ABSTRACT
Activation-induced cytidine deaminase (AID) initiates immunoglobulin (Ig) class switch recombination (CSR), somatic hypermutation (SHM), and gene conversion by converting DNA cytosines to uracils at specific genomic regions. In this study, we examined AID footprints across the entire length of an engineered switch region in cells ablated for uracil repair. We found that AID deamination occurs predominantly at WRC hot spots (where W is A or T and R is A or G) and that the deamination frequency remains constant across the entire switch region. Importantly, we analyzed monoallelic AID deamination footprints on both DNA strands occurring within a single cell cycle. We found that AID generates few and mostly isolated uracils in the switch region, although processive AID deaminations are evident in some molecules. The frequency of molecules containing deamination on both DNA strands at the acceptor switch region correlates with the class switch efficiency, raising the possibility that the minimal requirement for DNA double-strand break (DSB) formation is as low as even one AID deamination event on both DNA strands.
INTRODUCTION
At the mammalian immunoglobulin heavy chain (IgH) locus, the rearranged VH region is followed by tandem constant (C) regions, of which only the VH-proximal C region is expressed (1–4). Switching from the default expression of IgM (encoded by the μ constant region) to other isotypes (e.g., IgG, IgE, or IgA) occurs in antigen-stimulated B cells through an intrachromosomal deletion known as class switch recombination (CSR). CSR allows a B cell to express a different class of Ig (defined by the IgH C region) while maintaining the same antigen-binding specificity of the Ig. This is critical for the optimal elimination of infection in a humoral immune response because the C region mediates the effector function of Ig. Each C region is preceded by a large zone (2 to 10 kb) of repetitive DNA known as the switch (S) region, where DNA double-strand breaks (DSBs) are generated as requisite intermediates during CSR. A single CSR event involves the generation and joining of DSBs in the donor (e.g., μ) and acceptor (e.g., γ, ε, or α) S regions (1–4). Because DSBs are some of the most severe forms of DNA damage, dysregulated CSR can lead to oncogenic chromosomal translocations and cancer.
CSR is initiated by activation-induced cytidine deaminase (AID) binding to transcribed S regions and catalyzing DNA cytosine deamination that converts cytosines to uracils. Uracil in DNA can be either recognized as base damage and repaired by the base excision repair (BER) pathway or recognized as U-G mismatches and repaired by the mismatch repair (MMR) pathway (1–4). Apparently, both pathways are involved in the processing of AID-generated uracils, as the genetic ablation of components in either pathway (e.g., UNG2, APE1, MSH2, MSH6, MLH1, PMS2, and Exo1, etc.) inhibits CSR to various degrees (1–4). Intriguingly, BER and MMR in most cell types are error-free processes, but the processing of AID-generated uracils in B cells is definitively error prone, resulting in DSBs in S regions and point mutations in V regions. How these two pathways are diverted in B cells to affect opposite outcomes (error-free repair versus mutations and DSBs) remains an incompletely answered question.
AID is a single-strand-specific deaminase (5–9). The absolute requirement for transcription for somatic hypermutation (SHM) at V regions and CSR at S regions is generally viewed as a necessity to separate the two DNA strands to make them accessible to AID (8, 10). However, transcription itself is not sufficient to explain why the Ig V and S regions are premier targets for AID. While the uniqueness of V region transcription (for targeting AID during SHM) remains unclear, a unique feature of S region transcription has been well documented. Mammalian S regions are rich in G-cluster motifs on the nontemplate strand, thereby making a G-rich RNA transcript, which is postulated to promote transcription-dependent R-loop formation (11). R-loops consist of a displaced single-stranded nontemplate strand and an RNA-DNA hybrid formed between the RNA transcript and the template DNA strand. AID may directly target the single-stranded DNA (ssDNA) within the R-loop or bind the S region transcript that acts as a guide RNA to bring AID to the S region by base pairing (12). Theoretically, SHM requires deamination on only one DNA strand (13), whereas CSR likely requires lesions on both DNA strands since CSR requires the generation of DSBs. Thus, it is critical to understand how uracils are distributed across the two DNA strands at S regions during typical CSR. Because of this gap in knowledge, our understanding of how switch region uracils lead to DSBs is limited and largely speculative.
