Human Memory B Cells Transferred by Allogenic BoneMarrow Transplantation Contribute Significantly to theAntibody Repertoire of the Recipient1

Birgitte F. Lausen,2* Lotte Hougs,† Lone Schejbel,† Carsten Heilmann,* andTorben Barington†‡

The bone marrow is an important source of Abs involved in long-term protection from recurrence of infections. Allogenic bonemarrow transplantation (BMT) fails to restore this working memory. Attempts to overcome this immunodeficiency by immuni-zation of the donor have not been very successful. More needs to be known about transfer of B cell memory by BMT. We trackedmemory B cells from the donor to the recipient during BMT of a girl with leukocyte adhesion deficiency. Vaccination of herHLA-identical sibling donor 7 days before harvest induced Haemophilus influenzae type b (Hib) capsular polysaccharide (HibCP)-specific B cells readily detectable in marrow and blood. BMT did not lead to spontaneous production of HibCP Abs, but therecipient responded well to booster immunizations 9 and 11 mo after BMT. HibCP-specific B cells were obtained 7 days after thevaccinations, and their VH genes were sequenced and analyzed for rearrangements and unique patterns of somatic hypermutationsidentifying clonally related cells. Ninety (74%) of 121 sequences were derived from only 16 precursors. Twelve clones wereidentified in the donor, and representatives from all of them were detected in the recipient where they constituted 61 and 68% ofthe responding B cells after the first and second vaccinations, respectively. No evidence for re-entry of memory clones into the processof somatic hypermutation was seen in the recipient. Thus, memory B cells were transferred from the donor, persisted for at least 9 moin the recipient, and constituted the major part of the HibCP-specific repertoire. The Journal of Immunology, 2004, 172: 3305–3318.

T he recovery of the immune system after allogenic bonemarrow transplantation (BMT)3 requires at least 12 mo tobe completed. Both cellular and humoral immune func-

tions are affected, rendering the patients susceptible to serious in-fections (1, 2). B cells recover slowly in numbers and function andmay be of host or donor origin (3, 4). In cases of complete B cellchimerism, it is currently unknown to what extent the cells arerecruited de novo from the emerging marrow or by expansion ofdonor-derived mature or even memory B cells.

BMT recipients generally become seronegative to previous im-munizations during follow-up (5–9), and long-term immunity de-fined as persistent Ab is not maintained regardless of the immunestatus of the donor (8, 10). Therefore, reimmunization of long-termsurvivors after BMT is necessary (1, 8, 11). Nevertheless, severalobservations indicate that the effect of such immunizations may beinfluenced by the immune status of the donor, and it has been

proposed that immunization of the donor before transplantationcombined with immunization of the recipient may be beneficial forAb protection throughout the posttransplantation period (12).

Evidence for adoptive transfer of immunity from donor to re-cipient by BMT has so far been sparse and indirect. Sharing ofplasma Ab spectrotypes and Ids between donor and recipient hasindicated transfer of immune B cell progeny to the recipient (13–16), but it is not clear whether these cells were transferred asplasma cells or memory B cells. Also, the demonstration of sec-ondary-like Ab responses to the first immunization of the recipientafter BMT points to transfer of immunological memory (13, 17–19). However, these studies yield limited information as to whichcell types are responsible for this transfer (memory B cells, Thcells, Ag-loaded APCs). To settle these questions, studies areneeded using cellular markers coidentifying the level of cellulardifferentiation and clonal relatedness.

In this study, we used clonal tracking based on unique patternsof somatic hypermutations to study the transfer of vaccine-specificmemory B cells during BMT of a girl with leukocyte adhesiondeficiency (LAD) of type 1. We demonstrate that vaccination ofher HLA-identical bone marrow donor 7 days before harvest in-ducedHaemophilus influenzae type b (Hib) capsular polysaccha-ride (HibCP)-specific hypermutated memory B cells, which weretransferred to the recipient by BMT. These cells survived in therecipient for at least 9 mo, constituted a major part of the HibCP-specific B cell repertoire at this time point, and responded to re-peated immunizations of the recipient. Re-entry into the process ofsomatic hypermutation could not be demonstrated.

Materials and MethodsDonor-recipient pair

The patient was diagnosed as LAD type 1 at 7 mo of age. LAD is a rareinherited immunodeficiency characterized by defective adherence and migra-tion of leukocytes caused by mutations in the�2 integrin (CD18)-encoding

*Paediatric Clinic II, Juliane Marie Centre, and†Department of Clinical Immunology,Centre of Diagnostic Investigations, National University Hospital, Rigshospitalet,Copenhagen, Denmark; and‡Department of Clinical Immunology, Odense UniversityHospital, Odense, Denmark

Received for publication December 24, 2002. Accepted for publication December22, 2003.

The costs of publication of this article were defrayed in part by the payment of pagecharges. This article must therefore be hereby markedadvertisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.1 This study has received financial support from the Dagmar Marshall Foundation, theRosalie Petersen Foundation, the Lundbeck Foundation, and the Danish Medical Re-search Council (Grants 22-00-0151 and 22-01-0156).2 Address correspondence and reprint requests to Dr. Birgitte F. Lausen, Paediatric ClinicII, Section 4064, Juliane Marie Centre, National University Hospital, Rigshospitalet,Blegdamsvej 9, DK-2100 Copenhagen, Denmark. E-mail address: lausen@dadlnet.dk3 Abbreviations used in this paper: BMT, bone marrow transplantation; LAD, leuko-cyte adhesion deficiency; Hib,Haemophilus influenzae type b; HibCP, Hib capsularpolysaccharide; MNC, mononuclear cell; TT, tetanus toxoid; CDR, complementarity-determining region; AbSC, Ab-secreting cell; IgSC, Ig-secreting cell; FR, frameworkregion.

The Journal of Immunology

Copyright © 2004 by The American Association of Immunologists, Inc. 0022-1767/04/$02.00

gene (20). The diagnosis was based on the following clinical picture: leuko-cytosis with marked neutrophilia and recurrent bacterial infections (e.g.,omphalitis) combined with complete absence of CD18 on blood mononuclearcells (MNC) as judged by flow cytometry (data not shown). Lymphocyte sub-populations were normal including normal numbers of B cells. There werenormal proliferative responses to mitogens and allogenic cells. Plasma Ig lev-els were normal for the age with easily detectable Abs to tetanus toxoid (TT)and HibCP.

The patient was conditioned by etoposide, busulphan, and cyclophos-phamide, and received at 10 mo of age an unseparated bone marrow graft(2.33 � 108 cells/kg) from her 3-year-old healthy HLA-identical sister.The patient developed signs of veno-occlusive disease at day 19 after BMTbut recovered completely without signs of graft-vs-host disease at any timeafter BMT. IgG substitution was given at days 7 and 31 post-BMT.

The study was approved by the regional ethics committee, and writtenconsent was obtained from the parents.

Vaccination schedule

The donor was immunized 7 days before marrow harvest with HibCP-TTconsisting of 10 �g of HibCP covalently coupled to 24 �g of TT (Act-Hib;Pasteur Merieux Serum et Vaccines, Lyon, France) given s.c. in the su-prascapular region. She had previously received three routine immuniza-tions with TT, but had not received any HibCP-containing vaccine and hadno history of invasive Hib infection. Nevertheless, she carried low levels ofAbs to HibCP (1.1 �g/ml), indicating naturally acquired immunity. At theday of transplantation, heparinized blood and bone marrow aspirate wereobtained from the donor, and MNC were isolated using density gradientcentrifugation.

The recipient had received a single immunization with HibCP-TT at 5mo of age as a part of the childhood immunization program and was im-munized with the same vaccine 9 and 11 mo after BMT as a part of thestudy protocol. Heparinized blood was drawn on several occasions beforeand after transplantation, and MNC were isolated for routine take analysesand complementarity-determining region (CDR)3 length analyses. Further-more, blood was collected 7 days after the vaccinations for purification ofAg-specific cells. Serum samples were obtained before and 3–7 mo afterimmunization from both individuals. Total HibCP Ab levels were mea-sured by ELISA in a commercial laboratory.

Purification of HibCP-specific cells

HibCP-specific B cells including Ab-secreting cells (AbSC) were purifiedfrom blood MNC from the donor and the recipient (and donor bonemarrow) by adherence to paramagnetic polystyrene beads (Dynal, Oslo,Norway) coated with HibCP. The procedure has been described in detailpreviously (21). Some of these cells were immediately tested in enzyme-linked immunosorbent spot-forming cell assay to evaluate the numbers ofHibCP- and TT AbSC and total Ig-secreting cells (IgSC) as describedelsewhere (22). The remaining cells were frozen in aliquots in liquid ni-trogen after addition of an equal volume of freeze medium.

