Germinal centers (GCs) promote humoral immunity and vaccine efficacy. In GCs, antigen
(Ag)-activated B lymphocytes proliferate, are selected for high affinity antibody
(Ab), promote Ab class switching, and yield B cell memory
1, 2
. Whereas the cytokine milieu has long been known to regulate effector functions that
include the choice of immunoglobulin class
3, 4
, both cell-autonomous
5
and extrinsic
6, 7
metabolic programming have emerged as modulators of T cell-mediated immunity
8
. We now show that GC light zones are hypoxic and low oxygen (pO2) alters B cell physiology
and function. In addition to reduced proliferation and increased B cell death, low
pO2 impaired Ab class switching to the pro-inflammatory IgG2c Ab isotype by limiting
expression of the cytosine deaminase AID. Hypoxia induces HIF transcription factors
by restricting activity of prolyl hydroxyl dioxygenases (PHD), enzymes that hydroxylate
HIF-1α and HIF-2α to destabilize HIF through binding of Von Hippel-Landau protein
(pVHL)
7
. B cell-specific pVHL depletion led to constitutive HIF stabilization, decreased
Ag-specific GC B cells and undermined the generation of high-affinity IgG, switching
to IgG2c, early memory B cells, and recall Ab responses. HIF induction can reprogram
metabolic and growth factor gene expression. Sustained hypoxia or HIF induction via
pVHL deficiency inhibited mTOR complex 1 (mTORC1) activity in B lymphoblasts, and
mTORC1 haploinsufficient B cells had reduced clonal expansion, AID expression, and
capacities to yield IgG2c and high-affinity Ab. Thus, the normal physiology of germinal
centers involves regional variegation of hypoxia, and HIF-dependent oxygen sensing
regulates vital functions of B cells. We propose that restriction of oxygen in lymphoid
organs, which can be altered in pathophysiological states, modulates humoral immunity.
The micro-anatomy of secondary lymphoid organs and rapid proliferation of activated
lymphocytes in them
9
prompted testing for hypoxia. Using flow cytometry, HIF levels were found to be increased
in GC-phenotype B (GCB) cells compared to other B lymphocytes in spleens of immunized
mice (Fig. 1a; Extended Fig. 1a). Immune fluorescent microscopy revealed that HIF
was most elevated in germinal centers (Fig. 1b; Extended Fig. 1b). Low oxygen induces
HIF. However, HIF α subunits can be stabilized at normoxic pO2
10
, so we used chemical probes to mark hypoxic cells in vivo. Spleen, lymph nodes, and
Peyer’s Patches were analyzed after injection of pimonidazole or EF5
11
and staining with Ab that bind the adducts (Fig.1c–e; Extended Data Fig. 1b–h). Fluorescence
denoting hypoxia localized predominantly to the GC and the signal for each agent was
weaker in the IgD+ zone
1
. Flow cytometry detected EF5 only with GL7+ GCB cells (Fig. 1e), and a hypoxia-related
gene signature was enriched in GC B cells (Extended Data Fig. 1i). The EF5 and pimonidazole
signals only partially filled GC, which are subdivided into light and dark zones between
which B cells cycle iteratively to promote high-affinity Ab. EF5 labeling predominantly
overlapped a follicular dendritic cell marker (CD35) restricted to the light zone
(Fig. 1f). B lymphoblasts proliferate rapidly in the dark zone, whereas cell cycling
decreases in the light zone
1
. The most EF5-positive GCB cells had entered S-phase at lower rates (% BrdU+) (Fig.
1g, h) and more frequently activated an executioner caspase (Fig. 1i). Thus, activated
B cells experience hypoxia in GC, predominantly in their light zones. Strikingly,
the more hypoxic GCB cells proliferated less and had increased apoptotic signaling.
To test what impact hypoxia has on Ab class switching, activated B cells cultured
in hypoxia (pO2 of 1%) were compared to controls cultured at atmospheric (∼21%) or
venous (5%) pO2, using conditions that promote IgG1 or the pro-inflammatory isotype
IgG2c (Fig. 2a; Extended Data Fig. 2). Hypoxia restricted B cell population growth
(Fig. 2a, b), with increased caspase-3 activation and lower BrdU incorporation (Extended
Data Fig. 2a, b). Thus, O2 sufficiency promoted B cell proliferation by both improving
survival and increasing cell cycling. These effects were paralleled by an altered
balance in cell metabolism, as hypoxia promoted a higher glycolytic rate (Extended
Data Fig. 2c) in activated B cells. Conversely, PHD inhibition in vitro reduced O2
consumption, and gene expression profiling of fresh ex vivo B cells showed major differences
between non-GC and GC subsets (Extended Data Fig. 2d, e, respectively). Moreover,
IgG+ B cell frequencies were reduced at 1% pO2 (Fig. 2a; Extended Data Fig. 2f). The
enteric immune system is a site of physiological hypoxia
12
; notably, hypoxia did not decrease the frequency of IgA+ B cells in IgA-promoting
conditions (Fig. 2a; Extended Data Fig. 2f). Switching requires multiple B cell divisions
13
. When fluorescein partitioning was analyzed along with switching to IgG2c, hypoxia
reduced switching by B cells at the same division number (Fig. 2b). Thus, hypoxia
at levels of the GC light zone altered Ab class switching by a direct influence on
class choice in addition to reducing proliferation and reprogramming B cell metabolism
and survival.
CSR is executed by an activation-induced deaminase (AID), encoded by the Aicda gene
1, 3, 4
. In IgG switch conditions, Aicda mRNA and AID protein were reduced by hypoxia (Fig.
2c, d; Extended Data Fig. 2g). In contrast, AID was not reduced by hypoxia in IgA
switch conditions (Fig. 2d). Switch recombinase is directed to Ig heavy chain regions
by transcription factors that create accessibility marked by germ line transcripts
(GLT)
3, 4
. Hypoxia decreased induction of the transcription factor T-bet and the T-bet-dependent
Iγ2c GLT
14
(Fig. 2e, f), whereas Rora mRNA and the GLT Iα were not reduced in B cells at reduced
pO2 (Fig. 2e, f). DMOG reduced proliferation and increased apoptosis of B cells cultured
at 21% pO2 and severely restricted switching to IgG2c, whereas that to IgA exhibited
less impairment (Extended Data Fig. 3a, b). An inhibitor of HIF stabilization mitigated
the reduction of IgG2c-switched B cells by low oxygen (1% pO2) (Extended Data Fig.