To facilitate AID footprint delineation on a single-molecule basis, a number of genetic modifications are required. First, the experiment must be performed in UNG2−/− MSH2−/− cells such that uracil is not repaired and can be detected as a C-to-T mutation following PCR amplification (14). Second, the S region must be engineered to be short enough to allow PCR across its entirety while still supporting relatively efficient CSR. Third, the cell must be monoallelic for the S region that is being examined. This is because each AID-targeted allele can potentially yield two different PCR products. CSR occurs biallelically in stimulated B cells, with each cell containing two IgH alleles that can potentially yield four different PCR products (top and bottom strands of each allele). When performing PCR from a pool of BER/MMR-ablated cells, although one can differentiate top- and bottom-strand-derived PCR products (based on C>T versus G>A substitutions), it would not be feasible to determine which PCR products originate from the same allele. Introducing all of these modifications in a mouse model would be difficult (15, 16) but is relatively straightforward in a B cell line called CH12F3, which is capable of robust cytokine-dependent CSR in vitro (17). A subclone of CH12F3 (CH12F3.2A [here called CH12F3]) has a productive (successful VDJ rearrangement) IgH allele in the germ line configuration and a nonproductive allele (DJ join) that already underwent Sμ-Sα recombination (18). Therefore, this cell line is monoallelic for germ line Sμ or Sα and suitable for the delineation of AID footprints on a single-molecule basis.
In this study, AID footprints were analyzed across the entire length of an engineered Sα region from either a pool of cells or individual cell clones. We found that AID deaminations at Sα during CSR are rare and mostly distally spaced. The frequency of molecules harboring AID deamination on both DNA strands at Sα correlates with the CSR frequency, suggesting that DSB formation requires deamination on both DNA strands, and the minimal requirement might be as low as one deamination event on each strand. The implications of these findings are discussed.
RESULTS
Genetic manipulations required for AID footprint analyses in CH12F3 cells.To delineate AID deamination (termed AID footprints), several requisite genetic modifications were performed in the CH12F3 cell line. First, the endogenous Sα region was shortened to allow PCR amplification while still supporting relatively efficient CSR. Previously, we replaced endogenous Sα in CH12F3 cells with a 1.1-kb “core” using recombinase-mediated cassette exchange (19). This 1.1-kb core Sα represents the most uniformly repeated region of Sα and supports CSR to ∼30 to 40% of the wild-type (WT) level (19). In this study, the delineation of AID footprints was performed mostly on this core Sα. Second, the UNG2 and MSH2 genes were disrupted individually or in combination in WT as well as Sαcore cells by CRISPR/Cas9-mediated genome editing (Fig. 1). MSH2 disruption (Fig. 1A) was screened by Western blotting (Fig. 1B) and confirmed by sequencing of both alleles for CRISPR-mediated deletions. We found that MSH2 deficiency alone results in an ∼3-fold reduction in CSR efficiency (Fig. 1C), consistent with previous studies showing that MSH2 is required for maximal CSR efficiency (20, 21). UNG2 disruption (Fig. 1D) was screened by an activity assay (Fig. 1E) followed by Sanger sequencing confirmation of the frameshift deletions on both alleles. Although UNG2 deficiency alone in CH12F3 cells virtually abolishes CSR (Fig. 1F), AID footprint analyses were performed in cells deficient for both UNG2 and MSH2 (to ensure that AID-generated uracil is not repaired) and with the endogenous Sα region replaced by the 1.1-kb core. The resulting cell line is named Sαcore-dKO here.
CRISPR-mediated disruption of MSH2 and UNG2 in CH12F3 cells. (A) CRISPR target site in the MSH2 gene. (B) Loss of MSH2 protein in MSH2−/− cells shown by Western blotting. (C) Reduced CSR in MSH2−/− cells. Cells were stimulated with CIT (anti-CD40 plus IL-4 and TGF-β1) for 72 h. (D) CRISPR target site in the UNG gene. (E) Loss of uracil repair activity in the UNG2−/− cell extract. (F) Loss of CSR in UNG2−/− cells. Cells were stimulated with CIT for 72 h. FSC, forward scatter.