Cloning and sequencing of the H chain mRNA sequences

HibCP bead-purified cells from 5 � 106 MNC obtained on postvaccinationday 7 both from the donor (the purified cell suspension is referred to as

“Do” in the following) and from the recipient after the two experimentalvaccinations (referred to as “Rec1°” and “Rec2°,” respectively) werethawed, washed, and pelleted. RNA was extracted by the guanidiniumthiocyanate method (23), dissolved in sterile water, and stored at �80°Cuntil use. The RNA was used for H chain cDNA synthesis with the GeneAmp RT-PCR kit (PerkinElmer/Cetus, Emeryville, CA). For each vacci-nation occasion (Do, Rec1°, Rec2°), three 10-�l isotype-specific first-strand reaction mixes were set up with final concentrations of 1� PEIIPCR buffer (PerkinElmer/Cetus), 5 mM MgCl2 (Life Technologies, Rock-ville, MD), 1 mM dNTP (Pharmacia, Peapack, NJ), 1 U/�l of RNase in-hibitor (PerkinElmer/Cetus), and 2.5 U/�l of Moloney murine leukemiavirus reverse transcriptase (PerkinElmer/Cetus), and incubated at 42°C for30 min with 1 pmol of the relevant CH1 region primer: IgM133rc,IgA269rc, or IgG264rc (Table I). In addition, a new set of primers corre-sponding to regions situated further downstream in CH1 was designed forcDNA synthesis from mRNA obtained from the second post-BMT vacci-nation of the recipient (Rec2°): IgM140rc, IgA277rc, and IgG274rc (TableI). These primers were used to formally exclude amplification of contam-inant DNA derived from the donor or the recipient after the first post-BMTvaccination.

Four microliters of the isotype-specific first-strand reaction mixes fromeach vaccination occasion (Do, Rec1°, Rec2°) were used as template ineach of six independent 50-�l PCR with final concentrations of 0.2 mMdNTP, 1.5 mM MgCl2, 1� PEII PCR buffer, and 1.25 U of AmpliTaq andanti-Taq (PerkinElmer/Cetus). For each individual vaccination, two lengthsof PCR products were amplified for each isotype (IgM, IgA, and IgG)using 5 pmol/reaction of one of the primers: IgM126rc and IgM133rc (forRec2°, IgM133rc was replaced by IgM140rc) (Table I), IgA185rc andIgA269rc (Rec2°, IgA277rc), and IgG183rc and IgG264rc (Rec2°,IgG274rc), respectively. In all six reactions, 20 pmol of a degenerate VH3-specific signal peptide primer, VH3-sig-deg (Table I), was used. Thisprimer was designed to amplify all known signal peptide sequences of theVH3 family available in the GenBank database and in the directory of Kabat(24). The PCRs were hot-started and after an initial denaturation for 2 minat 94°C, 40 PCR cycles were performed, consisting of 30 s at 94°C, 1 minat 55°C, 1 min at 72°C, and a final 5-min step at 72°C.

The PCR products were cloned into the pCR2.1 vector by use of theOriginal TA Cloning kit (Invitrogen, San Diego, CA) as instructed by themanufacturer. Insert length was analyzed by running a control PCR usingthe flanking primers M13bio and M13RPbio (Table I). Plasmid DNA waspurified using the Quantum Prep Plasmid Miniprep kit (Bio-Rad, Hercules,CA). The plasmids were extracted with phenol/chloroform/isoamyl alcohol(25:24:1; pH 6.7/8.0; Amresco, Solon, OH) before use as template. Cloneswere sequenced corresponding to the entire variable domain (codons1–113) and a part of the first constant domain allowing identification ofisotypes and subclasses using the Ready Reaction kit (PerkinElmer/Cetus)and an ABI 373 automatic sequencer (PerkinElmer/Cetus) according to theinstructions of the manufacturer.

Candidate germline VH genes were assigned to each sequence based onmaximum homology (codons 1–94) with sequences of the GenBank data-base. Similarly, candidate germline JH minigenes were assigned based onmaximum homology from codons 101–113 with the published germlinesequences (25). DH segments were assigned to a germline DH gene if theymatched at seven continuous bases or more than eight continuous bases

Table I. Primers used in the study

Name Nucleotides

3-23c89rc 5�-TCTTTCGCACAGTAATAT-3�3-23cn1 5�-TGTGAGGTGCAGCTGT-3�3-23cn9mut-TET 5�-TET-CTGAGCAGGCTTTTTCTTGTG-3�3-23cn12-G 5�-AGTTTGGGCTGAGCTGGC-3�IgA185rc 5�-GCTGGCTGCTCGTGGTGTA-3�IgA269rc 5�-TGCACGTGAGGTTCGCTTC-3�IgA277rc 5�-AGGCATCTCTCAGGCCGGT-3�IgG183rc 5�-GCTGCTGAGGGAGTAGAGT-3�IgG264rc 5�-GGGTCCGGGAGATCATGAG-3�IgG274rc 5�-CTCACGTCCACCACCACGC-3�IgG140rc 5�-TGGACCAGGCAGCCCA-3�IgM126rc 5�-CGACGGGGAATTCTCACAGG-3�IgM133rc 5�-CCACGCTGCTCGTATCCG-3�IgM140rc 5�-GTCCTGTGCGAGGCAGCCA-3�M13bio 5�-TTTTCCCAGTCACGACGTTGTAAAACGACGGCCAG-3�M13RPbio 5�-AGCGGATAACAATTTCACACAGGAAACAGCTATGA-3�VH3-sig-deg 5�-G(C/T)(T/G)GCT(A/C)T(A/T)(A/T)TA(A/G)(A/G)(A/G)AGGTGTCCA-3�

3306 HUMAN MEMORY B CELLS TRANSFERRED BY ALLOGENIC BMT

except for a single mismatch. DH gene candidates (26) were considered inall reading frames and in both directions. N additions were defined asjunctional bases that could not be assigned to the VH, DH, or JH gene.

The sequences were named according to their origin and isotype andgiven a particular number, e.g., LIBPM322. Sequences from the donorwere named “LIBP,” from Rec1°, “ANBP,” and from Rec2°, “AN2P” . Thenext letter indicates the isotype: M, IgM; G, IgG; and A, IgA.

Estimation of polymerase errors

The frequency of RT PCR mutations was estimated from the sequencesencoded by the C region gene. Of 16,818 sequenced base pairs, 14 uniquesubstitutions were found and considered to be reverse transcriptase or Taqerrors. This led to an overall estimated prevalence of accumulated RT PCRmutations of 1 of 1201 bp, corresponding to 0.2 mutations per variabledomain (codons 1–94; 294 bp) after 40 PCR cycles. This is below thereported frequency of Taq errors of �10�4 substitutions/base/PCR cycle(27).

Amplification of germline gene sequences

Genomic DNA was isolated by a salting-out procedure (28) from 5 � 106

MNC obtained before vaccination of the donor. Five hundred nanograms ofDNA were used for PCR amplification of the 3-23 germline gene using 5pmol of gene-specific primers placed in framework region (FR)1 (3-23cn1)and FR3 (3-23c89rc), respectively (Table I). PCR was performed as de-scribed above but with 30 cycles instead of 40.

Measures to exclude cross-contamination

Great care was taken to avoid mixing of sequences derived from the threeexperimental vaccinations. Besides ordinary PCR precautions such as han-dling of cells and pre-PCR RNA and DNA samples in facilities other than

those used for post-PCR material, handling of samples from the three vac-cinations were separated in time too. Thus, PCR, cloning, and sequencingof material from the three vaccinations were done with weeks to monthsbetween; first from Do, then from Rec1°, and finally from Rec2°. In ad-dition, different primers amplifying different lengths and isotypes of thePCR products were used to allow for detection of colonies not containingthe proper PCR product. No indication of cross-contamination was seenthroughout the study. To formally exclude contamination of Rec2° sampleswith Rec1° or Do samples, specially designed primers (see above) wereused recognizing a template in the C region gene not present in any PCRproduct from the donor or first post-BMT vaccination of the recipient.

Clonal analysis

The software program TREECON for Windows (version 1.3b) was used togenerate a cluster analysis dendrogram based on the neighbor joiningmethod (29). All mutated sequences (codons 1–94) derived from the ca-nonical VH3-23 and 3-49RB germline genes were subjected to analysis. Thepresence of shared and unshared somatically acquired mutations was usedto calculate evolutionary distances and an evolutionary tree minimizingthese was formed using the respective germline sequences as roots.

Analysis of CDR3 lengths

For CDR3 length analysis, mRNA was extracted from thawed MNC usingthe Dynabeads mRNA DIRECT kit (Dynal). The mRNA was used forcDNA synthesis in an 80-�l reaction mix with a final concentration of 1.25�M oligo(dT)16 primer (DNA Technology, Aarhus, Denmark), and reac-tion conditions were otherwise as described above. The cDNA was storedat �20°C until use. The cDNA was used for IgG VH3-23-specific PCR asfollows. Four microliters of cDNA was used in a 20-�l PCR with a finalconcentration of 1� PCR buffer, 1.5 mM MgCl2, 0.2 U/�l PlatiniumTaqpolymerase (Life Technologies), 0.2 mM dNTP, and 0.1 �M isotype-spe-cific CH1 region primer (IgG183rc or IgM133rc) combined with a VH3-23signal peptide primer (0.1 �M; 3-23cn12-G; Table I). PCR was hot-startedand after an initial denaturation for 2 min at 94°C, 22 PCR cycles wereperformed, consisting of 1 min at 94°C, 1 min at 60°C, 1 min at 72°C, anda final 10-min step at 72°C. One microliter of these PCR products wasthereafter used as template in a 50-�l nested PCR with IgG140rc primer orIgM133rc primer combined with a fluorochrome-coupled (TET) VH3-23signal peptide primer 3-23cn9mut-TET (Table I); both primers were in afinal concentration of 0.1 �M. The reaction mix was otherwise as describedabove. PCR was hot-started, and after an initial denaturation for 2 min at94°C, 20 PCR cycles were performed, consisting of 1 min at 94°C, 1 minat 60°C, 1 min at 72°C, and a final 10-min step at 72°C. The lengths of thePCR products were analyzed in a fluorochrome-coupled fragment analysisusing PE Applied Biosystems (Foster City, CA) ABI Prism 310 geneticanalyzer.