3c). Akin to hypoxia, PHD inhibition and HIF stabilization impaired AID, T-bet, and
Iγ2c GLT induction in the presence of the IgG2c switch cytokine IFN-γ (Fig. 2c, e,
f; Extended Data Fig. 4a–c). In contrast, RNA for RORα and the Iα GLT were higher
in DMOG-treated cells than in controls (Fig. 2c). Thus, hypoxia reduced AID and GLT
induction in the conditions promoting IgG2c whereas Iα and AID levels were maintained
in IgA conditions, consistent with relative effects on class-switched B cell antigen
receptors (BCR).
pVHL destabilizes HIF by targeting hydroxylated alpha subunits for rapid proteasomal
degradation in most oxygen-sufficient environments
7, 15
. To model persistent hypoxic signaling in vivo, we used conditional Vhl loss-of-function
experiments. Mature B cells subjected to Vhl deletion yielded less Ag-binding GCB
cells after immunization, less IgG2c Ab, and a substantial decrease in cells secreting
Ag-specific IgG2c in primary (10) responses (Fig. 3a–c; Extended Data Fig. 5, 6).
Cycling between light and dark zone in GC promotes higher affinity Ab
16
so it was striking that for IgM and IgG1 pVHL depletion only impaired generation of
high-affinity anti-NP Ab (Fig. 3b). The defect in 10 responses substantially reduced
IgG2c of all affinities (Fig. 3b; Extended Data Fig. 6a), whereas Ag-specific IgA
was unaffected (Extended Data Fig. 6b). The effects of pVHL depletion on IgG2c and
high-affinity IgG1 Ab responses were HIF-dependent (Fig. 3b). Defects of Ab responses
were heightened in recall (20) immunity when compared to 10 responses (Extended Data
Fig. 5c, d compared to Fig. 3a). pVHL loss reduced the population of Ag-binding memory-phenotype
B (Bmem) cells, an effect mitigated by concomitant HIF depletion (Fig. 3c, Extended
Data Fig. 6d). Aicda mRNA induction in activated B cells was impaired in cells with
elevated HIF due to reduced Vhl (Fig. 3d; Extended Data Fig. 4c, d). Tbx21 mRNA and
T-bet protein levels also were lower in pVHL-depleted B cells (Fig. 3d; Extended Data
Fig. 4c). To test the significance of decreased AID and T-bet, we forced expression
of these proteins in activated B cells. T-bet did not increase the frequency of IgG2c–positive
B cells during PHD inhibition, though it bypassed the need for IFN-γ with control
B cells (Fig. 3e, Extended Data Fig. 4e). In contrast, forcing AID expression normalized
switching in these assays (Fig. 3e; Extended Data Fig. 4e). We conclude that the PHD/HIF/VHL
axis regulates the qualities of Ab responses, with modulation of AID levels as a major
mechanism for hypoxic influence on the Ig class preferences.
B cell activation, CSR, and development into Ab-secreting cells are effected by receptors
that stimulate the mammalian target of rapamycin (mTOR). Hypoxia and HIF-1 have been
shown either to inhibit or enhance mTORC1 activity in tumor or endothelial cells
17, 18
. In hypoxic and DMOG-treated B cells, BCR engagement elicited less phosphorylation
of proteins downstream from mTORC1 (Fig. 4a; Extended Data Fig. 7a). Depletion of
pVHL also reduced BCR-stimulated mTORC1 by a HIF-dependent mechanism (Fig. 4b). Thus,
hypoxia restrained mTORC1 in normal B cells. In vitro experiments suggest that HIF-mediated
limitation of increased amino acid transport contributes to this effect. B cell activation
increased leucine uptake and expression of transporters used for nutrient uptake;
HIF stabilization impaired this induction (Extended Data Fig. 7b–e). Moreover, adequate
supplies of leucine were crucial, and partially sufficient, for BCR re-activation
of mTORC1 in B lymphoblasts (Extended Data Fig. 7f). HIF depletion did not completely
restore either the Ab response or amino acid uptake to normal in pVHL-deficient B
cells. However, two additional mechanisms previously shown to suppress mTORC1 were
evoked in hypoxic B cells in vitro – steady-state ATP pools were halved, accompanied
by increased AMPK activity, and expression of the Redd1 gene increased (Extended Data
Fig. 8a–c).
Disruption of mTOR function by means that impair both mTORC2 and mTORC1 altered the
balance between class-switched and IgM Ab against specific Ag
19, 20
. In contrast, HIF stabilization only partially inhibited mTORC1 and spared mTORC2
(Extended Data Fig. 8d, e). Accordingly, we tested if partially reduced mTORC1 activity
impacts high-affinity Ab production, proliferation, AID levels, or biases of Ig class
switching using disruption of Rptor, which encodes a protein essential for mTORC1
21
. Rptor haplo-insufficiency in B cells reduced mTORC1 activity (Extended Data Fig.
9a) and yielded results of in vitro switching and humoral responses in vivo (Fig.
4c, d; Extended Data Fig. 9) similar to those obtained with hypoxia and the PHD/HIF/VHL
axis. IgG2c reductions were more substantial than those of IgM or IgG1 (Fig. 4c),
and NP-specific GC B cells and IgG2c anti-NP Ab-secreting cells (ASC) (Extended Data
Fig. 9b–d) were reduced. Partial mTORC1 loss reduced switching to IgG2c (Extended
Data Fig. 10a) and suppressed high-affinity IgG1 Ab production (Fig. 4d). IgG1 switch
conditions promoted higher expression of a tracking allele, AID-GFP, which was partially
reduced by Rptor hemizygosity (Fig. 4e), whereas IgG2c conditions led to less AID
in control cells and greater reduction in Rptor +/Δ B cells. Moreover, Rptor haploinsufficiency
led to reduced T-bet, and decreased both Tbx21 and Aicda mRNA in activated B cells
(Extended Data Fig. 10b, c). Pharmacological inhibition of mTOR with rapamycin substantially
reduced AID levels (Fig. 4f)
19, 20
and switching to IgG2c, an effect mitigated by forced AID and T-bet expression (Extended
Data Fig. 10d–f). Overall, localized hypoxia and HIF induction are normal features
of germinal center micro-physiology that modulate the output from lymphoid follicles,
effects similar to those of restricting mTORC1 activity.
