Because of its pivotal role in B cell survival and homeostasis and the greater sensitivity of MZ B cells to its presence, we examined whether BAFF overexpression in C57BL/6 BAFF transgenic mice might preferentially promote the survival of N^– B cells in the MZ. However, we found that the levels of N^– CDR3s in MZ B cells from these mice were the same as in wild-type mice (Fig. S2, available at http://www.jem.org/cgi/content/full/jem.20080559/DC1). There were no significant differences between C57BL/6 MZ and FO cells with respect to V_H and D_H gene usage, CDR3 amino acid usage, or hydrophobicity (Fig. S3 and S4), but MZ B cells had lower levels of somatic hypermutation than FO cells (P < 0.01; Fig. S5). Thus, the enrichment in the MZ with N^– cells is the most striking difference between the FO and MZ compartments.

Examining repertoire-based selection by a TdT^+/+/TdT^–/– BM chimera approach
To clarify whether N^– MZ cells represented a previously unidentified fetal B cell lineage or if CDR3 repertoire-based selection was responsible for the enrichment of these cells in the MZ, we used a novel approach wherein we competitively reconstituted irradiated C57BL/6 mice with TdT^+/+ and TdT^–/– donor BM cells from adult mice. Repertoire-based selection for N⁺ and N^– BCRs could then be clearly identified by skewing of the original reconstitution ratio. These studies also allowed us to examine the role of repertoire not only for mature cell fate decisions but also at each of the earlier developmental stages during B cell differentiation in these nontransgenic polyclonal mice.

CD45 (i.e., Ly5) allelic markers were used to distinguish TdT^+/+ from TdT^–/– donor-derived cells by flow cytometry. The surface markers and gating strategies used for the identification of BM and splenic B cell subsets are shown in Figs. S6–10 (available at http://www.jem.org/cgi/content/full/jem.20080559/DC1). With the exception of peritoneal B1 cells, which are known to be relatively radio resistant (25), autoreconstitution of lethally irradiated TdT^–/– hosts with endogenous B cells did not occur.

BM B cells favor the TdT^+/+ and Ig repertoires
During the early B cell developmental stages in the BM (Hardy pro-B fraction [Fr.] A, Fr. B+C, and pre-B Fr.D) the ratio of TdT^+/+/TdT^–/– cells remained constant (Fig. 2) and this baseline is marked in all figures. The chimeras were analyzed for all splenic and BM subsets and clear differences were apparent between many subsets.

View larger version (8K): [in this window] [in a new window] [Download PPT slide]   Figure 2. TdT+/+/TdT–/– reconstitution ratios remain constant in surface Ig– early BM B cell subsets. Lethally irradiated C57BL/6 TdT–/– mice received equal proportions of B220-depleted Ly5.1 TdT+/+ or Ly5.2 TdT–/– adult donor BM cells and were allowed to reconstitute for 5 wk before analysis. The TdT+/+/TdT–/– ratio in early BM B cell developmental stages is shown. Data were normalized with respect to the Fr. A compartment (dotted line; n = 21 mice). Error bars represent SEM.

View larger version (21K): [in this window] [in a new window] [Download PPT slide]   Figure 3. IgH and IgL selection in TdT+/+/TdT–/– chimeric mice. Mice were reconstituted with equal proportions of B220– donor BM TdT+/+ and TdT–/– cells. (A) The TdT+/+/TdT–/– ratio in BM and spleen B cell compartments of chimeric mice. Data were normalized with respect to the combined Fr. A–D compartment (dotted line; n = 21 mice). Data represent the mean ± SEM. Statistically significant differences are indicated (*, P < 0.05; **, P < 0.01; ***, P < 0.001). (B) IgL usage in BM and spleen B cell compartments of chimeric mice (n = 12–20 mice). Data represent the mean ± SEM. With the exception of the T1 compartment, TdT–/– cells had significantly higher frequencies of Ig+ cells than TdT+/+ cells. Statistically significant differences in light chain usage between TdT+/+ cells in each subset are indicated (*, P < 0.05; ***, P < 0.001), and between TdT–/– cells, statistical significance is shown (+, P < 0.05; ++, P < 0.01; +++, P < 0.001).

