Diversified bursal medullary B cells survive and expand independently after depletion following neonatal infectious bursal disease virus infection

Abstract

The primary immunoglobulin repertoire of chickens is generated not by gene rearrangement but by a subsequent process of gene conversion in proliferating immature B cells within the follicles of a specialized gut-associated lymphoid organ, the bursa of Fabricius. Neonatal infection with infectious bursal disease virus can eliminate almost the entire bursal B-cell compartment. Thereafter, two types of follicle reappear. Larger follicles, with rapidly proliferating B cells and normal structure, are correlated with partial recovery of antibody response. Smaller follicles, lacking distinct cortex and medulla, appear unable to produce antigen-responsive B cells. To understand the genesis of the two types of follicle, we analysed their V_L sequences and activation-induced deaminase mRNA levels. The results provide a model of bursal repopulation in which surviving bursal stem cells generate new follicles with normal morphology and function, while surviving medullary B cells continue to proliferate slowly, under the influence of stromal cells, giving rise to the smaller follicles. The latter remain fixed in a stage of development incapable of further gene diversification.

Introduction

The chicken has only a single rearranging variable region gene segment at either of the immunoglobulin (Ig) light and heavy chain loci.^1,2 Gene rearrangement therefore produces only a very limited Ig gene repertoire resulting from junctional diversity. A subsequent process of gene conversion, whereby sequences derived from an upstream array of divergent V(D)-region pseudogenes replace segments of the single rearranged V(D)J gene, is responsible for generating the primary antibody repertoire.² This process occurs in a proliferating immature B-cell population within the follicles lining the epithelium of a specialized gut-associated lymphoid organ, the bursa of Fabricius.^2,3 Like somatic hypermutation and class switch recombination,⁴ this process is dependent on activation-induced deaminase (AID).^5,6 After post-translational modification, mammalian AID is transported to the nucleus where, in combination with other factors, it causes the targeted deamination of DNA.⁷ This initial lesion in the Ig V-region may be resolved by somatic hypermutation or gene conversion, depending on the balance of DNA repair enzymes present in the affected cells.⁸

Bursal follicles are colonized at between 8 and 14 days of embryonic development by prebursal stem cells bearing surface Ig.⁹ Before hatch, the follicles contain a morphologically homogeneous population of rapidly proliferating B cells. After hatch, the B cells in the developing bursa become partitioned into distinct cortical and medullary populations, separated by a basement membrane. This is thought to result from the outward migration of B cells through the basement membrane to the newly formed cortex.¹⁰ The cortical B-cell population continues the rapid proliferation and gene conversion initiated in the embryonic bursa between days 16 and 18 of embryonic development.^2,11 The initiation of gene conversion is accompanied by a developmental change that includes the expression of the carbohydrate determinant LewisX (Le^X)¹¹ and a marked increase in levels of AID mRNA.¹² Many cells are lost by apoptosis in the rapidly proliferating cell populations, presumably as a consequence of selective pressures for expression of a functional B-cell receptor.^10,13 After hatch, mature B cells start to emigrate to the periphery and, despite the short half-lives of many,¹⁴ provide the antigen-responsive B cells required for humoral responses. The emigrant cells are predominantly derived directly from the cortical population.¹⁵ While it is clear that the function of the cortex is the generation of the primary B-cell repertoire, the function of the medulla is less clear. The medulla communicates with the lumen of the bursa, which is continuous with the gut lumen, through a specialized follicle-associated epithelium, which can transfer particulate matter from the lumen into the medulla¹⁶ and is therefore likely to have an antigen-presenting function. The extent to which this may be involved in the generation of tolerance or in the generation of response is not clear, although it is possible to obtain antibody responses to antigens taken up into the bursa from the lumen.¹⁷