Low oxygen confronts B cells in germinal center during an immune response. The findings
reveal that restricted oxygen supply or persistent induction of HIF transcription
factors in B cells limits proliferation, isotype switching, and levels of high-affinity
Ab. GC B cells undergo iterative selection to enhance antibody affinity
1, 2
so that the most suitable B cells survive, further mature, and continue to multiply.
Thus, the restriction of pO2 of the GC may slow proliferation and set a more stringent
threshold for critical survival signals. In addition, the IgG2c isotype has particular
functions in anti-microbial responses and inflammation due to the affinities of its
constant region with the spectrum of Fc receptors on cells
22
. Many patients suffering from hypoxemic lung disease exhibit lower serum IgG and
heightened susceptibility to respiratory infection
23
. Hypoxia also has been recognized as a major aspect of inflammation in disease states.
Intra-tumoral restrictions of oxygenation elicit indirect effects on immune function
in cancer and may also act directly on T lymphocytes
24, 25
. Moreover, hypoxia and neo-lymphoid tissue or tertiary lymphoid structures with GC,
plasma cells, and local Ab production are now recognized in a wide range of inflammatory
settings whose oxygen landscape is unexplored
26
. The hypoxia response program in intestinal epithelial cells limits local inflammation
12, 27, 28
, providing counter-regulation against activated neutrophils
27
. Analogous to this, the susceptibility of IgG2c to hypoxia may represent another
means for limiting pathology from unchecked inflammation in normal immunity.
Methods
Mice and B cell transfer models
Mice [C57BL/6 mice, CD45.1 congenic, Ig CH allotype disparate, Rag0
, AID-GFP Tg, pVHL cKO (Vhl
fl/fl; ERT2-Cre)
29
, pVHL; HIF1α; HIF2α triple cKO (Vhlfl/fl;Hif1afl/fl; Epas1fl/fl
;ERT2-Cre), and Raptor cKO (Rptor
fl/fl;ERT2-Cre)
30
] were housed in ventilated micro-isolators under Specified Pathogen-Free conditions
in a Vanderbilt University mouse facility and used at 6–8 wk of age following approved
protocols. Healthy mice of useful genotype were randomly selected for the experiments,
without preference to size, gender, or other potential confounding factor. All Figures
are based on data reproduced in independent biological replicates, typically conducted
weeks or months apart in time and involving different cages of donor and recipient
mice, and always with parallel handling and manipulation of the mice and cells of
samples to be compared. For adoptive transfer experiments, B cells (from 1–2 donor
mice of each genotype) were purified by depleting T cells using biotinylated anti-Thy1.2
Ab followed by streptavidin-conjugated microbeads (iMag™; BD Biosciences, San Jose
CA). Pooled WT CD4+ T cells and OT-II CD4+ T cells (4 and 1 × 106 cells per recipient,
respectively, typically from two donor mice of each background) were purified by positive
selection with L3T4 anti-CD4 microbeads and, in adoptive transfers into Rag0
or Ig CH allotype-disparate (IgHa) mice, mixed with pools of WT, VhlΔ/Δ, VhlΔ/Δ;Hif1aΔ/Δ;Epas1Δ/Δ,
or RptorΔ/+
B cells (5 × 106 cells per recipient) and injected intravenously (i.v.) into Rag0
or IgHa recipients. Recipient mice of similar ages (6–8 wk) were randomly selected
for the experiments, without preference to size or gender. Experiments using the conditional
Vhl alleles (VhlΔ/Δ
) were designed to avoid distortions rapidly imposed by systemic pVHL loss [e.g.,
extra-medullary hematopoiesis
29
]. Those using Rptorf/f
+drove excision with the same Rosa26-CreERT2 allele and with tamoxifen-initiated Cre
activity so as to be more directly comparable to the Vhl experiments and because of
distortions of B cell development observed even for heterozygotes with mb1-Cre (deletion
at outset of B lymphoid ontogeny) (Raybuck, A.L. and M. Boothby, unpublished observations).
Reagents
IFN-γ, IL-4 and mAb (purified, biotinylated, or fluorophore-conjugated) were from
BD Pharmingen or Tonbo Biosciences (San Diego CA) unless otherwise indicated. IL-5
was from Peprotech (Rocky Hill NJ), TGF-β and BAFF were from R&D Systems (Minneapolis,
MN). NP-BSA (for capture ELISA), NP-OVA, and NP-O-succinimide were obtained from Biosearch
(Novato CA). SRBC (sheep red blood cells), D-glucose, and 2-deoxyglucose were from
Thermo Fisher Scientific (Waltham MA). Tamoxifen, 4-hydroxy-tamoxifen, chicken ovalbumin,
all-trans retinoic acid and LPS were from Sigma-Aldrich Chemicals (St. Louis MO).
DMOG and oligomycin were from EMD Millipore (Billerica, MA). Fluorescent proteins
APC and rPE (Prozyme, Haward CA) were used for conjugation reactions with NP-O-succinimide
to generate fluor-conjugated NP.
Immunohistochemistry and detection of hypoxia
C57BL/6 mice were immunized with SRBCs (2×108 cells per mouse). At 1 week after immunization,
mice were injected with EF5
31
or pimonidazole HCl (Hypoxyprobe™) (Hypoxyprobe, Burlington, MA). Spleen, LN, and
Peyer’s Patches were embedded in OCT reagent and snap frozen on dry ice. Sections
of frozen tissue were fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton
X-100 in PBS, blocked with M.O.M™ (Vector Lab) followed by incubation with GL7-FITC,
IgD-PE, and anti-EF5-Cy5 Abs at 4 °C. For Hypoxyprobe™ detection, frozen sections
were stained with biotinylated anti-pimonidazole Ab followed by streptavidin-conjugated
Alex647 Ab. Biotinylated anti-CD35 Ab (BD Pharmingen, clone 8C12) followed by streptavidin-conjugated
PE Ab was used for indirect immunofluorescent detection of FDC. Quantification of
HIF-1α, EF5 and Hypoxyprobe™ fluorescent intensity within GC (total or light zone
as defined by CD35 staining) and follicular regions was performed using MetaMorph
Image processing software. For both negative controls and samples, the regions were
quantified in toto using the signal-averaged fluorescence intensity within each boundary
(e.g., CD35+ or GL7+) After subtracting the background MFI from negative control samples,
MFIs of HIF-1α, EF5 and Hypoxyprobe™ within GC (CD35+ and CD35neg zones) and GL7-negative
follicular region were obtained. Data are presented as average (± SEM) MFI for the
individual samples (n>20 GC for each condition, drawing evenly on three independent
experiments). In flow cytometric detection of hypoxic cells, BrdU incorporation, or
cleaved caspase 3, cell surface markers were stained by fluor-conjugated mAb, followed
by fixation (4% paraformaldehyde), permeabilization with saponin (0.2%), and stained
with anti-EF5-Cy5, or 2-step staining of pimonidazole according to supplier’s instructions.