BM Fr. F B cells have been considered to be recirculating FO-derived B cells. Therefore, although the Fr. F compartment is similar to BM transitional cells with respect to IgH and IgL selective preferences, it is striking that the composition of this compartment is much more skewed toward TdT^+/+ (P < 0.001) and Ig⁺ (P < 0.001) cells when compared with the splenic FO compartment (Fig. 3, A and B). This suggests that any recirculation that does occur between these compartments is not sufficient to homogenize the repertoires in these two distinct locations.

Entry into the splenic transitional pathway is skewed toward the TdT^–/– repertoire, but subsequent transitional development favors TdT^+/+ and Ig⁺ cells
The transition from BM Fr. E (AA4⁺CD23^–) to the first splenic transitional stage (T1) was marked by pronounced enrichment in TdT^–/– cells (Fig. 3 A).This may be attributable in part to the preferential retention of TdT^+/+ cells in the BM CD23⁺ transitional compartment. In contrast to this BM-to-spleen T1 transition, progression through the splenic T2 and T3 stages is characterized by an increased proportion of TdT^+/+ cells (Fig. 3 A). As in the BM transitional compartment, this is mirrored by an increased frequency of Ig⁺ cells (Fig. 3 B). However, significant differences are apparent between the BM transitional and splenic T2/T3 compartments as the BM transitional cells are more highly enriched in TdT^+/+ (P < 0.001) and Ig⁺ cells (P < 0.001; Fig. 3, A and B). These data support the idea that these compartments may represent distinct pathways and also demonstrate that repertoire plays an important role in this divergence.

The MZ compartment preferentially selects for N^– fetal-type specificities
The initial purpose in carrying out these chimera studies was to determine whether the enrichment of the MZ with N^– cells (Fig. 1 B) was caused by repertoire-based selection of these specificities. Our data demonstrate that the MZ compartment preferentially enriches for an N^– repertoire (Fig. 3 A). This is consistent with our sequence data (Fig. 1 B) and demonstrates that N^– fetal-type specificities, with their inherently shorter CDR3s, are selected into the MZ based on their BCR. This relative preference for N^– BCRs contrasts with FO B cells (Fig. 3 A) and other mature B cells in the BM (i.e., Fr. F; Fig. 3 A), which are enriched for the N⁺ repertoire. The reconstitution ratio among peritoneal B2 B cells resembled that in the splenic FO compartment (unpublished data). Thus, MZ B cells are unique among mature B2 B cells in their preference for N^– CDR3s.

Differential selection for the N⁺ and N^– repertoires within the FO I and FO II subsets
Two functionally distinct FO B cell compartments have recently been reported that are differentiated based on levels of IgM expression (27). FO I cells (IgM^low), which encompass the majority of mature FO cells, are thought to be Btk dependent, and strong BCR signals facilitate their development (7, 27). In contrast, FO II cells (IgM^high) have been proposed to represent an earlier stage capable of differentiating into either FO I or MZ cells depending on the BCR and intracellular signals they receive. Like MZ cells, FO II cells are proposed to develop in the absence of Btk.

We therefore subdivided our FO compartment into FO I and FO II subsets (Fig. S10). The N⁺ repertoire is preferentially enriched in the FO I compartment, whereas the FO II population is skewed in favor of the N^– repertoire (Fig. 4). This lower frequency of N⁺ CDR3s is consistent with the hypothesis that these FO II cells may give rise to MZ cells. As our FO gate includes some CD23⁺CD21^Hi cells, a small proportion of the FO-gated cells possessed an MZP phenotype, and these cells exhibited a similar selection ratio to that of the MZ compartment. These data show that H-CDR3 repertoire-based selection may contribute to this proposed mature cell fate branchpoint and that the TdT^–/– repertoire is disfavored by the FO compartment. This hypothesis is supported by our finding that intact TdT^–/– mice have smaller FO compartments than age-matched TdT^+/+ mice (Fig. S11, available at http://www.jem.org/cgi/content/full/jem.20080559/DC1).

View larger version (10K): [in this window] [in a new window] [Download PPT slide]   Figure 4. The TdT+/+/TdT–/– ratio differs in splenic FO I and FO II subsets. Data were normalized with respect to the combined Fr. A–D compartment (dotted line; n = 21 mice). *, P < 0.05; **, P < 0.01. Error bars represent SEM. The FACS gating strategy is shown in Fig. S10 (available at http://www.jem.org/cgi/content/full/jem.20080559/DC1).