Little is known about the signals that control bursal B-cell colonization, survival, proliferation, gene conversion, movement between compartments or emigration. Elegant investigations have shown the requirement for basal levels of B-cell receptor signalling for survival of bursal B cells up to the point of hatch, although this appears to be insufficient thereafter.¹⁰ The B-cell survival-promoting cytokine BAFF [B-cell activating factor belonging to the tumour necrosis factor (TNF) family (TNFSF13b)] is expressed in the bursa and can delay the rapid apoptosis of isolated bursal B cells removed from the follicular environment.^18,19 It has been suggested that bursal B-cell survival may be dependent on an autocrine activity of BAFF expressed by the bursal B cells themselves.¹⁸

Neonatal infection with a high dose of infectious bursal disease virus (IBDV) can cause almost complete destruction of the bursal B-cell population²⁰ comparable to that achieved through cyclophosphamide treatment.²¹ Following infection, there is a partial recovery of the numbers of bursal B cells. Using this experimental system to investigate the developmental potential of surviving B-cells, we have observed that there are two distinct types of follicle in the recovering bursa, larger follicles with a cortex and medulla, and smaller follicles without these structural compartments.²⁰ Birds with only small follicles did not produce detectable antibodies against IBDV or subsequently administered antigen. Those with both types of follicle produced both types of antibody, although at much lower levels than uninfected birds. As ability to mount Ig responses was correlated with the presence of the larger follicles, we hypothesized that these were essentially normal and productive, having been reconstituted from surviving bursal stem cells. In contrast, the small follicles were not able to support the complete programme of bursal B-cell development. They might represent either B cells arrested at an early stage by lack of necessary signals, or medullary B cells that have already completed the cortical stage of the bursal development programme and no longer have the capacity to generate an Ig repertoire by gene conversion. To test these hypotheses, we examined the diversity of Ig light-chain V-regions, and measured AID mRNA in these two distinct types of follicle. We also examined the two types of follicle with reagents that discriminate cortical and medullary compartments. The results support the hypothesis that the small follicles are the descendants of medullary B cells that maintain their phenotype in the absence of a discernible cortex.

Materials and methods

Animals and IBDV infection

Specific pathogen-free chickens were obtained from the Poultry Production Unit at the Institute for Animal Health, and maintained with feeding and water ad libitum, and filtered ventilation. An infectious dose of the virulent F52/70 strain of IBDV²² was administered, which had been determined by titration²⁰ to cause severe depletion of bursal lymphocytes without evident morbidity or mortality. Two-day-old RPRL Line 6 White Leghorn chicks were infected with 25 µl of purified virus in phosphate-buffered saline (PBS) per nares and 25 µl of virus per eye. All procedures were conducted in accordance with the UK Animals Scientific Procedures Act 1986 and local ethical review procedures.

Immunohistochemistry

The monoclonal antibody (mAb) AV20 recognizes a monomorphic determinant on the Bu-1 molecule, expressed by B cells.^23,24 The mAb BB3, which binds the cortico-medullary boundary in the normal bursa, was obtained from a mouse immunized with cells obtained from bursas reconstituted, following embryonic cyclophosphamide treatment, with cells infected with the HB1 retrovirus (JRY, unpublished data). The anti-Le^X Ab was obtained from Serotec (Oxford, UK). The mAb M1 binds chicken IgM.²⁵ Sections of tissue frozen in Tissue-Tek O.C.T. (Bayer, Newbury, UK) were fixed in acetone, treated with 0·03% H₂O₂ in PBS to inactivate endogenous peroxidase and washed three times in PBS before incubation with primary Abs. Bound Ab was detected using the Vectastain ELITE ABC kit for mouse IgG and the NovaRED substrate (Vector Laboratories, Peterborough, UK), as described in detail elsewhere.²⁰ Sections were counterstained with haematoxylin and permanently mounted for microscopy.