BrdU and cleaved caspase 3 were detected as described
32
.
Immunizations, and measurements of Ab responses
After collection of pre-immune sera, mice were immunized with NP16-ovalbumin (OVA)
(100 µg intraperitoneally) in alum (Imject, Thermo Fisher Scientific, Waltham MA)
as described
33
. Alternatively, this primary immunogen was mixed with NP-modified SRBC, followed
by a boost with NP-ovalbumin in alum. Relative levels of anti-NP Ab in immune sera
were assayed by ELISA on serial dilutions binding to either NP20-BSA (high valency,
to capture all affinities of Ab) or NP2 (low valency, to restrict binding to the high-affinity
Ab). Specific classes or isotypes were then detected using the series of isotype-specific
second Ab of the SBA Clonotyping System (Southern Biotech, Birmingham AL), as described
33
. Data for Ag-specific Ab are shown after subtraction of low OD values from pre-immune
controls analyzed together with the immune sera and were separately determined to
match values yielded by titration. Ab-secreting cells (ASC) were analyzed by ELISpot
as previously described
33
and quantitated using an ImmunoSpot Analyzer (Cellular Technology, Shaker Heights
OH). Ag-specific B cells were detected and enumerated using flow cytometry to score
B lineage-marked cells binding to fluor (APC or rPE)-conjugated NP, using a dump channel
(7-AAD and APC-conjugated mAb against IgD, F4/80, Gr1, CD11b, CD11c, CD4, and CD8)
to exclude non-specific signal.
Gene expression profiling
Mice were injected with SRBCs and sacrificed 10 days post-immunization. Single cell
suspensions from spleens were stained with α-B220 (RA3–6B2) and GL7. B220+GL7− and
B220+GL7+ splenocytes were sorted into Trizol reagent (Ambion). Total RNA was isolated
from biological replicates and provided to the Vanderbilt VANTAGE shared resource
for library construction and sequencing. Briefly, libraries were constructed from
poly-adenylated RNAs and sequenced with an Illumina HiSeq 2500 on an SR-50 run aiming
for 30M reads/sample. Reads were aligned to the mm10 mouse transcriptome using TopHat
and differential gene expression was determined using Cuffdiff as previously described
34
. Gene set enrichment analysis (GSEA) was performed using software available from
the Broad Institute (http://www.broadinstitute.org/gsea), which tested for enrichment
based on hypergeometric distribution with respect to published gene signatures. For
hypoxia regulated gene signature, GSEA plots comparing a gene set pre-ranked by log2
fold change in gene expression (GL7+ B220+ vs GL7− B220+) to a hypoxia signature published
by Eustace et al
35
. A significant enrichment was defined as having an FDR q value ≤0.05. The results
of RNA-Seq have been deposited in the NCBI Gene Expression Omnibus database under
accession code GSE77113.
In vitro B cell cultures for class-switched Ab production
Splenic B cells were purified (90–95%) by depleting T cells using biotinylated anti-Thy1.2
mAb followed by streptavidin-conjugated microbeads. For IgG1, B cells (0.5×106 cells/ml)
were activated with LPS or F(ab’)2 anti-IgM (Southern Biotechnology) and anti-CD40
(BD Pharmingen), cultured with BAFF and IL-4. For IgG2c, B cells (0.5×106 cells/ml)
were activated with LPS or anti-IgM and anti-CD40, cultured with BAFF and IFN-γ. For
IgA, B cells (0.5×106 cells/ml) were activated with LPS (1 µg/ml) or anti-IgM and
anti-CD40 and cultured with BAFF (10 ng/ml), TGF-β (5 ng/ml), IL-4 (10 ng/ml), IL-5
(10 ng/ml), and all-trans retinoic acid (RA) (10 nM) in IMDM medium supplemented with
10% FBS, Pen/Strep, L-glutamine, and β-mercaptoethanol. To analyze the partitioning
of cell division, purified B cells were stained with CellTrace™ Violet (Thermo Fisher
Scientific) according to manufacturer’s instruction or CFDA-SE as described previously
33
. Cells were cultured (4 d) at pO2 of 21%, 5% or 1% after which surface Ig was analyzed
by flow cytometry. In comparisons of all three [O2] (oxygen tensions), experiments
were performed by dividing one common pool of B cells and using two separate hypoxia
chambers maintained at constant pO2 using nitrogen.
Measurements of RNA and proteins
RNA was isolated using Trizol reagent (Invitrogen). After cDNA synthesis by reverse
transcription, expression of genes was analyzed in duplicate samples using SYBR green
PCR master-mix (Qiagen, Valencia CA) by quantitative real-time RT-PCR (qRT2-PCR).
Data are presented as values normalized to WT control and averaged over PCR normalized
to levels of internal control (actin). Primer pairs are detailed in a Table that,
along with conditions, is freely available upon request. Proteins in whole cell extracts
were separated by SDS-PAGE, transferred onto nylon membranes (Millipore), and then
incubated with rabbit antibodies against p-S6 (S235/236), p-p70S6K (S371), p-Akt (S473),
p-Akt (T308), p-ACC (S79), p-AMPK (T172) (Cell Signaling Technologies), or goat anti-Actin
(Santa Cruz) Abs followed by the appropriate fluorophore-conjugated, species-specific
secondary anti-Ig antibodies (Rockland Immunochemicals, and LI-COR). Proteins were
visualized and quantitated by laser excitation and infrared imaging (Odyssey, LI-COR).