Statistics

The hypergeometric distribution was used to evaluate the distribution ofsequences with different joints among the clusters defined by the neighborjoining method. The probability that any of the different joints would ap-pear by chance as often (or more often) as the one found to be mostnumerous in each cluster was calculated. Subgroups were compared by

FIGURE 1. Study design. Time schedule for vaccinations and bloodsamples. The donor was vaccinated 7 days before marrow harvesting andblood sampling (Do). The recipient was vaccinated 9 and 11 mo aftertransplantation. Blood samples were taken 7 days after the first (Rec1°) andthe second (Rec2°) vaccination of the recipient. Act-Hib, Vaccine contain-ing HibCP coupled to TT.

Table II. Numbers of IgSC, and TT and HibCP AbSC 7 days after immunization of the donor and the recipient with HibCP-TT (Act-Hib)a

IgSC and AbSC per Gram of Blood and Bone Marrow

Before purification Recovery after purification

TotalIgSC TT-AbSC HibCP-AbSC

Isotypedistribution

HibCP-AbSCof total

IgSC (%) TT-AbSC HibCP-AbSC

Isotypedistribution

HibCP-AbSCof total

IgSC (%)IgM(%)

IgG(%)

IgA(%)

IgM(%)

IgG(%)

IgA(%)

Donor blood 23,736 950 11,480 16 42 42 48 37 6,117 19 24 57 92 (90–95)Donor marrow 24,975 4,538 8,313 21 44 35 33 106 5,225 35 15 50 94 (92–94)Recipient 1st

immunization17,679 143 1,527 22 29 49 9 0 802 16 6 78 100 (74–100)

Recipient 2ndimmunizationb

14,692 62 338 59 ND 41 2 0 169 48 ND 52 100 (18–100)

a Isotype distribution is indicated for HibCP AbSC. Recovery after purification of HibCP-specific cells are given, and the purity of the latter is given in percentage of totalIgSC (right column). Confidence limits (95%) are given in parentheses (Poisson distribution).

b IgG-secreting cells were not measured.

3307The Journal of Immunology

Fisher’s exact test, the �2 test, and the Kruskal-Wallis test. In all analyses,two-sided p values of �0.05 were considered statistically significant.

ResultsClinical course

After transplantation at 10 mo of age, the patient recovered com-pletely from her immunodeficiency and rapidly developed com-plete chimerism with expression of CD18 on �90% of granulo-cytes, monocytes, T and B lymphocytes, and NK cells at 1 mo afterBMT. Levels of Igs and NK cells were normal by 4 mo post-BMT,and the numbers of B and T lymphocytes were normal after 6 and12 mo, respectively. Cyclosporine, given as posttransplantationimmunosuppression, was discontinued after 6 mo, and the recipi-ent remains healthy with a normal childhood development.

Responses to vaccination

The donor responded well to immunization with a single dose ofHibCP-TT. Serum Abs to HibCP increased from a prevaccinationlevel of 1.1 to 8.5 �g/ml 7 mo after immunization. The recipienthad received a single immunization with HibCP-TT at 5 mo of ageas a part of the childhood immunization program. Before BMT, theHibCP Ab level was 10.3 �g/ml, but 9 mo after BMT, the levelwas low, but detectable (0.31 �g/ml). At this time point, the re-cipient was immunized twice with HibCP-TT (9 and 11 mo afterBMT, respectively) (Fig. 1) and responded well, retaining a levelof 74.7 �g/ml 3 mo after the last immunization. B lymphocyteresponses were also evident from the appearance of HibCP- andTT-specific AbSC in the circulation as measured by an ELISPOTassay 7 days after the immunizations (Table II). Vaccine-specificAbSC were also numerous in the transplanted bone marrow, con-stituting �51% of the total number of IgSC, and 33% of the IgSCwere specific for HibCP. No HibCP AbSC could be detected in theblood of the recipient 3 days before or 7 days after transplantation(data not shown).

Purity of recovered cells

Fifty-three percent of the HibCP-specific AbSC was recovered af-ter purification from the blood after the first experimental immu-nizations of the donor and the recipient. Contamination with IgSCof other specificities was estimated to be low (�8%), assumingthat the recovery of TT-specific AbSC (�4%) was representativefor all cells with specificities other than that of HibCP (30) (TableII). The purity of cells obtained after the second experimental vac-cination of the recipient could not be accurately estimated becauseIgG AbSC were not measured due to technical reasons, but therecoveries of HibCP-specific AbSC of the IgM and IgA isotypeswere similar to those after the first vaccination (data not shown).

Applied VH3 gene repertoire

A total of 121 Ig VH gene sequences from the three experimentalvaccinations (49 IgM, 39 IgG (15 IgG1 and 24 IgG2), and 33 IgA(26 IgA1 and 7 IgA2)) were analyzed. As expected, the responseto HibCP immunization was markedly restricted with respect toVH, DH, and JH gene usage, and only nine different combinationsof these gene segments were found (Table III). Moreover, thespectrum of CDR3 lengths was restricted to one per gene combi-nation. One hundred and three sequences (85%) used one of twogene combinations: VH3-23 rearranged to JH6B1 or to D5-18 andJH4B1 encoding a 6-aa CDR3 with a conserved glycine in position95. These rearrangements are canonical for the human Ab responseto HibCP (31, 32), and they are used by the majority of HibCP-specific AbSC along with a Vk A2-encoded L chain expressing theHibId-1 Id (33, 34). Glycine 95 may be encoded by four differentnucleotide triplets: GGG, GGC, GGA, and GGT. For JH6B1-en- T

able

III.

Num

bers

ofse

quen

ces

usin

gsp

ecifi

cge

neco

mbi

nati

ons

and

CD

R3

sequ

ence

s(j

oint

s)am

ong

121

cano

nica

lan

dno

ncan

onic

alV

DJ

sequ

ence

sfr

omth

edo

nor

(Do)

and

from

the

first

(Rec

1°)

and

seco

nd(R

ec2°

)im

mun

izat

ions

ofth

ere

cipi

enta

Join

ts(N

o.)

VH

Gen

eD

HG

ene

J HG

ene

Sequ

ence

ofC

DR

3(C

odon

s95

–102

)

Num

ber

ofSe

quen

ces

Do

Rec

1°R

ec2°

Tot

al

Can

onic

alre

arra

ngem

ents

1.1

3-23

bJH

6B1

GGCTACGGTATGGACGTC

1820

1250

1.2

3-23

bJH

6B1

GGTTACGGTATGGACGTC

86

317

1.3

3-23

bJH

6B1

GGGTACGGTATGGACGTC

46

515

1.4

3-23

bJH

6B1

GGATACGGTATGGACGTC

47

314

23-

23b

D5-

18c

JH4B

1GGCTATGGTTTTGACTGC

51

17

Non

cano

nica

lre

arra

ngem

ents

33-

49b

D1-

1dJH

3B1

GATTCTAAGGTCGACGGGAAACTGGATGCTTTTGATATT

110

114

3-49

bD

3-10

JH6B

1GATAGAAAGTCTTTCTTTGCTTCGGGGAAATACGGTATGGACGTC

22

53-

49b

D3-

3JH

4B1

GAGCCGTTTTACGATTTTTGGAGTGGAACCCTTGGGTACTACTTTGACTAC

11

63-

21b

D3-

10JH

6B1

GCCTCGGTCTTACTATGGTTCGGGGAGTTATATGTCTACTACTACGGTATG

GACGTC

11

73-

21D

7-27

JH6B

1ATTGGGGGAATAACTGGGGACAACTACTACTACTACTACGGTATGGACGTC

11

83-

23b

DX

P2

JH4B

1GATGAGAGGGTGTATTATGATTACGTTTGGGGGAGTTACTCAGGTTACTAC

TTTGACTAC

11

93-

30D

3-16

�D

6-19

JH6B

1GATCTTAACGAAGGGGACGGGAGAGTCCAGTGGCTGGTCCCCTACTACTAC

TAC4TACGGTATGGACGTC

11

aW

here

pres

ent,

nucl

eotid

esfr

omD

Hge

nes

are

unde

rlin

ed.A

llse

quen

ces

have

been

subm

itted

toth

eE

urop

ean

Mol

ecul

arB

iolo

gyL

abor

ator

yD

ata

Lib

rary

unde

rac

cess

ion

nos.

AJ5

1921

9–A

J519

339.

Join

tnos

.6–9

are

iden

tified

asA

NB

PM22

9,A

NB

PM22

3,A

NB

PM23

0,an

dA

N2P

M45

4,re

spec

tivel

y.b

VH

gene

sha

vebe

enpr

evio

usly

desc

ribe

din

Hib

CP

resp

onse

s(3

1,34

,35

).c

Cou

ldal

sobe

DK

5-5/

DK

4.d

Cou

ldal

sobe

D1-

7or

1-20

.