View larger version (13K): [in this window] [in a new window] [Download PPT slide]   Figure 5. BrdU labeling kinetics in splenic B cells of chimeric mice. Continuous BrdU labeling was performed in mice in the final 6 d of a 5-wk reconstitution period (n = 2 mice per time point). Data were compiled by first identifying the TdT+/+ or TdT–/– cells in each splenic B cell population using the Ly5 congenic marker and then determining the proportions of BrdU+ cells in each TdT+/+ and TdT–/– subset separately. Data represent the mean ± SEM. Statistical significance is shown.

View larger version (50K): [in this window] [in a new window] [Download PPT slide]   Figure 6. The T1 compartment contains a discrete population of T1-MZP cells. (A) Splenic B cell populations were identified using the Allman and Hardy criteria (29). A discrete IgMHiCD21Hi population of cells is present within the T1 compartment. (B) B cell compartments were compared for their expression of several surface markers, including some that are expressed at high levels by the MZ compartment. These analyses were performed in intact B6 TdT+/+ mice (n = 5 mice). (C) TNP-Fluorescein-Ficoll binding by splenic B cells was assessed by flow cytometry (n = 3 mice). Data represent the mean ± SEM.

	DISCUSSION

Top ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS REFERENCES

 
A key finding of our study is that MZ and FO B cell compartments maintain a preference for distinctly different BCR repertoires. Although our comparison of FO and MZ V_H7183 and D_H gene usage suggests that, broadly speaking, the MZ IgH repertoire is quite diverse and employs similar V_H and D_H genes, our sequence analysis revealed that a subset of MZ cells do, in fact, possess a restricted, fetal-type repertoire characterized by the absence of N-region additions in the H-CDR3. Our mixed BM chimera approach using adult BM from TdT^+/+ and TdT^–/– mice demonstrated that the enrichment of these cells is caused by a selective preference in the MZ for B cells with BCRs lacking N nucleotides in their H-CDR3 junctions. In addition, our chimera data show significant differences in the proportions of TdT^+/+ and TdT^–/– cells among B cell subsets in the spleen and BM. These data challenge the simple linear pathway of B cell differentiation and suggest multiple pathways of differentiation caused by repertoire-based selection.

Two recent studies suggest that immature/transitional phenotype cells present in the BM represent a distinct pathway from that in the spleen and that these immature cells have the potential to mature in situ into mature BM B cells (26, 31). Our chimera data provide strong support for this hypothesis as well as a repertoire-based rationale for this bifurcation of the transitional cells into a splenic or BM developmental pathway. As the BM transitional and mature Fr. F compartments have a much greater preference for TdT^+/+ B cells than any splenic subsets, this suggests to us that the BM transitional pathway may primarily culminate in situ into mature Fr. F cells. The reported Btk dependence of BM transitional cells and Btk independence of splenic FO II cells also supports a largely distinct nature for these pathways (26, 27).

It has been reported that mature B cells in the BM reside in an extravascular perisinusoidal niche (32). Similar to the MZ, this location should provide these cells with ample opportunity for interaction with blood-borne pathogens. Indeed, it has been shown that humoral immune responses directed against blood-borne pathogens can be initiated in situ in the BM (32). Assuming that the BM and the MZ are both important sites for monitoring of blood-borne infections, our data clearly indicate that the BCR repertoires available in these distinct locations to fulfill this role are quite different.

The splenic FO and BM Fr. F compartments are both considered to form part of the recirculating pool of B cells in the body. Recent parabiotic studies strongly support the notion that mature BM B cells can recirculate (32). Therefore, it is surprising that increased BCR signaling, such as in Aiolos and other mutant mice (7, 33), results in the loss of mature BM B cells but the preservation of splenic FO cells. This suggests that the BCR signaling requirements for entering into and being retained in these compartments are qualitatively different. Furthermore, if recirculation of cells between the BM and spleen was both frequent and thorough, one might expect the reconstitution ratios between these compartments in chimeric mice to more closely resemble each other. Hence, the highly significant differences in reconstitution ratios of TdT^+/+/TdT^–/– B cells between the Fr. F and FO compartments are consistent with there being a selective repertoire-based retention of some mature B cells in the BM. In contrast, the similarity between reconstitution preferences of the FO and peritoneal cavity B2 compartments supports there being greater homogeneity in the repertoire preferences and/or the recirculation between these locations.