Ig light-chain V-regions from laser microdissected follicles

Frozen sections of recovering bursas were cut, and Bu-1⁺ cells stained using the mAb AV20²³ and the Vectastain ELITE ABC kit for mouse IgG (Vector Laboratories), according to the manufacturer's instructions. Sections were counterstained and dehydrated for laser microdissection using the AS LMD system (Leica Microsystems, Milton Keynes, UK). To minimize potential bias resulting from the effects of spatial distribution of clonally related cells, similar sized areas were dissected from large and small follicles. Microdissected fragments were incubated with proteinase K (20 µg/ml) in 100 mm NaCl, 25 mm Tris HCl, 10 mm ethylenediaminetetraacetic acid (EDTA) and 0·5%[volume/volume (v/v)] sodium dodecyl sulphate (SDS), pH 8·0, for 3 hr at 55°. DNA was prepared from the digested samples using the Qiagen PCR purification system (Qiagen, Crawley, UK), eluting with 30 µl of 10 mm Tris HCl, pH 8·5. Light-chain V-J genes were amplified with nested PCR using primers described elsewhere.⁵ PCR products were purified by agarose gel electrophoresis, cloned into pGEM-T (Promega, Southampton, UK) and sequenced using vector primers in the CEQ 8000 Genetic Analysis System (Beckman Coulter, High Wycombe, UK).

Detection of mRNA

The digoxigenin (DIG)-labelled antisense and sense probes for detection of AID and Bu-1 mRNA by in situ hybridization have been described previously.¹² For detection of BAFF mRNA, DIG-labelled antisense and sense probes were prepared as described for the AID probes.¹² Processing of sections, hybridization and detection of DIG were as described previously.¹² Methods for preparation of total RNA from pooled laser-dissected samples and quantitative RT-PCR measurement of AID mRNA and 28S ribosomal RNA, as well as the probes and primers used, were exactly as described previously.¹² Serial 10-fold dilutions of a reference RNA, bursa RNA for AID mRNA and liver RNA for 28S rRNA were assayed in the same run. The slopes (S for AID and S′ for ribosomal RNA) of linear regressions of C_T (threshold cycle) against the natural logarithm of the respective reference RNA concentration were determined using all dilution series from assay runs for which samples were to be compared. AID mRNA C_T values were adjusted (N_T) for differences in 28S rRNA using the equation N_T = C_T + (S/S′)(X − C_T′), where C_T is the C_T value for AID mRNA, C_T′ is the C_T value for 28S RNA, and X is an arbitrary constant. Thus the values compared are the amounts of AID mRNA relative to the amount of 28S rRNA in each sample. The mean C_T of triplicate analyses of each RNA sample was treated as a single measurement for statistical calculation. Log₂ ratios (R) of AID mRNA levels were calculated using the relationship log₂R = –(N_T1 − N_T2)/Slog_e. Standard error ranges were calculated assuming that C_T was normally distributed and subsequently transformed to a linear scale.

Results

Two distinct types of follicle in recovering birds

The bursas of some birds recovering from IBDV infection at 2 days of age contained two distinct types of follicle, distinguished by size and by the presence or absence of apparent cortico-medullary structure.²⁰ While both types of follicle were smaller than those in uninfected birds of similar age, both were composed mostly of cells expressing the B-cell marker Bu-1 (Figs 1a and b). Their smaller size, and the lack of cortico-medullary boundary, suggested that the small follicles might be precursors of the larger in a developmental process recapitulating that seen in the development of post-hatch follicles from the unstructured follicles in the embryonic bursa. However, the consistent presence of small follicles up to 50 days post-infection indicated that these were two distinct types of follicle. Staining of sections with an anti-Le^X Ab recognizing the carbohydrate antigen Le^X, expressed in embryonic bursal follicles from about embryonic development day 16,¹¹ revealed that the large follicles contained Le^X+ cells, like normal bursal follicles (Fig. 1c), while the small follicles did not (Fig. 1d).

Figure 1.

