One healthy female subject, age 56, was recruited for antibody repertoire high throughput sequencing using the 454 platform. The subject is Caucasian, a lifelong native of the state of Alabama, and was without a history of illness or repeated infection that could be related to abnormal immune function. The complete blood count was well within normal limits. Serum immunoglobulin levels were IgM 382, IgG 1,680, and IgA 368 mg/dL, respectively. Venous blood (100 cm³) was drawn by routine venipuncture and mononuclear cells were isolated using Ficoll-Paque Plus (GE Healthcare). CD19⁺ magnetic beads (Miltenyi Biotec MACS) were used to enrich for B cells. These CD19⁺ cells were further fractionated by CD27^± populations using CD27 magnetic beads (Miltenyi Biotec MACS) according to the manufacturer’s protocol. CD19⁺CD27⁺ B cells were stained with CD19 APC₇₈₀ (eBioscience), CD27 PE–Cy7 (BD Pharmingen), CD24 APC (BioLegend), and IgD FITC (Southern Biotech), and sorted into IgD⁺ memory B cells (CD19⁺/CD27⁺/IgD⁺/CD24⁺), IgD⁻ memory B cells (CD19⁺/CD27⁺/IgD⁻/CD24⁺), and plasmacytes (CD19⁺/CD27⁺/CD24⁻) using a high speed sorting cytometer (FACSAria III; Becton Dickinson). CD19⁺/CD27⁻ B cells were stained with CD19 APC₇₈₀ (eBioscience), CD24 APC (BioLegend), CD38 PE (BioLegend), and IgD FITC (Southern Biotech) and sorted into mature/naïve (CD19⁺/CD27⁻/IgD⁺/CD38⁺/CD24⁺), transitional (CD19⁺/CD27⁻/IgD⁺/CD38⁺⁺⁺/CD24⁺⁺⁺), and immature (CD19⁺/CD27⁻/IgD⁻) B cell subsets. Each B cell subset was then individually resuspended in 1 mL TRI reagent (Ambion) and archived at −80°C until processed for total RNA extraction. This work was performed in accordance with an Institutional Review Board approved protocol and informed consent was obtained from the subject at the University of Alabama at Birmingham, Birmingham, AL, USA.

For RNA extraction, 0.2 mL chloroform was added to the 1 mL sample, vortexed for 15 s, left to stand at room temperature for 5 min, then spun at 12,000 × g for 10 min at 4°C. The aqueous phase (~400 μL) was removed and to this an equal volume of 70% ethanol was added and then mixed by pipetting. This was applied immediately to an RNA-binding silica spin-column and subsequently processed according to the manufacturer’s protocol (Qiagen RNeasy micro column; catalog no. 74004). Purified total RNA was eluted in 14 μL RNase-free water. Oligo-dT primer was used to generate first-strand cDNA from ~100 ng input RNA using the SuperScript RT II synthesis kit (Invitrogen; catalog no. 11904-018) per the manufacturer’s protocol.

FastStart high fidelity PCR system (Roche; catalog no. 03-553-361-001) and an equimolar mix of eight optimized VH-FWD primers previously described for human IgH amplification (39, 40) coupled with a multiplex of 10-nucleotide uniquely barcoded CH-REV primers: IgM-rev, 5′-10 nt ID-GGTTGGGGCGGATGCACTCC-3′, and IgG-all-rev, 5′-10 nt ID-SGATGGGCCCTTGGTGGARGC-3′ were used to amplify V(D)JCμ and V(D)JCγ cDNAs from the cDNA template. Cycling conditions were as follows: 95°C denaturation for 3 min; 92°C for 1 min, 50°C for 1 min, 72°C for 1 min for 4 cycles; 92°C for 1 min, 55°C for 1 min, 72°C for 1 min for 4 cycles; 92°C for 1 min, 63°C for 1 min, 72°C for 1 min for 22 cycles; 72°C for 7 min. PCR amplicons were gel-purified (Zymo Research) before sequencing.