For measurements of the induction of S6K, S6 and Akt phosphorylation, purified B cells
were cultured 2 d in BAFF (10 ng/ml) and F(ab’)2 anti-IgM (1 µg/ml), washed, rested
18 h, and then re-stimulated (15 min) in the presence or absence of F(ab’)2 anti-IgM
(2.5 µg/ml). To test the effect of amino acid supply on S6K and S6 phosphorylation,
B lymphoblasts were washed, cultured in complete medium overnight, then rinsed, cultured
in amino acid-free RPMI1640 (US Biological, Salem, MA) for 1 hr, and re-stimulated
in the presence or absence of anti-IgM, with readdition of L-leucine (Sigma) or all
20 amino acids. For the induction of p-ACC and p-AMPK, purified B cells were cultured
for 2 d in LPS, BAFF, and IFN-γ.
Glycolysis and oxygen consumption assays
Purified B cells were cultured for 2 d at 37°C at pO2 of 21% (normoxia) or pO2 of
1% (hypoxia) in the presence of BAFF, LPS, and IFN-γ. To quantitate glycolysis, 1×106
viable cells were washed, pulsed with 10 µCi of 5-[3H]-glucose in 24-well plates (37°C,
1 h), and returned to their previous oxygen condition. Glycolytic conversion was then
quantitated as described
32
. Oxygen consumption rates were measured using Seahorse assays. Because this instrument
cannot be used in a hypoxia chamber, purified B cells (1×106 cell/ml) were activated
with 1 µg/ml LPS and cultured 48 hr with 10 ng/ml BAFF in complete IMDM medium supplemented
as described
32
in the presence or absence of 0.5 mM DMOG. After 48 hr, cultured B cells were washed
twice, resuspended in XF Base Media (Seahorse Bioscience, North Billerica MA) supplemented
with 2 mM L-glutamine, and equal numbers of Trypan Blue-excluding B cells (1.5×105)
were plated on extracellular flux assay plates (Seahorse Bioscience) coated with CellTak
(Corning) according to the manufacturer’s protocol. Before extracellular flux analysis,
B cell were rested (25 minutes at 37°C, atmospheric CO2) in XF Base Media. Oxygen
Consumption Rate (OCR) and Extracellular Acidification Rate (ECAR) were measured using
a XF96 extracellular flux analyzer (Seahorse Bioscience) before and after the sequential
addition of 10 mM D-glucose, 1 µM oligomycin, and 50 mM 2-deoxyglucose.
Amino acid uptake assay
Purified B cells were activated and cultured 2 d with LPS and BAFF. Viable cells were
washed and incubated with amino acid uptake buffer (5.4 mM KCl, 140 mM NaCl, 1.8 mM
CaCl2, 0.8 mM MgSO4, 5 mM D-glucose, 25 mM HEPES, and 25 mM Tris, pH 7.5) for 30 min
to deplete intracellular amino acids. Triplicate samples (1×106 cells/sample) were
incubated with 1 µCi of L-[3, 4, 5-3H] leucine (American Radiolabeled Chemicals, Inc)
in amino acid uptake buffer for 2 min at room temperature and immediately spun through
a layer of bromododecane (200 µl) into 8% sucrose/ 20% perchloric acid (25 µl). Tubes
were frozen in a dry ice/EtOH bath and cut with dog nail clippers to separate the
cells from unincorporated [3H]-leucine. 25 ul of 10% Triton X-100 and liquid scintillation
cocktail were added and the cell-associated 3H were measured by liquid scintillation
counting.
Statistical analysis
The primary analyses were conducted on pooled data points from independent samples
and replicate experiments (minimum two, generally three, biologically and temporally
independent replicate experiments for all data, with multiple independent samples
in the case of two biological replicates), using an unpaired two-tailed Student’s
t test with post-test validation of its suitability. Welch’s or Mann-Whitney testing
were used instead of the t-test where indicated based on statistical analysis of the
distribution of variances in the samples to be compared. Data are displayed as mean
(± SEM), i.e., ‘center values’ were mean as ‘average’. Results were considered statistically
significant when the p value of for the null hypothesis of a comparison was <0.05.
Since the extent or direction of difference between samples was unknown, and regulations
mandate reducing the number of animals used to the lowest feasible level, no statistical
methods were used to determine pre-specified sample sizes. The experiments were not
randomized and the investigators were not blinded during the experiments. Corrections
for multiple comparisons were not used. Statistical approaches for RNA-Seq-related
data are outlined in that section.
Extended Data
Extended Data Figure 1
Landscape of hypoxic cells in follicles and GCs of lymphoid organs
(a, b) Controls for anti-HIF-1α Ab staining of GC and portions of the surrounding
splenic follicle, as in Fig. 1a, b, with fluorescent signals at the same intensity
settings when analyzing samples processed together, using SRBC immunization of WT
and Hif-deleted mice and either anti-HIF-1α sera or non-immune rabbit IgG (rIgG1),
as indicated. Shown are (a) flow cytometry results of intracellular staining performed
after exposure of lymphoblasts of the indicated genotypes to 4-hydroxytamoxifen and
hypoxia, and (b) confocal images (40x magnification) as in Fig. 1a, b, respectively.
(c) Quantified data obtained from samples represented in Fig. 1c. Shown are the mean
(±SEM) specific fluorescence intensities of Hypoxyprobe™ (anti-pimonidazole) staining
in germinal centers (delimited as GL7+) and GL7− IgD+ follicular B cell regions after
subtracting background signal (mean fluorescence intensities in these regions after
anti-pimonidazole staining of samples from PBS-injected control mice). (d) Immunostaining
of EF5-modified cells. Shown are confocal microscopic images of spleen sections from
SRBC-immunized mice injected with EF5 (left) or PBS (right) 2 hr before harvest, followed
by direct immunofluorescent staining of frozen sections with anti-GL7 Ab, anti-IgD
and anti-EF5, representative of the quantified data presented in Fig. 1d. (e) Representative
images of mesenteric LN after injections and immunostaining as in Fig. 1c. (f) Low
(10x) magnification image of anti-pimonidazole immunohistochemistry on spleen sections
from SRBC-immunized mice injected with pimonidazole (left) or PBS (right) prior to
harvest. Among stained sections for both anti-pimonidazole and EF5, ∼75 % of GC sections
were unequivocally positive. (g) Representative images of Peyer’s Patches from non-immune,
EF5-injected mice processed as in Fig. 1c. (h) Representative images of spleen sections
from unimmunized mice injected with Hypoxyprobe™ (left) or PBS (right) 3 hr before
harvest, processed in parallel with sections from immunized mice injected with probe,
and imaged by confocal microscopy at the same time and settings as for the sections
from immunized mice. (i) GSEA plots comparing gene set pre-ranked by log2-fold change
in relative expression (GL7+/GL7-) in a hypoxia gene signature.