Two distinct types of follicle in recovering bursas. The left-hand panels (a, c) show serial sections of an uninfected 36-day-old bursa. The right-hand panels (b, d) show serial sections from the bursa of a bird, infected with infectious bursal disease virus (IBDV) at 2 days after hatch, 34 days post-infection (dpi). The upper panels (a, b) are stained with monoclonal antibody (mAb) AV20, to reveal Bu-1, and the lower panels (c, d) with anti-LewisX (Le^X) mAb. Arrowheads point to two distinct types of follicle: open arrowheads to two small Le^X-negative follicles, and the closed arrowhead to a large Le^X-positive follicle with an evident cortico-medullary boundary.

The antibodies M1 and BB3 can be used to distinguish the cortex and medulla of mature bursal follicles. M1, an anti-IgM antibody, stains only medullary B cells (Fig. 2a). M1 stained only cells within the medulla of the large follicles in the recovering bursa (Fig. 2b; l), but stained cells throughout the small follicles (Fig. 2b; s). BB3 stains an unidentified structure marking the cortico-medullary boundary in normal follicles (Fig. 2d). In recovering follicles, BB3 stained an internal structure delineating the cortico-medullary boundary in the large follicles (Fig. 2e; l). The BB3 staining of small follicles (Fig. 2e; s) revealed similar staining, but close to the edge of the follicles as revealed by the Bu-1 staining (Fig. 2f), with at most a single layer of external Bu-1⁺ cells. Thus, by both M1 staining and by the spatial relationship to the structure revealed by BB3, all the cells in the small follicles appeared like M1⁺ medullary B cells. Staining with M1 also indicated that these cells were indeed B cells and not an expanded population of other Bu-1⁺ cells.

Figure 2.

Structural differences in follicles in recovering bursas. Sections of an uninfected bursa (a, d) and a bursa 34 days post-infection (dpi) at 2 days of age (b, c, e, f) were stained with M1, anti-immunoglobulin M (IgM), BB3 or AV20, anti-Bu-1 antibodies, as indicated. The pairs (b, c) and (e, f), within dashed rectangles, are serial sections. Italicized letters on the micrographs indicate cortex (c) and medulla (m), or large (l) and small (s) follicles. Arrowheads at the left of panel (d) indicate the inter-follicle boundary (open) or the cortico-medullary boundary stained by BB3 (closed).

In situ hybridization using a BAFF antisense probe with sections of bursa revealed the presence of a high level of BAFF mRNA in a minor cell population scattered throughout the medulla of follicles (Figs 3a and b). In the recovering bursa, a similar population of BAFF⁺ cells was detected within small follicles (Fig. 3c) and in the central regions of large follicles (Fig. 3d), demonstrating a further similarity between the small follicles and the normal medullary structure.

Figure 3.

Detection of BAFF [B-cell activating factor belonging to the tumour necrosis factor (TNF) family (TNFSF13b)] mRNA by in situ hybridization. (a, b) Serial sections from a normal bursa hybridized with antisense BAFF probe (a) or control sense BAFF probe (b). (c, d) Hybridization of BAFF antisense probe to sections containing two small follicles (c) or a large follicle (d). The boundaries of follicles are indicated with dashed white outlines.

Ig light-chain V-region sequences in follicles from convalescent bursas

Serial sections were prepared from normal bursas and from the bursas of birds 51 days after infection. Staining with the Le^X Ab was used to confirm the distinction between large Le^X+ follicles, with distinct cortico-medullary structure, and small Le^X–, unstructured follicles. Three samples of similar area were collected, by laser microdissection, from successive sections from one normal follicle, and from three large and 10 small convalescent follicles. Rearranged Ig light-chain V-region genes were amplified by nested PCR from the pooled samples from each individual follicle, and sequenced. The sequences were aligned and analysed for the presence of non-germline sequence that could be attributed to pseudogene donors (supplementary data, Supplementary Figures S1–S3). Most identified pseudogene sequences were found in the pseudogenes of a different line of chickens.² The small number of other sequences, found in multiple independent PCR reactions, were treated as being derived from different pseudogenes specific to the Line 6 birds. The results of this analysis for the different types of follicle are summarized in Table 1. While no two identical sequences were recovered from among 36 sequences from a normal bursal follicle, duplicates were found in sequences from both large and small follicles from convalescent bursas. The extent of diversity was substantially lower in the sequences from the small follicles, although the average number of conversion tracts per gene was similar in all three classes of follicle. A small proportion of sequences were out-of-frame in both types of follicle from infected birds, but in none from the uninfected bird.