3308 HUMAN MEMORY B CELLS TRANSFERRED BY ALLOGENIC BMT

coded Abs, the two last nucleotides of these triplets are absent inthe germline segments and therefore are likely to be generated byN addition. Sequences differing in the last nucleotide of codon 95are therefore likely to represent independent rearrangements, al-though (a silent) somatic hypermutation is another possibility. ForJH4B1-encoded Abs, these nucleotides are present in the DH geneand therefore probably templated.

The finding of identical rearrangements among sequences fromdonor and recipient (Table III) is compatible with a clonal rela-tionship between the cells donating the mRNA. However, due tothe canonical nature of most rearrangements and paucity of ex-tended N segments in this response, clonality cannot be safelyinferred. Instead, clonality had to be established by analysis ofshared somatic hypermutations.

Somatic hypermutations

All sequences were compared with the germline VH gene withhighest homology in the GenBank database, and differences wereconsidered to represent somatically acquired hypermutations. Bythese criteria, 119 of 121 sequences were mutated with a medianprevalence of mutations of 5.1% (range, 0.7–7.8%) (codons 1–94).For comparison, the estimated prevalence of RT PCR errors was0.1% (see Materials and Methods). For the predominant VH gene,3-23, identity between the published sequence and the donor germ-line sequence was confirmed by sequencing of 14 cloned germlinePCR products generated by use of VH3-23-specific primers recog-nizing areas of FR1 and FR3 shared by the published germlinegene (data not shown). Thus, substitutions of the 103 canonicalsequences had to be somatically acquired and therefore potentiallyapplicable as clonal markers.

Definition of clonality

Two DNA sequences were considered clonally related if they de-rived from two B cells descending from the same precursor cell inwhich a functional H chain gene rearrangement had occurred. Thefirst criterion for clonal relationship was the requirement of iden-tical VDJ rearrangements. The second criterion was sharing of anumber of mutations not explainable by chance or selection by Ag.To establish safe limits for the numbers of shared mutations de-fining a clone, a sample of published HibCP-specific, canonicalDNA sequences (31, 32, 34, 35) and five sequences from this study

representing five different canonical joints were analyzed for mu-tations shared by chance. Among the 17 unrelated sequences,�5% of randomly selected pairs shared five or more identical mu-tations in the VH gene, and no pair shared more than six mutations(Fig. 2). A provisory requirement of at least eight shared mutationswas used to define a clone in this report. In Fig. 2, the sequencesof the five canonical joints were analyzed by pairwise comparisonsof shared somatic mutations. It is evident that several sequenceswithin each type of joint shared many more mutations than ex-plainable by chance and thus had to be clonally related.

Clonal analysis

Fig. 3 shows a cluster analysis for the 103 canonical sequencesbased on the neighbor-joining method. Fifteen clusters encompass-ing 80 (78%) of the sequences are indicated, in which all se-quences fulfil the second criterion of clonality, i.e., share at leasteight mutations with all the other sequences in the cluster. Becausethe genealogical tree was exclusively based on the V gene se-quences (codons 1–94), codons possibly shaped by the rearrange-ment process (i.e., codon 95) and the JH minigene-encoded codons96–113 could be used to test the reliability of the clustering. For12 of 15 clusters, only sequences with identical codon 95 and JH

gene had been allocated to each cluster. For the six major clusterswhere statistical analysis was meaningful (n � 4), all differed sig-nificantly from expectations if sequences were allocated randomly(Fig. 3; p � 0.03). In cluster C, however, one sequence(LIBPM261) used another JH gene (joint 1.1 rather than 2) indi-cating erroneously allocation to this cluster by chance. Alterna-tively, this sequence could have arisen by a PCR or cloning cross-ing-over artifact. This sequence was not considered truly clonallyrelated to the other sequences of that cluster, and the sequence isomitted in Table IV. The four sequences of cluster A using joint1.3 rather than 1.1 (ANBPM234, AN2PG497, ANBPG258, andAN2PG444), however, are readily explained by a single mutationevent in codon 95 in a precursor B cell in the genealogical tree andtherefore likely clonal members of the cluster despite the dif-ference in codon 95 (Figs. 3 and 4). Similarly, a single mutationevent may explain the presence of a sequence with joint 1.1(ANBPG289) in clone F. In all, these data strongly validate theidentification by this algorithm of 15 clonal progenies comprising79 canonical sequences.

FIGURE 2. Shared somatic mutations. Shared so-matic mutations among canonical HibCP-specificVH3-23 sequences (codons 1–94) analyzed by com-parison of all possible pairs of sequences and enumer-ation of the number of pairs that shared a defined num-ber of mutations. Sequences from donor and recipientwere pooled and divided according to the five jointsrepresenting different glycine codons in position 95and JH gene (Table III). Shared mutations among 17clonally unrelated HibCP-specific canonical VH3-23sequences (codons 1–94) are shown as control for ran-dom sharing of mutations. The presence within eachjoint group of several sequences sharing many moremutations than explained by chance indicates clonalrelationships among the cells from which the se-quences were derived.

3309The Journal of Immunology

FIGURE 3. Legend continues

3310 HUMAN MEMORY B CELLS TRANSFERRED BY ALLOGENIC BMT

Table IV shows the isotypes and origins of the 79 canonicalsequences allocated to the 15 identified clones. The clones con-sisted of from 2 to 20 DNA sequences each (median, 4). The me-dian number of mutations shared by all members of a single clonewas 9 (range, 8–14), but some sequences shared up to 23 muta-tions. Although most mutations tended to be concentrated in cer-tain areas of the gene (hot spots) (see Fig. 4), the clones clearlydiffered with respect to mutational patterns (clonal markers).

A similar analysis could be performed for the noncanonical se-quences. Fig. 5 shows a genealogical tree for the 14 3-49-derivedsequences. One cluster of 11 sequences (clone P) fulfils the secondcriterion of clonality. Unlike the remaining three sequences, the 11sequences share the same CDR3 sequence and the JH gene: JH3B1.Furthermore, they share a median of 16 (range, 13–16) mutationswith each other. Thus, these 11 sequences are definitively clonal.The remaining four noncanonical sequences using VH genes otherthan 3-49 (ANBPM229, ANBPM223, ANBPM230, andAN2PM454) had different joints (joint nos. 6–9 (Table III), re-spectively) and were therefore clonally unrelated.

In all, 90 sequences of the 121 DNA sequences (74%) could beassigned to 16 clonal progenies (Table IV).

Transfer of memory clones by BMT

After definition of the clones, the contributions from the two in-dividuals could be analyzed. The amplified part of the donor’sresponse was largely limited to 12 clones representing at the mostthe progeny of 12 precursor B cells (Table IV). Interestingly, rep-resentatives from all of these clones could be retrieved in the re-cipient (Table IV) constituting 27 sequences (61%) from the first

experimental vaccination of the recipient (Rec1°) and 25 se-quences (68%) from the second (Rec2°). Within most clones,memory cells of the IgM isotype were also transferred as evidentfrom the finding of IgM transcripts in the recipient sharing clonalmarkers with transcripts from the donor. In addition, four cloneswere obtained from the recipient without representatives in thedonor, and three new clones were obtained after the first and oneafter the second experimental immunizations. Nevertheless, �64%of the VH3 family gene sequences used by the recipient for theanti-HibCP responses were clearly derived from mutated (i.e.,memory) B cells transferred from the donor.

Clonal diversification before and after transfer

The numbers of mutations were higher for canonical IgA and IgGsequences (medians of 16 (5.5%) and 17 (5.7%) mutations incodons 1–113 (345 bases), respectively) compared with IgM se-quences (median, 11.5 (3.8%); p � 0.002). Because the isotypedistribution varied with the vaccination situation, switched andunswitched cells were studied separately in relation to the threeexperimental vaccinations. Interestingly, no significant differenceswere seen in the number of somatic hypermutations after the threeexperimental immunizations, neither for unswitched sequences(medians of 12, 11, and 17 for Do, Rec1°, and Rec2 °, respec-tively; p � 0.60; n � 49; Kruskal-Wallis test), nor for switchedsequences (medians of 15, 16, and 18; p � 0.17; n � 72). This wasalso the case when only clonally related sequences were studied(medians of 13, 15, and 17 for unswitched; p � 0.10; n � 31; and17, 16, and 20; p � 0.12; n � 59) and also when the analysis wasrestricted to clones with representatives in both donor and recipient

FIGURE 3. Cluster analysis of canonical sequences. Dendrogram demonstrating a cluster analysis based on the neighbor joining method for the 103vaccine-induced HibCP-specific canonical sequences (pooled data from all three vaccination situations). The germline VH3-23 sequence (codon 1–94) wasused as root. Clusters in which all sequences shared eight or more mutations with all other members of the cluster are indicated by gray background andpresumed to contain clonally related sequences only and are designated clones A–O. The five different joints are indicated by the following symbols: F,joint 1.1 (GGC-JH6B1); �, joint 1.2 (GGT-JH6B1); ‚, joint 1.3 (GGG-JH6B1); �, joint 1.4 (GGA-JH6B1); E, joint 2 (GGC-JH4B1). The validity of theclone assignment was assured by the preferential allocation of sequences with the same joint to each clone and tested statistically against random allocation(see Materials and Methods). �, This sequence was not considered to be a member of clone J, because it shared eight mutations with some but not all ofthe sequences in the clone.