The University of Texas Genomics Sequencing and Analysis Facility performed Roche GS-FLX 454 deep sequencing. CH-REV barcodes were examined to verify the integrity of each library after filtering raw data for read quality. Sequences were submitted to the ImMunoGeneTics (IMGT) database and IMGT/high V-QUEST web-based analysis tool (version 1.0.3) (41). The 11 CSV text files outputted by IMGT/highV-QUEST were then imported into IgAT immunoglobulin analysis tool for further deconstruction (42). Differences between populations were assessed, where appropriate, by Student’s t-test, two tailed; Fisher’s exact test, two tailed and d; χ², or Levene’s test for the homogeneity of variance. Analysis was performed with PRISM version 5 (Graph Pad). The standard deviation accompanies mean. Raw 454 sequence files were deposited to the NCBI Sequence Read Archive (Accession SRP037774).

CD19⁺ cells bearing the cell surface markers characteristic of immature, transitional, mature, memory IgD⁺, memory IgD⁻, and plasmacytes were isolated from the blood of a healthy female subject (43–47) (Figure 1). Following total RNA extraction, PCR was used to amplify cDNA copies of V(D)JCμ and V(D)JCγ transcripts using optimized VH-FWD primers previously described for human IgH amplification (39, 40). We obtained a total of 15,433 immature, 37,396 transitional, 47,781 mature, 43,558 memory IgD⁺, 28,142 memory IgD⁻, and 43,824 plasmacyte unique and in-frame IgH heavy chain reads. Of these, we obtained 1,240 immature, 1,354 transitional, 1,250 mature, 1,244 memory IgD⁺, 833 memory IgD⁻, and 1,714 plasmacyte reads that were of sufficient length to be identified as Igμ sequences, and 1,879 memory IgD⁻ and 3,347 plasmacyte reads that were of sufficient length to be identified as Igγ sequences. All of the unique Igμ and Igγ reads were deconstructed to assess the presence and extent of changes in these repertoires that had occurred as B cells progressed through the various developmental checkpoints.

The immature B cell subset is primarily composed of recent bone marrow emigrants. It expressed a highly diverse repertoire that differed from the subsequent transitional stage in that it contained the smallest contribution of germline V and J gene sequence to the CDR-H3 region (Figure 2). By family, V_H4 gene segments contributed the most, followed by V_H3, V_H1, V_H5, V_H2, and V_H6 (Figure 3). By individual V gene segments, V1–18, V1–69, V3–73, and V4–59 were most common. Across subsets, the immature B cell subset was enriched for V1–18, V3–30–3, and V3–74 (Figure 4). By D_H family, D_H3 was the most common, followed by D_H2 and D_H6 (Figure 5). By individual D gene segment, D2–2, D3–3, D3–22, D6–13, and D6–19 were favored. Across subsets, D1–26, D2–15, D3–10, D4–23, and D5–12 were more commonly used in the immature B cell lineage (Figure 6). By J_H gene segment, J_H4 was the most common, followed by J_H6, J_H5, and J_H3. Across subsets, immature B cells used J_H5 more frequently (Figure 7).

Amino acid usage in the CDR-H3 loops expressed by these immature B cells varied within a narrow range. When compared to transitional cells, immature B cells used less arginine, asparagine (p = 0.02), aspartic acid (0.04), glutamine (p = 0.009), glutamic acid (p = 0.02), tyrosine (p = 0.002), threonine (p = 0.0039), cysteine (p < 0.0001), and leucine (p = 0.02) (Figure 8). As a result of the decrease in the use of hydrophobic and hydrophilic amino acids, the immature repertoire exhibited the lowest prevalence of highly hydrophobic (hydrophobicity >0.7) CDR-H3 loops (p < 0.05) and the lowest prevalence of the highly hydrophilic (hydrophobicity ≤0.7) CDR-H3 loops of the six subsets examined (Figure 9).