Table 1.

Properties of V_L sequences obtained from laser-captured follicles

			Sequences
IBDV	Type	Follicles	Total	Different	Percentage diversity1	GC rate2	Percentage in frame
–	Normal	1	36	36	100	5.1	100
+	Large LeX+	3	99	69	70	6.1	94
+	Small LeX–	10	220	44	20	5.6	86

¹Percentage of sequences that are unique.

²Mean number of identified gene conversion tracts in each unique sequence. IBDV, infectious bursal disease virus; Le^X, LewisX.

Data for the individual follicles from infected birds are shown in Table 2. While the three large follicles showed similar levels of diversity and numbers of conversion tracts, the small follicles were much more variable, with from little or no diversity to about half that of the large follicles. The greater variability of gene conversion density between individual small follicles is evident from the histograms shown in Fig. 4. However, the mean numbers of conversions are not significantly different [Tukey Honestly Significantly Different (HSD) at 95% confidence]. Thus, while the smaller proportion of distinct sequences from the small follicles suggested that the rate of ongoing gene conversion in the small follicles was lower than that in the large follicles, the total historically accumulated burden of conversions in individual genes was similar in the two types of follicle. This allowed the hypothesis that the small follicles were composed of cells already diversified at the time of infection and no longer active in gene conversion, while the large follicles contained actively diversifying Ig genes.

Table 2.

V_L sequences from individual follicles from infected bursas

	Sequences
Follicles type1	Total	Different	Percentage diversity2	GC rate3	Out of frame
L	24	17	71	5.9	0
L	38	26	68	6.4	1/26
L	37	26	70	6.0	3/26
S	15	5	33	5.8	1/5
S	17	4	24	6.5	0
S	24	5	21	6.6	0
S	39	1	3	6.0	1/1
S	31	2	6	6.0	0
S	17	6	35	6.2	2/6
S	20	4	20	6.8	0
S	23	5	22	6.4	1/5
S	16	5	31	2.0	1/5
S	21	7	33	5.0	0

¹L, large LewisX (Le^X)⁺ S, small Le^X−.

²Percentage of sequences that are unique.

³Mean number of identified gene conversion tracts in each unique sequence.

Figure 4.

Distribution of numbers of identifiable conversion events in distinct immunoglobulin light chain variable genes of different classes of bursal follicles. The class of follicle is indicated at the top left of each panel. At the top right, numbers indicate the number of follicles examined, followed by the total numbers of distinct sequences. The minimum numbers of conversion events accounting for each sequenced gene were estimated using a computer algorithm as described in notes to Supplementary Fig. 4.

AID mRNA levels in different follicles

Serial sections from convalescent bursas were hybridized with antisense and control sense probes for the detection of Bu-1 mRNA (and hence B cells) and AID mRNA. The results are shown in Fig. 5. While the AID antisense probe clearly detected AID mRNA in the large follicles (Fig. 5b), the level of AID mRNA in the small follicles was lower than the limit of detection of this assay (Figs 5b, d and f).

Figure 5.

Activation-induced deaminase (AID) mRNA levels were lower in small follicles. (a, b) Serial sections, from the bursa 34 days after infection with infectious bursal disease virus (IBDV) at 2 days old, hybridized with antisense probes for Bu-1 mRNA (a) or AID mRNA (b). (c–f) Serial sections of a bursa, from an identically treated bird, which contained only small follicles, hybridized with antisense Bu-1 probe (c), antisense AID probe (d), control sense Bu-1 probe (e), or control sense AID probe (f). Solid arrowheads point to a large follicle. Open arrowheads point to the same small follicle in each set of serial sections.