Table IV. Isotype of 90 (of 121) mRNA sequences that could safely be allocated to definitive clones (A–P) and their origin from donor (Do) or fromthe first (Rec1°) or second (Rec2°) vaccination of the recipienta

Clone Joint Donor (Do) Rec1° Rec2°

Canonical clonesA 1.1: GGC-JH6B1 M, M M, M,b G1,b G1, G2, G2,

A1, A1, A1, A2G1, G1, G2,b G2,b

A1, A1, A1, A1B 1.1: GGC-JH6B1 M, MC 2: GGC-JH4B1 M, A1, A1, A1, A2 A2 G2D 1.2: GGT-JH6B1 M M, M G2, G2E 1.1: GGC-JH6B1 M A1F 1.2: GGT-JH6B1 G2, G2, G2, G2, A2, A2 G2, G2, G2, G2, G2c G2G 1.1: GGC-JH6B1 M, M M, MH 1.4: GGA-JH6B1 M, M, A1, A1 G1, G1I 1.1: GGC-JH6B1 G1, G2J 1.4: GGA-JH6B1 G2, G2, G2, G2 A1K 1.1: GGC-JH6B1 A2 M ML 1.1: GGC-JH6B1 M, G1M 1.1: GGC-JH6B1 M MN 1.3: GGG-JH6B1 M A2 A1O 1.1: GGC-JH6B1 M, M M, G1

Noncanonical cloneP 3:3-49-JH3B1 A2 M, M, M, M, M, G1,

G1, G1, A1, A1

a Isotypes and subclasses: M, IgM; G1, IgG1; G2, IgG2; A1, IgA1; A2, IgA2.b Joint 1.3: GGG-JH6B1.c Joint 1.1: GGC-JH6B1.

3311The Journal of Immunology

FIGURE 4. (Legend continues)

3312 HUMAN MEMORY B CELLS TRANSFERRED BY ALLOGENIC BMT

(i.e., exclusion of clones I, J, L, and M; p � 0.19 and 0.14). Neitherwere any significant differences found when members of the larg-est clones were studied separately (Fig. 6, clones A and F; p �0.17). These results were not due to negative selection of aminoacid-replacing mutations, because no difference in the prevalenceof silent mutations was noted either ( p � 0.67 and 0.65 forswitched and unswitched canonical sequences; data not shown).

Also, the shape of the genealogical trees (Fig. 6) are compatiblewith little if any somatic diversification (and selection) betweenthe immunizations. Thus, sequences from the three vaccinationsappear more or less randomly distributed in the trees with no clearindication of sequential acquirement of mutations from vaccina-tion to vaccination. Of the 121 studied sequences, only four pairsof sequences with identical VDJ sequences were found (clone D,AN2PG481 (G2) and AN2PG443 (G2); clone F, LIBPG316 (G2)and ANBPG284 (G2); clone H, LIBPA232 (A1) and LIBPA214(A1); and clone I, AN2PG568 (G1) and AN2PG489 (G2)). Inclones J and P, 3 and 5 sequences with identical VDJ sequenceswere found, respectively. Clone P is somewhat atypical, with 10 of11 sequences seen after the last immunization and none after thefirst post-BMT immunization. However, this apparent repertoireshift is likely to be a sudden amplification of a pre-existing, mu-tated clone rather than a sign of continued intraclonal diversifica-tion. This is judged by the high number of shared mutations (14)combined with the paucity of unique mutations. In fact, 5 of thesequences had identical VDJ sequences. Moreover, 13 of the 14shared mutations were also found in the sequence LIBPA231,which was isolated from the donor, strongly indicating that most ifnot all mutations in clone P had occurred in the donor. Similarobservations were made for clones C, F, G, K, N, and O (Fig. 6),

which all contain donor-derived sequences with high numbers ofmutations similar to or even outnumbering the mutations in recip-ient-derived sequences. In the case of clone F, the two mostheavily mutated sequences were of donor and recipient origin, re-spectively, but shared all mutations, proving that the B cell pro-viding the sequence ANBPG284 definitely did not mutate in therecipient.

Influence on the general Ab repertoire of the recipient

The finding of �33% of the IgSC being of HibCP specificity in thetransferred bone marrow raised the question as to whether the gen-eral Ab repertoire of the recipient was skewed after the transplan-tation. Because almost all of the HibCP-specific cells used thecanonical VH3-23 gene, transcripts of this gene were amplifiedusing H chain-specific downstream primers followed by fluoro-chrome-coupled fragment analysis to reveal the CDR3 lengthspectrum of the molecules. Fig. 7 reveals that the expressed IgGVH3-23 repertoire was almost as diverse after transplantation asbefore (a–e), and that canonical size transcripts found amongHibCP-specific cells from the donor and recipient (g and h) con-stituted only minor fractions (if any) of the repertoire at these timepoints despite the two restimulations with Ag. Similar results wereobtained for IgM and IgA (data not shown) with the exception thatthe IgM repertoires were bell-shaped polyclonal at all time points(example given in Fig. 7f). In conclusion, although the transferredcells had a dramatic effect on the HibCP-specific Ab repertoire,little if any effect was detected on the general Ab repertoire.

DiscussionThe normal bone marrow plays a central role in the maintenance ofhumoral immunity. Not only is it the principal site of B cell lym-phopoiesis after birth (36); it is also a source of high-affinity Absprotecting the individual from reinfection for years. After infec-tions like polio, yellow fever, smallpox, and measles, humans maybe protected by Abs for most of their lives (37–41). How theimmune system manages this remarkable task is not understood.Until recently, it was generally assumed that plasma cells wereshort lived with life spans of a few days or perhaps weeks (42).However, new studies in mice indicate that individual plasma cellsof the bone marrow may live and produce Ab for at least a yearafter immunization, and do so independently of the persistence ofAg (37, 43, 44). Although these plasma cells may be transferred torecipient mice by BMT, the generated Ab levels are, however, low,suggesting that long-lived plasma cells either fail to home to theproper compartment or need signals for their activity only pro-vided by the immune host. Other murine studies fail to detectlong-lived plasma cells (45). In these studies, continued productionof Ab depends on the persistence of Ag (and T cell help), andinvolves continued recruitment of new plasma cells by activationand differentiation of memory B cells. The basis for these discrep-ancies is not clear, but it may be related to differences in the Agsused (replicating vs nonreplicating Ags).

The lifetime of a human bone marrow plasma cell is yet un-known. Unlike the mouse, humans depend on the production ofAbs for decades, and it is possible that recruitment of new plasmacells from the memory compartment or even de novo may beneeded to obtain lifelong protection by Abs. After birth, an in-creasing number of Ag-experienced, somatically hypermutated

FIGURE 4. Sequences from clone A. Entire variable domain (codons 1–113) of 20 clonally related vaccine-induced HibCP-specific cDNA sequencesfrom clone A (joints 1.1 and 1.3). Dots mean identity with the VH3-23 germline sequence given above. Capital and lowercase letters indicate amino acidreplacing and silent substitutions, respectively. N indicates nucleotides probably added by N addition. Sequence name, isotype, and origin from the donor(Do) or from the recipient after the first (Rec1°) or second vaccination (Rec2°) are indicated on the left.

FIGURE 5. Cluster analysis of noncanonical 3-49RB sequences. Den-drogram demonstrating a clonal relationship between 11 of the 14 vaccine-induced HibCP-specific 3-49RB sequences in clone P (shaded area). Thegermline 3-49RB sequence (codon 1–94) was used as root. The three dif-ferent joints among the sequences are indicated as follows: F, joint 3(3-49RB; D1-1; JH3B1); �, joint 4 (3-49RB; D3-10; JH6B1); E, joint 5(3-49RB; D3-3; JH4B1). The validity of the clone assignment was assured bythe preferential allocation of sequences with the same joint to the clone andtested statistically against random allocation (see Materials and Methods).

3313The Journal of Immunology

FIGURE 6. Legend continues

3314 HUMAN MEMORY B CELLS TRANSFERRED BY ALLOGENIC BMT

IgM�IgD�CD10� memory B cells are found in the human bonemarrow (46). Isotype-switched memory B cells are also found inthe marrow, but in lower numbers. It is possible that memory cells

constitute a source of plasma cells, and it has been proposed thatthey divide in a self-renewing (stem cell-like) way, allowing con-tinued production of plasma cells without exhausting the memorypool (47). Recent data from Bernasconi et al. (48) support thishypothesis by showing that human memory B cells respond topolyclonal stimuli by differentiating into plasma cells in vitro. Thissets focus on the memory B cells present in the bone marrow as anindirect source of long-term Ab production in humans.

Memory B cells are resting but Ag-experienced B cells (49)characterized by the expression of CD27 and the presence of so-matic hypermutations in the rearranged genes (VDJ and VJ) en-coding the H and the L chains of their Ag receptors, respectively.They constitute 40% of circulating B cells in adults and may bedivided into three populations by expression of surface Ig:IgM�IgD� (15%), IgM�IgD� (10%), and isotype-switchedIgM�IgD� (15%) (50). Mutations are abundant and acquired bylargely stochastic processes (51–53). The primary site of somatichypermutation is in the germinal center in which memory cells aregenerated in response to thymus-dependent Ags (54). Other sitesof mutation probably exist, because memory B cells may also begenerated in knockout mice lacking germinal centers (55). Humanslacking germinal centers due to CD40 ligand deficiency also pro-duce somatically mutated B cells (56). Mutations are inherited bythe cellular offspring and therefore may be used as clonal markerssuitable for tracking of memory cells and their progeny in a BMTcontext. However, clonal tracking is only technically feasible if theB cell response is oligo- or monoclonal.

