chromosomes. The fact that in 487 men we did not observe any DAZ gene copy loss without sY587 SFV loss shows that most gr/gr deletions do not occur through unequal cross over in the 135 kb segment situated between the two sY587 sites of each DAZ gene pair. In contrast, the type 1b, 2c, 3d, and 4a deletions (fig 2), representing 15 of the 32 gr/gr deletions (eight of 17 independent events), must have their proximal breakpoint in the 70 kb situated between sY587 in the 3′ end of DAZ2 and the distal end of the P2 palindrome, identifying this region as a relative hotspot of unequal crossing over. Our results show that the gr/gr deletion frequency is independent of the length of homology. Thus, the incidence of gr/gr deletions in the general population provides no support for their negative effect on fertility.

gr/gr deletions and normospermia

In the study that found a significant association between gr/gr deletions and spermatogenic failure (9/246 infertile men and 0/148 normospermic men; p<0.014, Fisher’s exact test, one-sided), the men in the normospermic population were characterised by a total sperm count of >40×10⁶. In our study, we identified two infertile men with >40×10⁶ spermatozoa in their ejaculate who carry distinct gr/gr deletions (13-1485 and 13-3446; fig 2). These two men were part of a group of 24 men, from infertile couples, who had sperm counts of >20×10⁶ per millilitre. This high incidence of gr/gr deletion (8.3%) in a group characterised by a high sperm count indicates that most gr/gr deletions have little or no effect on sperm production.

In the published association study, gr/gr deletions were initially identified by a negative result for the P2/P1 junction marker, sY1291.¹⁶ This strategy was chosen, prior to the discovery of the b2/b3 inversion polymorphism,²⁸ on the basis of an earlier theoretical description of gr/gr deletions.¹⁷ When a gr/gr deletion occurs on a b2/b3-inverted chromosome, however, u3 is deleted instead of the P2/P1 junction. Although a u3 marker was used in that study,¹⁶ all 27 gr/gr deletions identified were negative for the P2/P1 marker, and none for the u3 marker, and it is unclear whether u3 was screened for in all men or only in men negative for the P2/P1 junction. This same study did not identify the u3-gr/gr deletion in two Hg-N3 individuals, demonstrating that the authors’ screening strategy did not detect all gr/gr deletions. In contrast, we found nine u3-deleted chromosomes out of 32 gr/gr deletions. Thus, the study showing a link between gr/gr deletion and reduced sperm production may not include 25% of the total number of gr/gr deletions present in the groups of men studied. The proposed link between gr/gr deletion and reduced sperm count should, therefore, be disregarded until the u3-gr/gr deletions, which remove more genes than the P1/P2-gr/gr deletions,²⁸ are taken into account.

AZFc genes and human fertility

It has been proposed that partial AZFc deletions may affect male fertility by reducing the dose of certain AZFc genes.¹⁶ In consequence, any gr/gr deletion should reduce fertility and should be subject to negative reproductive selection. The results of our study reveal weaknesses in the main lines of evidence advanced in support of this model, suggesting that gene copy loss through gr/gr deletion often has little effect on human fertility.

We suggest, therefore, that an alternative model in which most gr/gr deletions are neutral, and the rate of gr/gr deletion is much lower than that of complete AZFc deletion, should also be considered. The effects of gr/gr deletions would be determined by the presence of sequence dependent loss-of-function differences between AZFc genes, and would only be seen when the functional copy is deleted and the mutated copy retained. No effect of gr/gr deletion would be seen on Y chromosomes with identical gene copies, or if the mutated copy was deleted. The effect of gr/gr deletion would, therefore, be expected to vary on Y chromosomes of different haplogroups.

The relatively strong association observed in our study between infertility and SFV loss at CDY1-7750 without a reduction in copy number, represents the first evidence that, in conjunction with partial AZFc deletion or gene conversion, sequence dependent loss-of-function differences between AZFc gene copies may frequently be the basis for spermatogenic failure. The further study of these variants could lead to the identification of the AZFc genes that are necessary for efficient spermatogenesis in man.