Frequency and distribution of AID deamination at S regions.The engineered core Sα region made it possible to delineate AID footprints across an entire switch region. To accumulate enough events for an overview of the frequency and distributions of AID footprints across the Sα region, Sαcore-dKO cells were stimulated with CIT (anti-CD40 plus interleukin-4 [IL-4] and transforming growth factor β1 [TGF-β1]) for either 1 or 4 weeks (with the necessary splitting of confluent cultures), followed by PCR, cloning, and sequencing of core Sα from cell pools.
As can be seen, after 1 week of stimulation, every sequenced molecule contains at least one C-to-T conversion (AID footprint) (Fig. 2). Some molecules accumulate more than 10 (Fig. 2). The mutation frequency in the core Sα region after 1 week of stimulation was 6.8 × 10−3/bp and increased to 1.3 × 10−2/bp after 4 weeks (Fig. 2). Deamination occurred predominantly at WRC sites (where W is A or T and R is A or G) (89% for 1 week and 80% for 4 weeks) (Fig. 2), which are known AID hot spots. Deamination at non-WRC sites increased slightly over time, possibly due to a gradual depletion of WRC sites. We estimated that CH12F3 cells double every 12 to 16 h when cultured in CIT medium (17, 22). Therefore, for a week, there are 10 to 14 generations. Based on this estimation, we calculate that less than one (∼0.7 to 0.9) AID deamination event occurs per cell per generation within this core Sα target sequence. This number is likely an overestimation since early conversions are propagated into daughter cells and accumulate over generations. Although perhaps not every cell in the population responds to stimuli, the above-described estimation nevertheless indicates a very low frequency of AID deamination at the acceptor S region during CSR.
Frequencies and distributions of AID footprints at pre-Sμ or Sα regions. Pie charts show the frequency of AID deamination. The number in the pie center indicates the number of molecules sequenced. Each slice indicates the proportion of sequences containing various numbers of mutations. The graphs on the right side of the pie charts show the distribution of AID footprints across the switch regions. Each vertical line indicates the percentage of C-to-T changes at that position. Lines above or below the horizontal line indicate deamination on the nontemplate or template strand, respectively. Red indicates an AID hot spot, and blue indicates a non-hot spot. (A) Sα region from cells stimulated for 1 week. (B) Sα region from cells stimulated for 4 weeks. (C) Pre-Sμ region from cells stimulated for 1 week. (D) Pre-Sμ region from cells stimulated for 4 weeks.
Deamination on the template strand is roughly equal to that on the nontemplate strand (48% versus 51% at 1 week and 44% versus 56% at 4 weeks, respectively). Also, AID footprints appear evenly distributed across the entire length of the S region (Fig. 2), in contrast to SHM, where the mutation frequency rises sharply ∼150 bp downstream of the transcription initiation site and then decreases exponentially over a 1- to 2-kb region (23, 24).
The donor Sμ region was not engineered to facilitate complete PCR amplification. To assess deamination at Sμ, a small region at the beginning of Sμ (called pre-Sμ) was examined as a surrogate for Sμ (25–27). The mutation frequencies at pre-Sμ were 5.6 × 10−3/bp for week 1 and 1.3 × 10−2/bp for week 4 (Fig. 2), similar to those of core Sα. Also similar to core Sα, AID deamination at the pre-Sμ region was roughly equal between the two DNA strands (Fig. 2). There was a slightly higher number of deamination events at non-WRC sites (25% after 1 week and 27% after 4 weeks of stimulation) at pre-Sμ. These results indicate that AID targets both DNA strands and deaminates cytosines predominantly at WRC motifs in vivo.
Few mutations were found in nonstimulated cells (1/14,971 for core Sα (cSα) and 3/5,850 for pre-Sμ). Because they are exclusively C-to-T mutations, they are likely deamination products by a basal level of AID expression in unstimulated cells rather than PCR errors.