In this study, we exploit the well-characterized and geneticallyrestricted Ab response to the capsular polysaccharide from the im-portant childhood pathogen Hib (31, 57, 58) to perform clonaltracking of memory B cells. We clearly demonstrate that afterrecent immunization of the donor—at a time point where numer-ous vaccine-specific plasmablasts are present in the marrow—thegraft contains Ag-experienced and somatically hypermutated Bcells (i.e., memory B cells) capable of persisting in the recipient fora prolonged time (at least 9 mo). This persistence was not accom-panied by significant production of Ab in the recipient, because Ablevels were low at 6 mo after BMT. Nevertheless, memory B cellsclearly resided in the body in positions and numbers allowing themto constitute a major part of the responding population at subse-quent immunizations. In fact, 61% of VH3 family transcriptscloned after the first post-MT immunization of the recipient rep-resented B cell clones present in the transplant. For the remaining39%, it is unknown whether they originated from donor memorycells, but some of them may have done so, because minor clonesfor statistical reasons may have been overlooked among the donor-derived sequences. These proportions are probably representativefor the overall response, because virtually all Abs to HibCP useVH3 genes in normal individuals (31, 34, 57, 59). However, itcannot be excluded that HibCP-specific B cells induced after BMTmight differ by including B cells using VH genes from familiesother than VH3. Whereas about one-half of the general populationof B cells in normal adults uses VH genes from the VH3 family (51,60), this family is less represented shortly after BMT (61, 62).However, the use of VH3 genes catches up and reaches the normallevel in the first year after BMT (61, 63). It is therefore unlikelythat the recipient lacked sufficient naive cells using VH3 genes atthe time of vaccination (9 and 11 mo post-BMT), and indeed Fig.

FIGURE 6. Dendrograms of 16 identified clonal progenies (clones A–P) showing the clonal evolution, isotypes, and relation to vaccination episodes.The number of mutations in each sequence (codons 1–113) is indicated at the bottom of the figure. Sequences from the donor are named LIBP (E, �);from the first immunization of the recipient, Rec1°, are named ANBP (U, u); and from Rec2° are named AN2P (F, f). The next letter indicates the isotype:M, IgM; G, IgG; and A, IgA. In addition, every sequence is given a particular number. E, U, F, IgM sequences. �, u, f, IgA and IgG sequences.

FIGURE 7. CDR3 length spectra before and after BMT. The appliedVH3-23 repertoire before (a) and 7 mo to 4 years after transplantation (b–f).The individual PCR fragments in each reaction were measured according tolength (horizontal scale; base pairs) and amount (vertical scale; arbitrary units).The length of the CDR3 region corresponding to each peak is illustrated at thetop. The expressed IgG repertoire was almost as diverse after transplantationas before. f, The bell-shaped polyclonal appearance of the IgM VH3-23 rep-ertoire is shown. The IgM spectrum is aligned to the corresponding CDR3length peaks of the IgG spectra to account for different placement of primers.g and h, The repertoire of Ag-purified HibCP-specific cells from donor andrecipient, respectively, consisting almost exclusively of canonical length rear-rangements, is shown. Rearrangements of this length constituted only a minorfraction of the total repertoire of the recipient at other time points.

3315The Journal of Immunology

7 documents polyclonal involvement of B cells using the canonicalVH3-23 gene at all time points before and after BMT.

Interestingly, the donor-derived cells retained or reacquired thememory phenotype after the first post-BMT immunization of therecipient, because 66% of the mRNA cloned after the second post-BMT vaccination could be tracked to the donor. It is noteworthythat even the apparent repertoire shift represented by the appear-ance of 3-49RB-derived clones after the second posttransplant im-munization (Table IV) could be tracked to the donor and mustrepresent a selective expansion of the progeny of memory B cellspresent in the donor before transplantation. The clonal relationshipbetween these sequences was unequivocally determined by thesharing of 12 nucleotide substitutions in the VH region and aunique 17-bp N addition between the VH and the DH gene. It isinteresting that the donor-derived sequence was isotype switchedto the most downstream CH gene (C�2), from which no furtherisotype switching is possible, whereas the recipient-derived se-quences comprised unswitched sequences and sequences switchedto upstream CH genes (C�1 and C�1). This clearly shows that atleast some of the transferred memory B cells must have been un-switched, and that some of them refrained from isotype switchingeven after two rounds of antigenic stimulation. The same patternwas seen in five of the canonical clones (clones A, D, G, K, M, andP; Table IV). This is compatible with the idea of a stem cell-likephenotype of memory B cells allowing them to stay undifferenti-ated (and unswitched) and yet generate isotype-switched offspring(47). The IgM�IgD�CD10� memory B cells could be candidatesfor such a function, because they are preponderant in the humanbone marrow (46) and retain the capacity to switch isotype unlikeIgM-only and isotype-switched memory B cells, as demonstratedin vitro (64).

Another question is whether hypermutated B cell progeny mayre-enter the process of somatic hypermutation when rechallengedwith Ag. The fact that B cells are sometimes found with mutationprevalences exceeding by far the normal level of �5% mutatednucleotides in the VDJ region could point to such a possibility(65), but so far, no direct evidence for this has been given. Thefinding in mice that the number of mutations (and affinity) some-times increases after reimmunization (66) is not found in otherexperimental systems (67) and could alternatively be explained byselection of rare variants with many mutations generated duringthe primary immune response. In adoptive transfer experiments inmice, continued mutation of memory B cells was not seen (68, 69),and the present study finds the same in this donor-recipient pair.Thus, mutation prevalences were �5% among cloned transcriptsfrom Ag-purified isotype-switched HibCP-specific cells in the do-nor, which is close to the general average of transcripts from theperipheral blood (51), and these values did not increase signifi-cantly after the first or even the second immunization of the re-cipient (Fig. 6). Also, when the genealogical tree structures for theindividual clones were examined, they were found to be compat-ible with expansion without mutation in the recipient, and in someclones, individual donor-derived sequences contained as many mu-tations as the most mutated sequences derived from the recipient.This indicates that memory B cells transferred by BMT are unableto increase their affinity or to participate in other immune re-sponses where further somatic hypermutation would be required.This could contribute to the humoral immunodeficiency followingBMT. However, it should be noted that the experimental designmay have overlooked a low level of somatic hypermutation in therecipient for statistical reasons. Moreover, RT-PCR favors detec-tion of circulating cells rich in mRNA like plasmablasts and Ab-forming cells, whereas resting memory cells may have been over-looked either because of less mRNA content or because they were

not circulating on postvaccination day 7 where blood samples weretaken. However, memory cells mutated after the first immunizationof the recipient should reveal themselves by seeding hypermutatedprogenies into the circulation as Ab-forming cells after the secondimmunization which was apparently not the case. RT and Taqerrors were calculated to one mutation per 1200 bp, which corre-sponds to 0.2 per VDJ sequence (�4% of all mutations). Becausethese errors affect donor and recipient sequences equally, this hasno consequences for the conclusions concerning the reinduction ofmutations in the recipient. Because these errors for statistical rea-sons are unique for each affected sequence, they will not affect thedefinition of clones, which is based on shared mutations irrespec-tive of unshared ones. For the same reason, the RT-PCR errors willonly reveal themselves terminally in the tree where the terminalbranch will increase by one mutation in approximately one in fivebranches. This has no influence on the conclusions in this study.

However, failure to mutate after BMT does not prove that mem-ory B cells under normal conditions are unable to re-enter theprocess of somatic hypermutation. The failure could relate to theBMT procedure itself. Thus, it is possible that the ability to re-enter somatic hypermutation is confined to a subpopulation of cellsnot present in the graft or restricted to compartments that adop-tively transferred cells do not reach. Another possibility is that theBMT procedure induces an intrinsic deficiency of somatic hyper-mutation in B cells in general as suggested by Suzuki et al. (63)and recently by in vitro studies by Glas et al. (70). Finally, thefailure to mutate in the present study could be related to the lackof CD18, which may not be fully reconstituted by BMT in allcompartments of the immune system in this recipient. However,we find this unlikely, because somatic hypermutation was demon-strated in the recipient before BMT showing that CD18 is notobligatory for this process (data not shown).

This study adds evidence to that provided by other groups (12,13, 18, 19) indicating that BMT recipients may benefit from im-munization of their donors, but optimal immunization schedulesconcerning both the donor and the recipient must be studied in thecontext of the relevant vaccine. It is unknown whether immuniza-tion before marrow harvest is mandatory for transfer of memorycells. The close time relationship between the immunization of thedonor and the BMT in our study could have facilitated the efficienttransfer of HibCP-specific memory B cells, because one-third ofthe IgSC in the bone marrow were HibCP specific at that timepoint. This might raise the concern that recent immunization of thedonor could skew the repertoire of the recipient to a similar extent.However, this was not the case in this study, in which we findbroadly distributed CDR3 length spectra of which the canonical(HibCP-specific) rearrangement constituted only a few percent—atmost—of the VH3-23 repertoire.