Extended Data Figure 2
Altered B cell survival, proliferation and metabolism in reduced pO2
(a) Increased executioner caspase-3 activation in hypoxic B cells. Shown (left panels)
are representative flow histograms of activated (cleaved) caspase-3 (CC3) in the B
cell gate after activated B cells were cultured in pO2 of 21% (normoxia) and 1% (hypoxia).
B cells were stimulated with BAFF, LPS, and IFN-γ, cultured (4 d) at the indicated
oxygen tension and processed for detection of activated caspase-3 using fluorescent-conjugated
active caspase-3 Ab. Panel to right displays the mean (±SEM) quantitative data for
the frequencies of B cells positive for caspase-3 cleavage in three independent replicate
experiments. (b) O2 sufficiency enhances cell cycle rates. As in (a), but cells were
pulsed with BrdU and frequencies of S-phase during the cultures are displayed in relation
to IgG2c switching, with a representative result (left panels) and quantitation of
the overall B220+ cell populations in three independent replicate experiments (right
panel). B cells were cultured (4 d) with BAFF, LPS, and IFN-γ at the indicated oxygen
levels, pulsed (4 hr) with BrdU, and then stained with anti-IgG2c, -B220, and -BrdU
Ab after fixation, permeabilization, and processing. (c, d) Pools of purified WT B
cells were stimulated with BAFF and LPS, divided, and cultured (2 d) in pO2 of 21%
(normoxia) and 1% (hypoxia). (c) Rates of glycolysis were measured after return to
their previous oxygen conditions, using equal numbers of surviving B cells after culture
as detailed in the Methods. The bar graph shows mean (±SEM) rates of glycolysis measured
in three independent experiments. (d) Inhibition of PHD activity decreases cellular
respiration of B lymphoblasts. Purified B cells were activated and cultured (2 d)
with LPS and BAFF in the presence or absence of the PHD inhibitor DMOG (0.5 mM). Oxygen
Consumption Rate (OCR) was measured with cultured viable B cells (1.5×105 cells) as
detail in the Methods. The bar graph shows mean (±SD) OCR of technical triplicate
measurements in one experiment representative of three independent replicates with
similar results. (e) Metabolic gene expression profile of GL7+ GC B cells. Genes showing
significant expression changes in GL7+ GC B cells were mined for genes important for
the indicated cellular processes. The heat map depicts values for the indicated genes
shown as the value derived as log10 of the fragments per kilobase per million (reads)
after adding 1 to each value (FPKM+1). (f) Hypoxia limits switch to IgG among B cells
activated via BCR and CD40. As in Fig. 2a, except that the B cell preparations were
activated by cross-linking their surface IgM and CD40 without addition of LPS. (g)
Quantified mean fluorescence intensities for GFP expression in the full set of replicate
experiments conducted as in Fig. 2d, presented as mean (±SEM) data for each condition
of culture (pO2 of 21, 5, or 1%, with cytokines and retinoic acid for Ig class switch
conditions as indicated, and as for Fig. 2a, b).
Extended Data Figure 3
HIF stabilization alters B cell survival, proliferation and class switched Ab level
(a) Purified WT B cells were activated and cultured (4 d) with LPS and BAFF in the
presence or absence of DMOG, after which frequencies of cells with cleaved caspase
3 or BrdU uptake, as indicated, were measured as in Extended Data Fig. 2 (representative
result from one experiment among n=3 independent replicate experiments). (b) Purified
WT B cells were activated and cultured in conditions for switching to IgG1, IgG2c,
and IgA, as in Fig. 2a, b, but at atmospheric (21%) pO2 in the presence or absence
of DMOG, after which the frequencies of surface IgG1, IgG2c and IgA among B220+-gated
cells were measured as in Fig. 2 and detailed in the Methods. FACS plots display the
surface levels of IgG1, IgG2c, and IgA on B220+-gated cells in one experiment representative
of three independent replicates. (c) HIF inhibition impedes the hypoxia-induced alteration
of Ab class switch choices. B cells were activated and cultured (4 d) with BAFF, LPS,
and the indicated switching condtions as in Fig. 2a (IL-4, IgG1; IFN-γ, IgG2c; retinoic
acid, TGF-β, IL-4 and −5, IgA) at pO2 of 21% (normoxia) or 1% (hypoxia) in the presence
or absence of the HIF inhibitor Bay 87–2243. FACS plots displaying the surface levels
of IgG1, IgG2c, and IgA on B220+-gated cells in one representative result among three
independent experiments are shown.
Extended Data Figure 4
Hypoxia and PHD inhibition repress T-bet induction
(a, b) B cells from WT mice were activated and cultured in LPS, BAFF, and IL-4 or
IFNγ for 4 days under normoxic and hypoxic conditions (a) or cultured with and without
DMOG at pO2 of 21% (b). Shown are results of immunoblots using anti-T-bet Ab along
with actin as a loading control. Data are one representative result from among three
independent experiments. (c) HIF-dependent regulation of T-bet expression by pVHL.
B cells from WT or conditionally deleted Vhl and Vhl; Hif1a; Epas1 (Vhl, and V;H1;H2,
respectively) cKO mice were activated and cultured (4 d) in LPS and BAFF in the presence
or absence of IFN-γ, as indicated. Results of one representative immunoblot (from
among three independent experiments) probed for HIF1α, T-bet and actin are shown.