A signal-strength hypothesis has been proposed based on the analysis of mice that are deficient in a variety of signaling molecules (7). Broadly speaking, this suggests that, within a permissible range, weaker signals promote MZ differentiation, whereas stronger BCR signals favor FO and B1 cell fates, although one study suggests that the FO cell fate is the default cell fate in the absence of BCR signals (12). Selection based on signal strength would clearly result in the differential assortment of specificities into the mature B cell compartments. When considering our data in the context of a signal-strength model, it is perhaps of some import that the splenic T3 compartment was enriched in TdT^+/+ cells. It has been proposed that this compartment represents a reservoir for anergic and autoreactive B cells and not an intermediate developmental stage that gives rise to mature B cells (34, 35). Therefore, our data may reflect an inherently higher degree of self-reactivity in the TdT^+/+ repertoire. Consistent with this, autoimmune-prone mice exhibit decreased disease severity and mortality when TdT deficiency is introduced, which provides a compelling argument for reduced harmful autoreactivity in the TdT^–/– repertoire (36–38). It has been suggested that the fetal repertoire is enriched in auto- or polyreactive specificities (39); however, it may be that such polyreactivity is beneficial as opposed to pathogenic. It has been suggested that shorter CDR3s will leave more space in the binding site for antigen to enter and to contact the CDRs 1 and 2 (40), thereby perhaps increasing CDR3 promiscuity. Furthermore, mouse D_H genes predominantly use reading frame 1 (RF1) but N^– CDR3s are even more biased in their RF1 usage. Therefore, it may be that skewed RF usage combined with their shorter CDR3s gives N^– cells the selective suitability for being selected into the MZ compartment.

From the immature B stage onwards, TdT^–/– B cells in our chimeric mice more frequently used Ig. This might suggest that the TdT^–/– repertoire requires more receptor editing than the TdT^+/+ repertoire because increased Ig usage is a hallmark of receptor editing. However, the transit time through the pre-B compartment is the same for TdT^–/– and TdT^+/+ cells in intact mice (unpublished data), arguing that there is not increased receptor editing in TdT^–/– cells. It has been suggested that there is a direct correlation between CDR3 length and the potential for interaction between individual heavy and light chains (41). Therefore, it may be that some N^– H-CDR3 have an inherently better "fit" with Ig chains. Indeed, there have been several reports where IgH using V genes common in fetal life can only poorly associate with SLC (42, 43).

Because the proportions of N^– H-CDR3 were similar for both productive and nonproductive sequences in intact mice as well as in adult mice reconstituted with adult BM, this suggests that these N^– H-CDR3 are likely to arise from TdT-deficient precursors present within the adult BM. Although a portion of N^– MZ CDR3 in intact mice may genuinely derive from the fetal period, the source of the N^– B cells produced in the adult mouse is uncertain. These N^– B cells from the BM may simply arise by chance and then be selected into the MZ by virtue of their N^– BCR. Alternatively, the possibility exists that adult BM may contain an as yet unidentified discrete TdT-deficient B cell precursor population, the progeny of which are preferentially selected into the MZ. Because we did not deplete our donor BM cells of CD19⁺ cells, a possible source of N^– precursor cells could be the recently reported CD19⁺B220^– B1 precursor population (9). However, this precursor population did not appear to give rise to MZ cells in transfer studies (9), and it is unknown whether this small population in the adult BM expresses TdT. Interestingly, the extent of enrichment of both N^– specificities and Ig usage that we observed in the MZ is very similar to that reported for B1 cells (44, 45). In addition, the MZ and B1 compartments are both enriched for antibacterial and anticarbohydrate specificities and are thought to act as a first line of defense against invading pathogens (6). Therefore, it is intriguing to speculate that this "fetal-type" subset of MZ cells contributes to the reported functional similarities between B1 and MZ cells. Furthermore, we propose that a fraction of B1 and MZ N^– cells may be replenished from N^– precursors in the adult BM.