It is concluded that, after recent immunization of the donor,memory B cells may be transferred to the recipient, stay there forat least 9 mo, and constitute the major source of cells respondingto vaccination of the recipient without perturbing the general rep-ertoire of the recipient. At least some of the transferred cells arenot isotype switched (but Ag experienced and hypermutated) andmay stay unswitched despite antigenic stimulation in the recipient.Transferred memory acquired little if any somatic hypermutationsin this setting.

References1. Witherspoon, R. P., R. Storb, H. D. Ochs, N. Flournoy, K. J. Kopecky,

K. M. Sullivan, H. J. Deeg, R. Sosa, D. R. Noel, K. Atkinson, and E. D. Thomas.1981. Recovery of antibody production in human allogeneic marrow graft recip-ients: influence of time posttransplantation, the presence or absence of chronicgraft-versus-host disease, and antithymocyte globulin treatment. Blood 58:360.

3316 HUMAN MEMORY B CELLS TRANSFERRED BY ALLOGENIC BMT

2. Lum, L. G. 1990. Immune recovery after bone marrow transplantation. Hematol.Oncol. Clin. North Am. 4:659.

3. Storek, J., and A. Saxon. 1992. Reconstitution of B cell immunity following bonemarrow transplantation. Bone Marrow Transplant. 9:395.

4. Haddad, E., F. L. Deist, P. Aucouturier, M. Cavazzana-Calvo, S. Blanche,G. D. S. Basile, and A. Fischer. 1999. Long-term chimerism and B-cell functionafter bone marrow transplantation in patients with severe combined immunode-ficiency with B cells: a single-center study of 22 patients. Blood 94:2923.

5. Wahren, B., G. Gahrton, A. Linde, P. Ljungman, B. Lonnqvist, O. Ringden, andV.-A. Sundqvist. 1984. Transfer and persistence of viral antibody-producing cellsin bone marrow transplantation. J. Infect. Dis. 150:358.

6. Ljungman, P., M. Wiklund-Hammarsten, V. Duraj, L. Hammarstrom, B. Lon-nqvist, T. Paulin, O. Ringden, M. S. Pepe, and G. Gahrton. 1990. Response totetanus toxoid immunization after allogeneic bone marrow transplantation. J. In-fect. Dis. 162:496.

7. Ljungman, P., V. Duraj, and L. Magnius. 1991. Response to immunizationagainst polio after allogeneic marrow transplantation. Bone Marrow Transplant.7:89.

8. Ljungman, P., I. Lewensohn-Fuchs, V. Hammarstrom, J. Aschan, L. Brandt,P. Bolme, B. Lonnqvist, N. Johanson, O. Ringden, and G. Gahrton. 1994. Long-term immunity to measles, mumps, and rubella after allogeneic bone marrowtransplantation. Blood 84:657.

9. Parkkali, T., T. Ruutu, M. Stenvik, T. Kuronen, H. Kayhty, T. Hovi, R.-M.Olander, L. Volin, and P. Ruutu. 1996. Loss of protective immunity to polio,diphtheria and Haemophilus influenzae type b after allogeneic bone marrowtransplantation. APMIS 104:383.

10. Parkman, R., and K. I. Weinberg. 1997. Immunological reconstitution followingbone marrow transplantation. Immunol. Rev. 157:73.

11. Parkkali, T., H. Kayhty, T. Ruutu, L. Volin, J. Eskola, and P. Ruutu. 1996. Acomparison of early and late vaccination with Haemophilus infuenzae type bconjugate and pneumococcal polysaccharide vaccines after allogeneic BMT.Bone Marrow Transplant. 18:961.

12. Molrine, D. C., E. C. Guinan, J. H. Antin, S. K. Parsons, H. J. Weinstein,C. Wheeler, C. McGarigle, P. Blanding, N. R. Phillips, K. Kinsella, et al. 1996.Donor immunization with Haemophilus influenzae type b (HIB)-conjugate vac-cine in allogeneic bone marrow transplantation. Blood 87:3012.

13. Saxon, A., R. Mitsuyasu, R. Stevens, R. E. Champlin, H. Kimata, and R. P. Gale.1986. Designed transfer of specific immune responses with bone marrow trans-plantation. J. Clin. Invest. 78:959.

14. Wimperis, J. Z., M. K. Brenner, H. G. Prentice, E. J. Thompson, andA. V. Hoffbrand. 1987. B cell development and regulation after T cell-depletedmarrow transplantation. J. Immunol. 138:2445.

15. Wimperis, J. Z., D. Gottlieb, A. S. Duncombe, H. E. Heslop, H. G. Prentice, andM. K. Brenner. 1990. Requirements for the adoptive transfer of antibody re-sponses to a priming antigen in man. J. Immunol. 144:541.

16. Labadie, J., M. J. D. van Tol, N. H. Dijkstra, M. van der Kaaden,C. M. Jol-van der Zijde, G. G. de Lange, F. E. Zwaan, and J. M. Vossen. 1992.Transfer of specific immunity from donor to recipient of an allogeneic bonemarrow graft: evidence for donor origin of the antibody producing cells.Br. J. Haematol. 82:437.

17. Lum, L. G. 1990. Recapitulation of immune ontogeny: a vital component for thesuccess of bone marrow transplantation. In Bone Marrow Transplantation.R. Champlin, ed. Kluwer Academic Publishers, Boston, p. 27.

18. Labadie, J., M. J. D. van Tol, N. H. Dijkstra, F. E. Zwaan, and J. M. Vossen.1992. Transfer of specific immunity from donor to recipient of an allogeneic bonemarrow graft: effect of conditioning on the specific immune response of the graftrecipient. Br. J. Haematol. 80:381.

19. Ilan, Y., A. Nagler, R. Adler, E. Naparstek, R. Or, S. Slavin, C. Brautbar, andD. Shouval. 1993. Adoptive transfer of immunity to hepatitis B virus after Tcell-depleted allogeneic bone marrow transplantation. Hepatology 18:246.

20. Etzioni, A., and J. M. Harlan. 1999. Cell adhesion and leukocyte adhesion de-fects. In Primary Immunodeficiency Diseases: A Molecular and Genetic Ap-proach. H. D. Ochs, C. I. E. Smith, and J. M. Puck, eds. Oxford Univ. Press, NewYork, p. 375.

21. Barington, T., L. Hougs, L. Juul, H. O. Madsen, L. P. Ryder, C. Heilmann, andA. Svejgaard. 1996. The progeny of a single virgin B cell predominates thehuman recall B cell response to the capsular polysaccharide of Haemophilusinfluenzae type b. J. Immunol. 157:4016.

22. Barington, T., and C. Heilmann. 1992. An improved haemolytic plaque assay forthe detection of cells secreting antibody to bacterial antigens. J. Immunol. Meth-ods 146:129.

23. Chomczynski, P., and N. Sacchi. 1987. Single-step method of RNA isolation byacid guanidinium thiocyanate-phenol-chloroform extraction. Anal. Biochem. 162:156.

24. Kabat, E. A. 1991. Sequences of Proteins of Immunological Interest. NationalInstitute of Health, Bethesda, MD.

25. Mattila, P. S., J. Schugk, H. Wu, and O. Makela. 1995. Extensive allelic sequencevariation in the J region of the human immunoglobulin heavy chain gene locus.Eur. J. Immunol. 25:2578.

26. Corbett, S. J., I. M. Tomlinson, E. L. L. Sonnhammer, D. Buck, and G. Winter.1997. Sequence of the human immunoglobulin diversity (D) segment locus: asystematic analysis provides no evidence for the use of DIR segments, invertedD segments, “minor” D segments or D-D recombination. J. Mol. Biol. 270:587.

27. Tindall, K. R., and T. A. Kunkel. 1988. Fidelity of DNA synthesis by the Ther-mus aquaticus DNA polymerase. Biochemistry 27:6008.

28. Miller, S. A., D. D. Dykes, and H. F. Polesky. 1988. A simple salting out pro-cedure for extracting DNA from human nucleated cells. Nucleic Acids Res.16:1215.

29. Van de Peer, Y., and R. De Wachter. 1994. TREECON for Windows: a softwarepackage for the construction and drawing of evolutionary trees for the MicrosoftWindows environment. Comput. Appl. Biosci. 10:569.

30. Hougs, L., T. Barington, H. O. Madsen, L. P. Ryder, and A. Svejgaard. 1993.Rapid analysis of rearranged � light chain genes of circulating polysaccharide-specific B lymphocytes by means of immunomagnetic beads and the polymerasechain reaction. Exp. Clin. Immunogenet. 10:141.

31. Pinchuk, G. V., C. Nottenburg, and E. C. B. Milner. 1995. Predominant V-regiongene configurations in the human antibody response to Haemophilus influenzaecapsule polysaccharide. Scand. J. Immunol. 41:324.

32. Hougs, L., L. Juul, H. J. Ditzel, C. Heilmann, A. Svejgaard, and T. Barington.1999. The first dose of a Haemophilus influenzae type b conjugate vaccine re-activates memory B cells: evidence for extensive clonal selection, intraclonalaffinity maturation, and multiple isotype switches to IgA2. J. Immunol. 162:224.