Single-molecule analyses of AID footprints.DSBs at donor and acceptor S regions are essential intermediates of CSR. The generation of DSBs requires AID-initiated lesions, likely on both DNA strands. Thus, it is critical to delineate how AID deaminations are distributed across both DNA strands of individual molecules. After AID deaminates an S region, the two DNA strands are no longer completely complementary; each can give rise to a different PCR product. In experiments where genomic DNA from a pool of cells is used as the PCR template, it is impossible to identify the original pair for each molecule. However, initial attempts to amplify S regions from single cells yielded very few PCR products. Therefore, we used an alternative approach (Fig. 3A): cells were stimulated for a brief period of time (24 h) before cytokine withdrawal and then individually seeded into 96-well plates to allow clones to develop in the absence of stimuli. The choice of 24 h is based on observations and previous results showing peak levels of AID expression at this time point (28, 29). Prolonged stimulation was avoided to limit multiple rounds of AID-induced mutations. For each clone, the PCR product consists of two “species,” one derived from the template strand and the other derived from the nontemplate strand. In an individual cell, if no AID deamination has occurred or if a cell harboring deamination has replicated prior to single-cell seeding, the two species within a PCR are identical, as is the case for the majority of the clones analyzed (289/307). However, if even a single AID deamination event occurred, the two species will be different, resulting in a double peak(s) on a sequencing chromatogram (Fig. 3B).
Strategy for AID footprint analysis of individual S regions. (A) AID-mediated cytosine deamination generates uracil in DNA and converts a normal C-G base pair to a U-G mispair. Cell division separates the mispair, in the absence of uracil repair, into daughter cells, resulting in two different “species” (one derived from the top strand and the other derived from the bottom strand). Sanger sequencing of the PCR product from each cell clone results in a double peak on the chromatogram at this position. (B) Representative chromatograms showing double peaks as an indication of PCR product heterogeneity from individual cell clones. (Left) Double peak of C and T reflecting AID deamination on the nontemplated DNA strand. (Right) Double peak of G and A reflecting AID deamination on the template DNA strand. Sequence heterogeneity was confirmed by cloning of the PCR product followed by Sanger sequencing.
Using this method, a total of 307 PCRs (from individual cell clones) were sequenced, 18 of which showed one or more double peaks (C/T or G/A) in the chromatogram, indicating the existence of two species in these PCRs. These 18 PCR products were cloned into a plasmid and then transformed into Escherichia coli to separate the two species (among different bacterial colonies). After Sanger sequencing of at least 12 bacterial colonies from each cloning, AID deamination profiles on both DNA strands from these 18 molecules were reconstituted (Fig. 4). In 16 of the 18 cases, both species were successfully recovered. In the other two cases, only WT sequences were recovered, suggesting that the initially observed double peaks were sequencing artifacts. No PCR product yielded more than two species after cloning and sequencing, supporting the fact that these were authentic clones. No mutations that were pure C-to-T or G-to-A conversions were found in any of the PCR products, suggesting that no AID deamination had occurred before single-cell seeding. All but one AID deamination occurred at a WRC hot spot motif (Fig. 4).
AID footprints of individual Sα regions. Each line represents a single molecule of the Sα region. Dots above or below the line indicate deamination on the nontemplate or template strand, respectively. Red indicates AID hot spots, and blue indicates non-hot spots. Numbers at the bottom are coordinates of nucleotide positions in the Sα region.
Out of the 16 molecules that harbor distinct AID deamination patterns between the two DNA strands, 8 molecules have deamination on only one of the strands (Fig. 4, molecules 2 through 5, 8, and 11 through 13), and the other 8 molecules contain deamination on both strands (Fig. 4, molecules 1, 6, 7, 9, 10, and 14 through 16). It has been suggested that CSR requires proximate single-strand breaks on both DNA strands, at overlapping AID hot spots (i.e., WGCW); however, only one molecule shows evidence of such staggered deamination at an overlapping AID hot spot motif (Fig. 4, molecule 14). Most deamination events observed in this study are isolated and quite distal to each other, suggesting that the majority of the DSBs at acceptor S regions may result from distally located nicks. Clustered deamination events are observed in a number of molecules (Fig. 4, molecules 10, 11, and 14 through 16), evident for a limited degree of AID processivity (30, 31) in vivo. Of note, this clustering occurs mostly on one DNA strand.