A splenic precursor population for MZ cells was first tentatively identified because of its absence in several strains of knockout mice lacking MZ B cells (11, 46). Subsequent reports support the hypothesis that this population gives rise to MZ cells (7, 28). It was termed T2-MZP because of its similarity to T2 cells. However, considering the similar IgH selective preferences exhibited by the T1 and MZ compartments in our chimeric mice along with the similarities between MZ and T1 cells in surface phenotype (both are IgM^HiIgD^LoCD23^–), we considered that a more direct pathway of differentiation from the T1 to the MZ compartment was also likely. By inspecting the T1 compartment for CD21^Hi cells, we have identified a discrete subpopulation which is a likely intermediate population along a direct T1MZ differentiation pathway. We designate these as T1-MZP cells. Although there is strong evidence that CD21^HiCD23⁺ splenic B cells have MZ precursor potential, we propose that some MZ differentiation can also start even earlier, at the T1 stage of splenic B cell development, and that these cells may bypass the T2 (CD23⁺) stage altogether, resulting in heterogeneity in the origin, repertoire, and, possibly, function within the MZ compartment. This is supported by the phenotypic similarities between T1-MZP and MZ cells and the similarities in IgH selective preferences between the T1 and MZ compartments of our chimeric mice. We have integrated the findings of this study with the work of others and present a model for repertoire-based B cell development in Fig. 7.

View larger version (41K): [in this window] [in a new window] [Download PPT slide]   Figure 7. Model for B cell development incorporating H-CDR3 selective preferences. This schematic merges our repertoire data with other models of B cell development (5, 26, 27, 31). We have included the additional dimension of H-CDR3 repertoire-based selection along the y-axis of this figure.

	MATERIALS AND METHODS

Top ABSTRACT RESULTS DISCUSSION MATERIALS AND METHODS REFERENCES

 
Mice.
All mice used in this study were bred and maintained at The Scripps Research Institute, and these animal studies were approved by The Scripps Research Institute Institutional Animal Care and Use Committee. TdT^–/– and BAFF transgenic mice were backcrossed to C57BL/6 for 10 generations (47). Unless otherwise indicated, mice used in this study were aged between 6 and 10 wk.

Flow cytometry and cell sorting.
Fc-receptors were blocked with anti-CD16/32 (BioLegend), and then cells were stained with different combinations of the following antibodies: CD19-PacificBlue/-PECy7, CD23–Alexa Fluor 488/-PE, B220-PECy7, CD180-PE, and IgD-PE (clone 11-26c.2a; BioLegend); CD93-APC (clone AA4.1), CD45.1-PECy5.5, IgM-PE/-APC/-PECy7 (clone 11/41), CD23-PECy7, CD1d-PE, CD36-PE, and CD11a-PE (eBioscience); and -PE, -biotin (clone R26-46), CD21-FITC/-PE, and CD43-FITC (BD Biosciences). These antibodies were used in combination with streptavidin-conjugated PE (Biomeda), APC (Biomeda), or Pacific Blue (BioLegend).

FACS analysis was performed on a digital LSRII (BD Biosciences) and data were analyzed using FlowJo software (Tree Star, Inc.). Cells were sorted on a FACSDiva or FACSAria (BD Biosciences). Sequence data from intact C57BL/6 and BAFF transgenic mice are derived from two to three independent cell sorts.

DNA and RNA isolation, cDNA preparation, PCR, cloning, and sequencing.
DNA and RNA were isolated from cell sorter–purified populations using AllPrep DNA/RNA kits (QIAGEN). PCR amplifications of DNA were performed for 40 cycles. Primers used for V_H7183 V(DJ) rearrangements were AF303 (5'-CAGCTGGTGGAGTCTGGGGGA-3') and AF591 (5'-CTTACCTGAGGAGACGGTGA-3'). 5–15 independent PCRs were performed on each sample to minimize PCR-generated duplicates, and at least two independent preparations of sorted cells were used. PCR products were gel purified and ligated into TOPO vectors (Invitrogen). Minipreps were made using QIAGEN kits and were sequenced with ABI-automated capillary sequencers (Thermo Fisher Scientific). The CDR3 was defined as including codon 95 through codon 102. Sequences that conformed to the triplet RF were identified as productive sequences. During sequence analysis, assigning an identity to prospective D_H genes required a minimum of 5-bp homology with known D_H genes. Nucleotides that could not be accounted for by V_H, D_H, or J_H genes or their associated P nucleotides were categorized as N nucleotides. Duplicate sequences were only counted once, and no additional sequences were subsequently obtained from such PCRs.