Of the six subsets examined, the transitional CDR-H3 repertoire was the most heavily enriched for longest CDR-H3 loops (Figure 2). This bias for increased length reflects greater preservation of V(D)J gene segment sequence (Figure 2). Conversely, transitional B cell CDR-H3s were enriched for N nucleotide addition, averaging total 19.68 nucleotides and 9.75 nucleotides at the D → J junction (Figure 2). This was the first in a general pattern of diminishing N addition with maturation. Compared to the immature B cell fraction, there was a significant decrease for V_H1 family gene segments (p < 0.001) (Figure 3). By V gene segment, the use of V1–69, V2–26, V3–7, V3–11, V3–15, V3–21, V3–30, V3–33, V3–NL1, V4–28, V4–31, V4–61 was greater than in immature B cells, whereas use of V1–2, V1–3, V1–8, V1–18, V1–24, V1–46, V1–58, V2–5, V3–9, V3–21, V3–23, V3–30–3, V3–48, V3–66, V3–73, V3–74, V4–34, and V4–39 was decreased (Figure 4). The transitional B cell CDR-H3 loop utilized higher levels of D_H6 gene segments (not significant), with lower levels of D_H5 (p = 0.005) than immature B cells (Figure 5). By D gene assignment, a significant increase in D1–1 (p = 0.09) and D2–2 (p = 0.0002) usage in transitional B cells was observed when compared with the immature fraction, with a compensatory decrease in D1–26 (p = 0.01), D2–15 (p < 0.0001), D3–10 (p = 0.005), D4–23 (p = 0.0026), and D5–12 (p = 0.0004) (Figure 6). The use of J_H6 (p = 0.0008) was greater than in immature B cells, while the use of J_H4 (p = 0.01) and J_H5 (p = 0.09) was decreased (Figure 7).

CDR-H3 loops of these transitional cells used more arginine, lysine, asparagine (p = 0.02), aspartic acid (p = 0.04), glutamine (p = 0.009), glutamic acid (p = 0.02), tyrosine (p = 0.001), threonine (p = 0.003), cysteine (p < 0.0001), and leucine (p = 0.02), while using less tryptophan, serine, glycine, alanine, methionine, and phenylalanine than immature B cells (Figure 8). Of the six subsets studied, transitional B cells exhibited the higher prevalence of charged sequences as compared to the immature fraction (Figure 9). The contrast to the immature population was the most striking, suggesting specific gain of charged CDR-H3s in the transition from the immature to the transitional B cell stage. Conversely, the prevalence of highly hydrophobic CDR-H3s increased when compared to the immature B cell fraction.

The mature B cell population was at the median for total CDR-H3 length and for the relative contributions of germline (Figure 2). Conversely, mature B cell CDR-H3s were enriched for N nucleotide addition, averaging 18.22 nucleotides total and 10.22 nucleotides at the V → D junction (Figure 2). In comparison to the transitional B cell repertoire, mature B cells exhibited similar expression of V_H family gene usage (Figure 3). An increase in V4–59 (p = 0.01) and a decrease in the use of V4–61 (p < 0.0001), respectively, were observed when compared to the transitional and mature fractions (Figure 4). Use of the D_H2 (p = 0.01) family in general, and the D2–2 gene segment (p = 0.01) in particular, was lower than in transitional cells (Figures 5 and 6). There was an increase in the use of J_H1 (p = 0.0004) with a decrease in the use of J_H6 (p = 0.002) (Figure 7).

CDR-H3 loops demonstrated an increase in the use of glutamine (p = 0.007), with a decrease in tyrosine (p = < 0.0001), cysteine (p = 0.0001), and valine (p = 0.04) (Figure 8). As a result, the mature B cell repertoire was enriched for the use of hydrophobic and charged CDR-H3 loops when compared with immature and transitional subsets (Figure 9).

The Igμ repertoires of the memory IgD⁺ and memory IgD⁻ blood B cells were distinguishable and divergent from both mature B cells and from each other. The memory IgD⁺B cell CDR-H3 region exhibited a greater contribution of germline D_H and J_H gene sequences than memory IgD⁻ (Figure 2). Memory IgD⁺ B cells used V_H4 (p = 0.008) family gene segments more frequently than memory IgD⁻ B cells, and V_H1 (p = 0.03) family gene segments less frequently. The memory IgD⁻ B cells used V_H1 (p = 0.03) gene segments more frequently and V_H4 (p = 0.03) gene segments less frequently than mature B cells (Figure 3). By individual gene V_H gene segment, the most prominent differences between memory IgD⁺and IgD⁻ reflected increased use of V3–23 (p = 0.0003), V4–34 (p = 0.001), V4–61 (p = 0.004) in the former, and decreased use of V1–8 (p = 0.002) and V4–74 (p = 0.02) in the latter (p < 0.0001) (Figure 4), with the exception of V4–31 (p = 0.02, memory IgD⁻) and V4–59 (p = 0.02, memory IgD⁺ and p = 0.01, memory IgD⁻), which was increased among mature B cells (Figure 4).