Levels of AID mRNA in laser-dissected follicles were also assessed by quantitative RT-PCR (Fig. 6). The measured level of AID mRNA detected in the small follicle samples, relative to 28S rRNA, was about 16-fold lower than that in either the large follicles from infected birds or the follicles from uninfected birds.

Figure 6.

Quantitative reverse transcriptase–polymerase chain reaction (RT-PCR) measurement of activation-induced deaminase (AID) mRNA in laser microdissected follicles from bursas of infectious bursal disease virus (IBDV)-infected birds. AID mRNA and 28S rRNA were analysed in RNA from n individual laser-dissected follicles from an uninfected 34-day-old bird (U, n =2), and from large (L, n =15) and small (S, n =13) follicles from the bursas of birds 34 days after infection with IBDV at 2 days of age. AID mRNA levels are expressed relative to 28S rRNA, and normalized to the level in the follicles from uninfected birds. The left-hand panel shows the results on a log₂ scale and the right hand panel shows the same data transformed onto a linear scale. Error bars show the standard error calculated from the log₂ data. Welch's two sample t-test (one-sided) confirmed that the level of AID mRNA, relative to 28S rRNA, was lower in small than in large follicles (P < 0·000002).

These observations provide further evidence that the small follicles, unlike the large follicles, were no longer active in gene conversion, and consequently that the traces of conversions evident in their Ig genes were probably from events preceding the IBDV infection.

Discussion

The bursas of some birds neonatally infected with IBDV contained follicles with normal morphology whose presence is necessary for subsequent development of Ig-secreting and antigen-responsive B cells.²⁰ In cyclophosphamide-treated birds, the development of normal follicles is dependent on the addition of bursal stem cells.²¹ Thus it is likely that the normal follicles in birds recovering from IBDV infection were dependent on the survival of small numbers of bursal stem cells, which are present in the neonatal bursa.²⁶ It is not clear whether the development of normal follicles from these bursal stem cells is dependent on follicular stromal elements left after decimation of the B-cell population, or whether they form by a complete recapitulation of the normal embryonic development pathway.

Small follicles, which may be similar to the small follicles present in all birds following neonatal IBDV infection, were also observed in cyclophosphamide-treated birds.²¹ While the large follicles expressed the carbohydrate antigen Le^X, a marker for the initiation of gene conversion in the embryonic bursa,¹¹ the small follicles did not. This suggested that there might be differences in the gene conversion process in the two types of follicle. Therefore we investigated the nature of the cells in the small follicles by comparing them with large follicle cells with respect to gene diversification status, as well as investigating their relationship to stromal components.

The stochastic nature of gene conversion events, the likely occurrence of neutral and reverting conversions, and the obscuring of traces by overlapping conversions preclude accurate measurement of the number of conversion events that have affected any one Ig V-region. Nevertheless, counting of recognizable conversion tracts in sequences from large follicles could clearly distinguish the complexity of the conversion history of the component B cells from the less extensive conversion in sequences from day 16 of embryonic bursal development, when conversion has just started.¹² On the basis of light-chain V-region sequences from one set of sequences we obtained from a single small follicle, which were entirely germline, we considered the possibility that the small follicles might be derived from bursal stem cells developing without the stromal-derived signals required to initiate gene conversion. The more extensive analyses described here showed that the Ig light-chain genes from the small follicles, with the exception of a minority of sequences from one follicle, bore the traces of similarly extensive gene conversion to that found in the large productive follicles. However, the low level of AID mRNA in the small follicles, and the much lower overall diversity of the Ig sequences recovered from them, indicated that gene conversion was active at a much reduced level, if at all, in the B cells within the small Le^X– follicles. We conclude that the B cells in the small follicles were descendants of small numbers of surviving B cells that already possessed diversified Ig genes at the time of infection.