33. Lucas, A. H., R. J. Langley, D. M. Granoff, M. H. Nahm, M. Y. Kitamura, andM. G. Scott. 1991. An idiotypic marker associated with a germ-line encoded �light chain variable region that predominates the vaccine-induced human anti-body response to the Haemophilus influenzae b polysaccharide. J. Clin. Invest.88:1811.

34. Lucas, A. H., J. W. Larrick, and D. C. Reason. 1994. Variable region sequencesof a protective human monoclonal antibody specific for the Haemophilus influ-enzae type b capsular polysaccharide. Infect. Immun. 62:3873.

35. Adderson, E. E., P. G. Shackelford, A. Quinn, P. M. Wilson, M. W. Cunningham,R. A. Insel, and W. L. Carroll. 1993. Restricted immunoglobulin VH usage andVDJ combinations in the human response to Haemophilus influenzae type b cap-sular polysaccharide. J. Clin. Invest. 91:2734.

36. Nunez, C., N. Nishimoto, G. L. Gartland, L. G. Billips, P. D. Burrows,H. Kubagawa, and M. D. Cooper. 1996. B cells are generated throughout life inhumans. J. Immunol. 156:866.

37. Slifka, M. K., M. Matloubian, and R. Ahmed. 1995. Bone marrow is a major siteof long-term antibody production after acute viral infection. J. Virol. 69:1895.

38. Paul, J. R., J. T. Riordan, and J. L. Melnick. 1951. Antibodies to three differentantigenic types of poliomyelitis virus in sera from North Alaskan Eskimos.Am. J. Hyg. 54:275.

39. Sawyer, W. A. 1931. The persistence of yellow fever immunity. J. Prev. Med.5:413.

40. Black, F. L., and L. Rosen. 1962. Patterns of measles antibodies in residents ofTahiti and their stability in the absence of re-exposure. J. Immunol. 88:725.

41. Panum, P. L. 1849. Beobachtungen uber das Maserncontagium. Virchows Arch.1:492.

42. Ho, F., J. E. Lortan, I. C. M. MacLennan, and M. Khan. 1986. Distinct short-livedand long-lived antibody producing cell populations. Eur. J. Immunol. 16:1297.

43. Manz, R. A., A. Thiel, and A. Radbruch. 1997. Lifetime of plasma cells in thebone marrow. Nature 388:133.

44. Manz, R. A., M. Lohning, G. Cassese, A. Thiel, and A. Radbruch. 1998. Survivalof long-lived plasma cells is independent of antigen. Int. Immunol. 10:1703.

45. Ochsenbein, A. F., D. D. Pinschewer, S. Sierro, E. Horvath, H. Hengartner, andR. M. Zinkernagel. 2000. Protective long-term antibody memory by antigen-driven and T help-dependent differentiation of long-lived memory B cells toshort-lived plasma cells independent of secondary lymphoid organs. Proc. Natl.Acad. Sci. USA 97:13263.

46. Paramithiotis, E., and M. D. Cooper. 1997. Memory B lymphocytes migrate tobone marrow in humans. Proc. Natl. Acad. Sci. USA 94:208.

47. Fearon, D. T., P. Manders, and S. D. Wagner. 2001. Arrested differentiation, theself-renewing memory lymphocyte, and vaccination. Science 293:248.

48. Bernasconi, N., E. Traggiai, and A. Lanzavecchia. 2002. Maintenance of sero-logical memory by polyclonal activation of human memory B cells. Science298:2199.

49. Schittek, B., and K. Rajewsky. 1990. Maintenance of B-cell memory by long-lived cells generated from proliferating precursors. Nature 346:749.

50. Klein, U., K. Rajewsky, and R. Kuppers. 1998. Human immunoglobulin(Ig)M�IgD� periperal blood B cells expressing the CD27 cell surface antigencarry somatically mutated variable region genes: CD27 as a general marker forsomatically mutated (memory) B cells. J. Exp. Med. 188:1679.

51. Brezinschek, H. P., R. I. Brezinschek, and P. E. Lipsky. 1995. Analysis of theheavy chain repertoire of human peripheral B cells using single-cell polymerasechain reaction. J. Immunol. 155:190.

52. Foster, S. J., T. Dorner, and P. E. Lipsky. 1999. Somatic hypermutation of V�J�rearrangements: targeting of RGYW motifs on both DNA strands and preferentialselection of mutated codons within RGYW motifs. Eur. J. Immunol. 29:4011.

53. Juul, L., V. Andersen, A. Svejgaard, and T. Barington. 1997. The normally ex-pressed � immunoglobulin light chain gene repertoire and somatic mutationsstudied by single-sided specific polymerase chain reaction (PCR); frequent oc-currence of features often assigned to autoimmunity. Clin. Exp. Immunol. 109:194.

54. Jacob, J., G. Kelsoe, K. Rajewsky, and U. Weiss. 1991. Intraclonal generation ofantibody mutants in germinal centres. Nature 354:389.

55. Karrer, U., C. Lopez-Macias, A. Oxenius, B. Odermatt, M. F. Bachmann,U. Kalinke, H. Bluethmann, H. Hengartner, and R. M. Zinkernagel. 2000. An-tiviral B cell memory in the absence of mature follicular dendritic cell networksand classical germinal centres in TNFR1�/� mice. J. Immunol. 164:768.

56. Chu, Y.-W., E. Marin, R. Fuleihan, N. Ramesh, F. S. Rosen, R. S. Geha, andR. A. Insel. 1995. Somatic mutation of human immunoglobulin V genes in theX-linked HyperIgM syndrome. J. Clin. Invest. 95:1389.

3317The Journal of Immunology

57. Scott, M. G., J. J. Tarrand, D. L. Crimmins, D. W. McCourt, N. R. Siegel,C. E. Smith, and M. H. Nahm. 1989. Clonal characterization of the human IgGantibody repertoire to Haemophilus influenzae type b polysaccharide. J. Immunol.143:293.

58. Adderson, E. E., P. G. Shackelford, A. Quinn, P. M. Wilson, and W. L. Carroll.1993. Diversity of immunoglobulin light chain usage in the human immune re-sponse to Haemophilus influenzae type b capsular polysaccharide. Pediatr. Res.33:307.

59. Adderson, E. E., P. G. Shackelford, A. Quinn, and W. L. Carroll. 1991. RestrictedIg H chain V gene usage in the human antibody response to Haemophilus influ-enzae type b capsular polysaccharide. J. Immunol. 147:1667.

60. Suzuki, I., L. Pfister, A. Glas, C. Nottenburg, and E. C. B. Milner. 1995. Rep-resentation of rearranged VH gene segments in the human adult antibody reper-toire. J. Immunol. 154:3902.

61. Fumoux, F., V. Guigou, D. Blaise, D. Maraninchi, M. Fougereau, and C. Schiff.1993. Reconstitution of human immunoglobulin VH repertoire after bone marrowtransplantation mimics B-cell ontogeny. Blood 81:3153.

62. Storek, J., L. King, S. Ferrara, D. Marcelo, A. Saxon, and J. Braun. 1994. Abun-dance of a restricted fetal B cell repertoire in marrow transplant recipients. BoneMarrow Transplant. 14:783.

63. Suzuki, I., E. C. B. Milner, A. M. Glas, W. O. Hufnagle, S. P. Rao, L. Pfister, andC. Nottenburg. 1996. Immunoglobulin heavy chain variable region gene usage in

bone marrow transplant recipients: lack of somatic mutation indicates a matura-tional arrest. Blood 87:1873.

64. Werner-Favre, C., F. Bovia, P. Schneider, N. Holler, M. Barnet, V. Kindler,J. Tschopp, and R. H. Zubler. 2001. IgG subclass switch capacity is low inswitched and IgM-only, but high in IgD�IgM�, post-germinal center (CD27�)human B cells. Eur. J. Immunol. 31:243.

65. Fischer, M., and R. Kuppers. 1998. Human IgA- and IgM-secreting intestinalplasma cells carry heavily mutated VH region genes. Eur. J. Immunol. 28:2971.

66. Berek, C., J. M. Jarvis, and C. Milstein. 1987. Activation of memory and virginB cell clones in hyperimmune animals. Eur. J. Immunol. 17:1121.

67. Fish, S., E. Zenowich, M. Flemming, and T. Manser. 1989. Molecular analysis oforiginal antigenic sin. I. Clonal selection, somatic mutation, and isotype switch-ing during a memory B cell response. J. Exp. Med. 170:1191.

68. Dell, C. L., Y. Lu, and C. L. Claflin. 1989. Molecular analysis of clonal stabilityand longevity in B cell memory. J. Immunol. 143:3364.

69. Siekevitz, M., C. Kocks, K. Rajewsky, and R. Dildrop. 1987. Analysis of somaticmutation and class switching in naive and memory B cells generating adoptiveprimary and secondary responses. Cell 48:757.

70. Glas, A. M., E. H. N. van Montfort, J. Storek, E. G. N. Green, R. P. M. Drissen,V. J. Bechtold, J. Z. Reilly, M. A. Dawson, and E. C. B. Milner. 2000. B-cell-autonomous somatic mutation deficit following bone marrow transplant. Blood96:1064.

3318 HUMAN MEMORY B CELLS TRANSFERRED BY ALLOGENIC BMT