Igμ from memory IgD⁺ B cells used D3 (p = 0.01) family D_H gene segments less frequently than memory IgD⁻ cells (Figure 5). When compared with mature B cells, the memory IgD⁺ Igμ repertoire also used D2 and family D_H gene segments more frequently and D3 family D_H gene segments less frequently (not significant). Finally, memory IgD⁻ B cells appeared to use D3 family D_H gene segments more frequently than mature B cells, although this preference did not achieve statistical significance. By individual D_H gene segment, the memory IgD⁺ Igμ repertoire displayed increased use of D6–13 (p = 0.01); and a decrease in use of D3–3 (p = 0.01) (Figure 6). Divergent usage of J_H2 and J_H6 was also observed (Figure 7). The memory IgD⁺ Igμ repertoire used J_H6 more frequently than the memory IgD⁻ (p = 0.001) or mature B cell Igμ repertoire (p = 0.009); and J_H2 (p < 0.0001) less frequently than memory IgD⁻. J_H usage in the memory IgD⁻ Igμ repertoire was very similar to that observed in mature B cells, with the exception of an increase in memory IgD⁻ J_H2 usage as compared to the mature B cells (p = 0.007) (Figure 7).

The CDR-H3 loop of the memory IgD⁺ B Igμ repertoire contained more proline (p = 0.01), tyrosine (p = 0.01), and alanine (p = 0.005); but less arginine (p = 0.001), and tryptophan (p = 0.04) than memory IgD⁻ B cells (Figure 8). The increase in tyrosine reflected increased use of J_H6, rather than increased use of reading frame 1. Indeed, use of reading frame 1, 2, and 3 were similar between the memory fractions (Figure 10). When compared to mature B cells, the memory IgD⁺ Igμ repertoire was similarly enriched for glutamine (p = 0.02) and tyrosine (p = 0.03), and depleted of phenylalanine (p = 0.01). The memory IgD⁻ Igμ repertoire also contained more arginine (p = 0.005) and less proline (p = 0.01) than the mature B cell Igμ repertoire. The memory IgD⁻ Igμ repertoire relatively contained a higher percentage of highly charged CDR-H3s (hydrophobicity >0.7) (1.85%) when compared to the Igμ repertoires of subsequent B cell fractions (Figure 9).

In comparison to the other Igμ and Igγ repertoires, the CDR-H3 component of the plasmacyte Igμ repertoire exhibited the fewest N nucleotides at both the V → D and D → J junctions, respectively. As a result, not only the Igμ repertoire relatively enriched for germline V(D)J sequence, but also exhibited the shortest average length (Figure 2).

By V_H family, plasmacytes exhibited higher usage of V_H4 than either memory B cell population, and lower usage of V_H2, V_H3, and V_H5 (Figure 3). These differences were most affected by increased use of V4–34 (p = 0.007, p < 0.0001) when compared to both the memory IgD⁻ and IgD⁺Igμ repertoires and decreased use of V5–51 (p = 0.01) when compared to the memory IgD⁻ Igμ repertoire (Figure 4).

The distribution of D_H gene family usage among the plasmacyte Igμ repertoire was similar to that of the mature B cell Igμ repertoire, but differed for individual families with the two memory B cell Igμ repertoires. There were no statistically significant differences in the use of D_H gene segments between the memory IgD⁺ and the plasmacyte Igμ repertoires. When compared to the memory IgD⁻ Igμ repertoire, the plasmacyte Igμ repertoire used D_H5 gene segments more frequently (p = 0.04) (Figure 5). By individual D_H gene segment, plasmacytes used D6–25 more frequently (p = 0.006) and D7–27 less frequently (p = 0.03) than mature B cells. Plasmacytes used D6–25 more frequently (p < 0.0001), and D4–11 (p = 0.01), D6–13 (p = 0.04), and D6–6 (p = 0.03) less frequently than the IgD⁺ memory Igμ repertoire. Finally, plasmacytes used D5–24 (p = 0.007) and D6–25 (p = 0.0001) more frequently, and D3–9 (p = 0.03) less frequently than the memory IgD⁻ Igμ repertoire (Figure 6).