(d) HIF superinduction by pVHL depletion in B cells at 1% pO2. WT and B cells after
conditional Vhl f/f deletion were activated, cultured in 1% pO2 as in Extended Data
Fig. 1a, and analyzed by flow cytometry after processing together for indirect immunofluorescent
staining of intracellular HIF-1α as in Fig. 1a and Extended Data Fig. 1a. Numbers
denote the mean fluorescent intensity of the B cells of each type. (e) Flow cytometric
data from one representative experiment as in Fig. 3e, in which B cells were transduced
with MIT, MIG, MIT-T-bet or pMx-GFP-AID retrovectors, and cultured with BAFF and LPS
± IFN-γ in the presence or absence of DMOG. The frequencies of surface IgG2c+ events
among B220+ cells analyzed 4 d after transduction are shown, with flow data from one
experiment of three independent experiments.
Extended Data Figure 5
VHL regulates Ag-specific Ab production
(a) Schematic outline of adoptive transfer experiments. B cells purified from tamoxifen-treated
WT, Vhl
f/f, or Vhl
f/f; Hif1a
f/f; Epas1
f/f CreERT2+ mice were transferred into recipients after mixing with CD4+ T cells
(polyclonal : Ag-specific = 4:1). Recipients were analyzed after primary (10) immunization
or, for memory responses, after the 10 and a recall immunization. (b) As in Fig. 3a,
except B cells from WT or conditionally deleted Vhl cKO mice were mixed with CD4+
OT-II TCR transgenic T cells, transferred into Ig CH allotype-disparate recipient
mice, followed by immunization with NP-ovalbumin and harvest 3 wk after primary immunization.
Donor- ([b] allotype) and recipient-derived ([a] allotype) NP-specific IgM and IgG1
levels in the sera were analyzed by ELISA. The mean (±SEM) absorbance data averaging
independent samples [n=8 (WT) and 7 (Vhl cKO)] obtained in two separate transfer experiments
(measured on the same ELISA plate) are shown. (c, d) As in Fig. 3a, WT or Vhl Δ/Δ
(Vhl cKO) B cells were mixed with WT CD4+ T lymphocytes (a 4:1 mixture of polyclonal
and OVA-specific OT-II cells), and transferred into Rag0
recipients that were then immunized with NP-ovalbumin, and analyzed for NP-specific
Ab levels 3 wk after primary (10) immunization (c) or, for memory response, 9 wk after
the 10 and 1 wk after a recall immunization (panel d). [n= 5 independent recipients
per genotype in two independent experiments.] (c) Mean (±SEM) ELISA data for all-affinity
IgM anti-NP from the same samples as Fig. 3b are shown. (d) Impaired immune memory
follows interference with B cells’ hypoxia response system. Terminal sera obtained
from the recipient mice (Fig. 3a) 1 wk after recall immunization were analyzed by
ELISA for all-affinity anti-NP Ab of the indicated isotypes at the same time as the
10 response samples (Fig. 3a; Extended Data Fig. 5c).
Extended Data Figure 6
HIF-dependent regulation of antigen-specific B cell population and antibody response
by pVHL
(a, b) As in Fig. 3, WT, Vhl Δ/Δ (“Vhl cKO”), or Vhl Δ/Δ Hif1a Δ/Δ Epas1 Δ/Δ (“V;H1;H2
cKO”) B cells were mixed with WT CD4+ T lymphocytes (a 4:1 mixture of polyclonal and
OVA-specific OT-II cells), transferred into Rag0
recipients that were then immunized with NP-ovalbumin and analyzed for NP-specific
Ab levels after primary (10) immunization as in Fig. 3b, c. Using the same mice and
samples as for Fig. 3b, c, cells in spleen secreting IgG2c anti-NP were quantified
by ELISpot and averaged as frequencies of Ab-secreting cells (ASC) in the sample (a).
Mean (±SEM) frequencies for all samples (n=9 each) are shown. (b) Anti-NP IgA levels
in the sera of the samples used in Fig. 3b were quantified by ELISA. (c, d) VHL regulation
of Ag-specific GC and memory B cells is HIF-dependent. As in Fig. 3b, c, WT, pVHL-depleted
(Vhl Δ/Δ , Vhl cKO), or pVHL, HIF-1α, HIF-2α-depleted (Vhl Δ/Δ Hif1a Δ/Δ Epas1 Δ/Δ,
“V;H1;H2 cKO”) B cells were mixed with CD4+ T cells (4: 1 polyclonal: OVA-specific),
transferred into Rag0
mice, immunized with NP-SRBC along with NP-OVA, boosted with NP-OVA at 3 wk after
primary immunization, and analyzed at 1 wk after the boost. Shown are the mean (±SEM)
frequencies or numbers of Ag (NP)-binding B cells of GC- (IgD− GL7+) (c), and early
memory (IgD− GL7− CD38hi) phenotypes (d) derived from each donor population and recovered
in the recipient mice, as determined by enumeration and flow cytometric phenotyping
with fluor-conjugated NP.
Extended Data Figure 7
Hypoxia interrupts impairs an activation-induced feed-forward loop in which mTORC1
increases leucine uptake by B cells
(a) PHD inhibition attenuates mTORC1 activity. WT B cells were activated with αIgM
and cultured (2 d) in BAFF, rested 20 h in the presence or absence of DMOG, and then
re-stimulated (20 min) with αIgM. Shown are immunoblots probed with anti-HIF1α, anti-p-S6K,
anti-p-S6, and anti-S6 Ab along with anti-actin as a loading control. Data are the
results from one representative experiment among three independent replicates. (b–f)
Hypoxia and HIF stabilization reduce leucine uptake and mTORC1 activation. (b, c)
Reduced leucine uptake (b) and Slc7a5 mRNA encoding the large neutral a.a. transporter
LAT1 (c) with inhibition of PHD proteins or mTOR. WT cells were analyzed after culture
in 1% O2 or at pO2 of 21%, in presence of vehicle, PHD inhibitor (DMOG), or mTORC1
inhibitor (Rapa) as indicated. (b) Mean (±SEM) B cell uptake of leucine, in n=3 independent
experiments. (c) Mean (±SEM) relative mRNA level, normalized to actin (n=3 independent
experiments). (d, e) Activated B cells of the indicated genotypes were assayed for
leucine uptake (d) and induction of the Slc7a5 gene encoding a large neutral a.a.
transporter (e). (d) Mean (± SEM) leucine uptake by the cultured cells, normalized
in each independent experiment (n=3) to activated WT cells. (e) VHL loss leads to
HIF-dependent attenuation of Slc7a5 mRNA levels. WT or cKO B cells of the indicated
genotypes were activated and cultured at 21% O2 as in Fig. 3d. Mean (± SEM) qPCR results
normalized first to actin for level within a sample, and then to the WT control in
each independent experiment (n=3). (f) Leucine stimulates mTORC1 activity in activated
B cells. Activated WT B cells, divided and cultured overnight in medium lacking or
sufficient for the indicated a.a., were restimulated and analyzed as in Fig. 4a, b.