Embryonic bursal follicles lack distinct cortical and medullary compartments and contain populations of B cells similar to that of the juvenile cortex, with rapid proliferation and active gene conversion. The less rapidly proliferating medullary B-cell population develops after hatch, as cells migrate through the basement membrane where they establish the rapidly proliferating cortical population.¹⁵ The proportions of the medullary population composed of descendants of cells that never entered the cortex, or cells that have moved back across the basement membrane from the cortex, is unknown. As bursal emigrant B cells appear to derive directly from the cortex,¹⁵ the medullary B-cell compartment is likely to have different functions, including selection by antigen taken up from the lumen by follicle-associated epithelial cells and the generation of humoral responses to antigen accumulated by that route. Like medullary B cells, but in contrast to cortical B cells, the cells in the small follicles following IBDV infection all stained with the M1 antibody, were located within the basement membrane structure revealed by the BB3 antibody, and were associated with a population of cells with high levels of BAFF mRNA. Thus we propose that the small follicles arise from a surviving population of medullary B cells, which are more resistant than cortical cells to destruction by the effects of IBDV infection, but are unable to step backwards in their developmental programme to establish a cortical population active in gene conversion. The slower rate of proliferation of medullary B cells would account for the smaller size of the resulting follicles.

We previously failed to find evidence of lower levels of AID mRNA in the medulla than in the cortex of mature bursal follicles.¹² This contrasts with clear evidence of lower levels of AID mRNA in the small follicles after IBDV infection. However, we were unable to detect, by microscopic examination, any small follicles connecting to the bursal lumen through the follicle-associated epithelium typical of normal bursal follicles. AID expression in normal medulla may be antigen driven, resulting in either receptor editing or somatic hypermutation,¹² and thus be absent from the small follicles because they are not exposed to luminal antigen.

Koskela et al.¹⁸ reported the expression of BAFF in magnetically sorted bursal cells expressing the marker Bu-1, as well as demonstrating expression in B-cell lines. They proposed an autocrine loop responsible for maintaining the bursal B cells. Our in situ analysis of BAFF expression was insufficiently sensitive to detect BAFF mRNA in the major B-cell populations of either cortex or medulla, but did reveal a minor scattered population of cells, restricted to the medulla, with detectable levels of BAFF mRNA. These results do not necessarily conflict. A relatively low level of BAFF expression by the overwhelming number of B cells in the cortex could be sufficient for the autocrine maintenance of cortical B cells, while the medullary B cells might depend on a relatively high level of expression from a small population of stromal cells. However, it is possible that the BAFF-expressing stromal cells also express the Bu-1 marker. Formal evidence that bursal B cells depend on BAFF for survival is lacking, but the invariable association of the small-follicle B cells with the cells containing high levels of BAFF mRNA does suggest that BAFF from the latter may play a part in allowing the former to survive and expand, and even that the anti-apoptotic effects of BAFF may account for the relative resistance of those cells founding the small follicles to the apoptotic effects of IBDV infection.²⁷ Thus it is more likely that the small follicles form from the remnants of depleted follicles, where a few medullary cells survive in association with BAFF-expressing stromal cells within the collapsed basement membrane, than that they are generated de novo, as we propose for the large follicles.

The observations we have reported allow the conclusion that the small follicles found in all birds after recovery of bursas depleted of B cells by IBDV infection are founded by a few surviving B cells that have escaped destruction, and that these cells do not recapitulate the normal developmental programme of rapid expansion and gene conversion in follicles of normal morphology. This is in spite of the fact that the bursal stroma does still have the capacity to support normal follicle development if bursal stem cells are available. Thus these cells represent a post-cortical stage of B-cell development, which supports the proposition that medullary B cells are derived from embryonic or cortical B cells that have irreversibly completed the gene diversification programme.

Abbreviations

AID

activation-induced deaminase

BAFF

B-cell activating factor belonging to the tumour necrosis factor family (TNFSF13b)

DIG

digoxygenin

IBDV

infectious bursal disease virus

mAb

monoclonal antibody

PCR

polymerase chain reaction