By J_H gene segment, the plasmacyte Igμ repertoire displayed similar levels of J gene segments when compared to the mature B cell Igμ repertoire. Plasmacytes expressed higher levels of J_H2 (p = 0.01), J_H4 (p = 0.01); and lower levels of J_H6 than memory IgD⁺ B cells (p = 0.0004). Finally, plasmacytes expressed lower levels of J_H2 (p = 0.04) than memory IgD⁻ B cells (Figure 7).

When compared with the mature B cell Igμ repertoire, plasmacytes expressed lower levels of asparagine (p = 0.02), alanine (p = 0.01), and leucine (p = 0.007) in the CDR-H3 loop. When compared with memory IgD⁺ B cells, plasmacytes expressed lower levels of asparagine (p = 0.001), aspartic acid (p = 0.02), glutamine (p = 0.04), tyrosine (p = 0.001), and alanine (p = 0.0002); and higher levels of tryptophan (p = 0.02) and phenylalanine (p = 0.01). When compared with the memory IgD⁻ Igμ repertoire, plasmacytes expressed lower levels of arginine (p = 0.02), lysine (p = 0.009), and isoleucine (p = 0.02) (Figure 8).

When comparing the relative prevalence of either highly charged or highly hydrophobic CDR-H3 loops, plasmacytes were enriched for charged CDR-H3 loops (0.84%) in comparison to the five other Igμ repertoires (Figures 9 and 11). The distribution of highly hydrophobic CDR-H3 loops decreased in plasmacytes (1.54%) as compared to memory IgD⁻ B cells (1.85%), and returned to the comparable levels of memory IgD⁺ B cells (1.56%) (Figure 9).

The Igγ repertoires expressed by memory IgD⁻ B cells and plasmacytes were distinguishable and uniquely different from each other. While the average length and V(D)J gene segment length was very similar between the memory IgD⁻ and plasmacytes, differences in the N-region additions were observed. The memory IgD⁻ B cell CDR-H3 region exhibited a greater number of N nucleotide addition at the V-D junction (10.56 nucleotides) as compared to the plasmacytes. Conversely, plasmacytes contained more N nucleotide addition at the D–J junction than memory IgD⁻ B cells (10.08 nucleotides) (Figure 2). Memory IgD⁻ B cells used V_H1 (p = 0.03), V_H2 (p = 0.0001), and V_H3 (p = 0.0003) family gene segments more frequently than plasmacyte; and V_H4 (p < 0.0001) family gene segments less frequently (Figure 3). This pattern is due to an increase in individual gene V_H gene segment, the most prominent differences between memory IgD⁻ and plasmacytes reflected increased use of V1–2 (p = 0.03), V1–8 (p = 0.003), V2–5 (p = 0.0003), V3–7 (p = 0.01), V3–15 (p = 0.001), V3–30 (p = 0.005), and V4–40–2 (p = 0.01), in the former, and decreased use of V4–4 (p = 0.0007) and V4–34 (p < 0.0001) in the latter (Figure 4).

The memory IgD⁻ Igγ repertoire used D6 (p = 0.01) family D_H gene segments less frequently than plasmacyte Igγ (Figure 5). By individual D_H gene segment, the memory IgD⁻ Igγ repertoire displayed increased use of D5–24 (p = 0.005) and decreased use of D2–21 (p = 0.03) (Figure 6). The memory IgD⁻ Igγ repertoire used J_H6 less frequently than plasmacytes (p = 0.0006) (Figure 7).

The CDR-H3 loop of the memory IgD⁻ Igγ repertoire contained more asparagine (p < 0.0001) and aspartic acid (p = 0.01); but less tyrosine (p = 0.04), cysteine (p = 0.03), and leucine (p = 0.01) than plasmacyte Igγ (Figure 8). The plasmacyte Igγ repertoire was relatively enriched for hydrophobic amino acids, which was reflected by a higher percentage of hydrophobic CDR-H3s (hydrophobicity > 0.7) (1.54%) when compared to the memory IgD⁻ (1.12%) (Figure 9).

The Igμ and Igγ repertoires of analyzed cell types expressed similar distribution of D_H reading frames, with reading frame 1 having greatest preference, followed by reading frame 2 and reading frame 3 (Figure 10), while the μH chain plasmacytes used reading frame 3 less likely than memory IgD⁻ B cells (p = 0.03) (Figure 10).