Extended Data Figure 8
Hypoxia promotes AMPK activity and induction of the mTORC1 inhibitor REDD1 without
repressing mTORC2
(a) B cells were activated and grown (2 d) in LPS and BAFF at the indicated pO2 and
in the presence or absence of IFN-γ as indicated. ATP concentrations in equal numbers
of cells were then assayed. In each of three replicate experiments with similar results,
the [ATP] measured for cells at conventional (21%) pO2 without IFN-γ was set as 1,
and the mean (±SEM) levels in each sample relative to this reference are shown for
three biological replicates. (b) Immunoblot results after probing membranes with anti-p-ACC,
anti-p-AMPK(T172), and actin are shown for one representative experiment. Numbers
indicate the level of signal for cells cultured in hypoxia or DMOG as compared to
the reference value of the sample cultured in conventional (21%) pO2, after normalization
of each sample according to its loading. (c) Results of a representative qRT2-PCR
experiment measuring Redd1 mRNA in WT B cells (activated and cultured as in panel
b), with each sample first normalized to Actin mRNA and then to vehicle-treated cells.
(d, e) Effect of VHL, hypoxia, and DMOG on Akt phosphorylation in B cells. (d) B cells
were activated with anti−IgM and BAFF, cultured (2 d) and rested (20 h) under conditions
of hypoxia, or normoxia in the presence or absence of DMOG, after which cells were
re-stimulated (20 min) with anti−IgM. (e) As in (d), B cells from WT or conditionally
deleted Vhl cKO mice were activated with anti−IgM in the presence of BAFF, cultured
(2 d) and rested (20 h), after which cells were re-stimulated (20 min) with anti−IgM.
Shown are results of immunoblots probed with antibodies directed against p-Akt (T308),
p-Akt (S473), and Akt. Numbers show the quantitation of signal relative to B cells
that were not restimulated, after adjustment of each sample for loading as determined
by total Akt. Data shown are from one representative experiment among three independent
replicates.
Extended Data Figure 9
mTORC1 regulates expansion of Ag-specific B cells and Ab class spectrum
(a) Results of immunoblots using anti-Raptor, and anti-p-S6 along with anti-S6 Ab
as a loading control. B cells (WT or haploinsufficient for Raptor) were activated
with F(ab’)2 anti−IgM and BAFF, cultured (2 d) and rested (20 h), after which cells
were re-stimulated (20 min) with F(ab’)2 anti−IgM. Data are from one representative
experiment among three independent replicates. (b) Recipient Ab controls for effect
of mTORC1 on class-switched Ab responses. As in Fig. 4c, WT or Raptor-haploinsufficent
B cells (from heterozygous mice that were ROSA26-CreERT2, Rptor f/+ and converted
to Δ/+ by tamoxifen injections) were mixed with CD4+ OTII TCR transgenic T cells,
transferred into Ig CH allotype-disparate recipient mice, immunized with NP-OVA, and
harvested at 3 wk after primary immunization. Donor- ([b] allotype) (in Fig. 4) or
recipient-derived ([a] allotype) NP-specific IgG1 and IgG2c levels in the sera were
analyzed by ELISA. Mean (±SEM) absorbance data averaging samples [n=9 (WT) vs. 8 (Rptor
+/Δ)] obtained in three separate experiments (measured on the same ELISA plate). (c–e)
WT or Rptor Δ/+ B cells were mixed with CD4+ T cells (polyclonal : OVA-specific =
4: 1) and transferred into Rag0
mice and immunized with NP-OVA. Shown are the recoveries of Ag (NP)-binding WT vs
Rptor Δ/+ B cells of GC-phenotype (B220+ GL7+ IgD−) (c) and early memory (B220+ CD38+
GL7− IgD−) (d). (e) Generation of Ag-specific IgG2c-secreting cells depends on mTORC1.
Mean (±SEM) results of ELISpot assays quantitating NP-binding IgG2c (b allotype) Ab-secreting
cells (ASC) from the experiments in Fig. 4c, d, and Extended Data Fig. 9b, quantified
as described in Extended Data Fig. 6a.
Extended Data Figure 10
mTORC1 is rate-limiting for AID expression and switching to IgG2c
(a) A division-independent mechanism dependent on mTORC1 quantity in B cell switching
to IgG2c. Flow cytometric data in the B cell gate, displaying CFSE partitioning (fluorescein
emission intensities) versus the IgG2c, were all from one experiment representative
of three independent biological replicates. WT or Rptor Δ/+ B cells were stained with
CFSE and cultured with LPS, BAFF and IFN-γ, and analyzed by flow cytometry. (b) WT
or Rptor Δ/+ B cells were cultured (2 d) with LPS, BAFF, and IFN-γ. Mean (±SEM) levels
of mRNA encoded by the Aidca (left) and Tbx21 (right) genes measured in three independent
replicate experiments by qRT2-PCR normalized to actin in the sample and then to the
level in WT cells (set as relative level of 1). (c) Immunoblots probed for Raptor,
T-bet, and actin, as indicated, using B cells as in (b). [representative of n=3 independent
experiments] (d) mTOR promotes switching to IgG by division- independent mechanisms.
As in panel a, but CFSE-stained WT B cells were activated and cultured for 4 days
with LPS, BAFF and IFN-γ in the presence or absence of rapamycin vs vehicle. (e, f)
mTORC1 regulation of AID level in collaboration with T-bet determines efficient switching
to IgG2c. B cells were transduced with MIT, MIG, MIT-T-bet or pMx-GFP-AID retrovectors,
and cultured with BAFF and LPS ± IFN-γ in the presence or absence of rapamycin (5
nM). (e) Representative flow data, from one experiment among three independent replicates,
derived as in Extended Data Fig. 4e. (f) Mean (±SEM) frequencies of surface IgG2c+
events among B220+ cells analyzed 4 d after transduction are shown (n=3 independent
experiments).
Supplementary Material
1