H-CDR3 spectratype analyses.
cDNA was synthesized with QuantiTect Reverse Transcription kits (QIAGEN). A V_HJ558 primer AF572 (5'-TCCAGCACAGCCTACATGCAGCTC-3') in combination with the J_H primer AF591 were used to amplify cDNA transcripts. PCR amplifications were performed for 40 cycles. A 5' 6-FAM–labeled AF591 primer was subsequently used to fluorescently label PCR fragments, and samples were processed on an ABI 310 Genetic Analyzer (The Scripps Research Institute Center for Nucleic Acid Research). Spectratype data were analyzed using ABI Prism Genotyper software (Applied Biosystems).

Chimeric BM reconstitutions.
Lethally irradiated (1,100 rad) C57BL/6 TdT^–/– mice were used as recipients for BM transfers. Donor cells were harvested from the BM of 6-wk-old C57BL/6 TdT^+/+ (Ly5.1), and TdT^–/– mice and were depleted of B cells using anti–B220-PE, and anti-PE–conjugated magnetic beads (Miltenyi Biotec). 5 x 10⁶ cells were injected into the tail vein of each irradiated mouse, and chimeric mice were analyzed 5 wk after irradiation.

The data normalization process for our chimera data was a direct single step conversion to take into account variations in the actual reconstitution ratios of individual mice. In Fig. 2, we present data normalized with respect to the Fr. A compartment. After demonstrating that there was no difference between Fr. A, Fr. B+C, and Fr. D, we subsequently normalized the datasets with respect to the combined Fr. A–D compartments to facilitate the combined presentation of data from experiments using different BM surface-staining procedures that did not always allow the separate identification of the Fr. A compartment.

In mice reconstituted with equal proportions of TdT^+/+ and TdT^–/– BM precursors, if the reconstitution ratio in the combined Fr. A-D of a mouse was 0.7, and not precisely 1 as expected, then the Fr. A–D compartment is normalized to 1 and the reconstitution ratios in all other compartments of that mouse are multiplied by (1/0.7) to maintain their relativity. The normalized reconstitution ratios in each individual B cell compartment were averaged for all replicates, and the variability in reconstitution ratios among individual mice is directly reflected in the SE bars in the figures.

BrdU labeling.
The BrdU-labeling kinetics of spleen cells was assessed in TdT^+/+/TdT^–/– chimeric mice. These mice received i.p. injections of 100 µl (10 mg/ml) BrdU (Sigma-Aldrich) every 12 h for the duration of the indicated exposure periods. After tissue harvesting, cells were processed using BrdU flow kits (BD Biosciences) according to the manufacturer's protocol.

TNP-Fluorescein-Ficoll binding assay.
Mice were injected (tail i.v.) with 500 µl of 1 mg/ml TNP-fluorescein-ficoll (Biosearch Technologies, Inc.) in PBS. 1 h later, mice were killed and spleens were harvested. Single cell suspensions were then surface stained for B cell subset markers. FACS analysis was performed on a digital LSRII, and TNP-ficoll binding by B cells was revealed by fluorescein fluorescence.

Statistical analysis.
Analyses were performed using Prism 4 (GraphPad Software, Inc.). Where appropriate, differences between populations were assessed by ², Fisher's exact test (two-tailed), one-way analysis of variance (ANOVA) with Bonferroni's Multiple Comparison test, Student's T test, and two-tailed or two-way ANOVA. Data are represented as means ± SEM (where applicable).

Online supplemental material.
Fig. S1 shows that TdT deficiency restricts H-CDR3 length. Fig. S2 demonstrates that BAFF overexpression does not alter the levels of N^– H-CDR3 in FO and MZ B cell populations. Fig. S3 shows V_H7183 and D_H gene usage in MZ and FO B compartments. Fig. S4 shows the amino acid usage and hydrophobicity of MZ and FO H-CDR3. Fig. S5 displays the somatic hypermutation levels in MZ and FO H-CDR3. Figs. S6–10 show the flow cytometry gates used to identify individual B cell subsets. Fig. S11 demonstrates that C57BL/6 TdT^–/– mice have smaller FO compartments than wild-type mice. Online supplemental material is available at http://www.jem.org/cgi/content/full/jem.20080559/DC1.