There has been evidence that CSR initiates at Sμ and secondarily involves downstream S regions (32), suggesting that cells harboring DSBs in acceptor switch regions may already have DSBs in the donor Sμ region. In this regard, the frequency of DSBs at the acceptor switch regions may dictate CSR frequency. Interestingly, the frequency (2.6%) of molecules harboring mutations on both DNA strands at Sα correlates well with the CSR efficiency (2.7% ± 0.4%) when the same cells are complemented with UNG2 and MSH2 cDNAs and stimulated for 24 h (Fig. 5). These data suggest that the minimal requirement for DSB formation at the switch region may be as low as one uracil on each DNA strand.
CSR efficiency of Sαcore-dKO cells complemented with UNG2 and MSH2. (A) Representative histograms from CSR assays after 24 h of CIT stimulation. (B) Average CSR efficiency of Sαcore-dKO cells complemented with UNG2 and MSH2 after 24 h of CIT stimulation. Error bars indicate standard deviations of the means from three independent experiments.
DISCUSSION
This study provides new insight into how AID facilitates deamination in S regions during CSR. First, AID targets template and nontemplate strands at similar frequencies. Because AID is an ssDNA deaminase, it was thought that AID would preferentially target the nontemplate strand (10), especially in the context of an R-loop structure (4). This might suggest that deamination occurs predominately after R-loops are processed by cellular enzymes that expose ssDNA on both DNA strands. One candidate for such an activity is the RNA exosome, which had already been implicated in CSR (33). The RNA exosome degrades myriad RNA species and may have direct and indirect roles in CSR (34, 35). Other candidates for such an activity would be RNase H enzymes. Mammals have two distinct RNase H enzymes, RNase H1 (single subunit) and RNase H2 (three subunits) (36). Although neither has been implicated in CSR, in vitro, E. coli RNase H1 can convert R-loops into structures termed “collapsed R-loops” that contain single-stranded regions on both DNA strands (as a result of a misalignment of switch repeats) (4, 11). A recent structural study demonstrated that AID has two distinct DNA-binding grooves, nicely explaining why AID has higher binding affinity and deaminase activity for branched or G4 DNA structures than linear DNAs (37). Collapsed R-loops contain numerous branched DNA structures; this could explain how AID deaminates both DNA strands (3). It also provides an explanation for the observation that although the primary sequences of switch regions are not highly evolutionarily conserved, the repetitive nature of switch regions is uniformly observed in all species. Second, AID footprints distribute evenly across the entire length of an S region, coincident with the random distribution of recombination breakpoints across S regions. This feature is also consistent with the R-loop/collapsed R-loop model, since R-loops initiate and terminate randomly across the S region (11). In contrast, SHM is not evenly distributed over a distance (23, 24), suggesting that AID deaminates S and V regions via distinct mechanisms. However, S region deamination via an SHM-like mechanism (e.g., stalled transcription bubble) cannot be formally ruled out.
In this study, we examined AID deamination on individual molecules within a single cell cycle. Interestingly, most AID deamination events are distally located. This appears at odds with the widely conjectured mechanism by which DSBs form predominantly by deamination at overlapping AID hot spots (e.g., 5′-AGCT-3′), resulting in nicks just across from one another. Surprisingly, only 1 of 8 molecules harboring AID deamination on both strands displayed this pattern. One caveat is that distally located AID footprints on primary sequences may actually be spatially close in a collapsed R-loop. Additional secondary structures such as G4 quadruplexes might also contribute to bringing spatially distant sites into proximity, although there has been no clear evidence of G4 quadruplex formation in S regions in living cells. Previous studies proposed a model involving MMR-mediated strand excision as a means to create DSBs from distally located nicks, potentially explaining why MMR components are required for fully efficient CSR (2, 38, 39). Interestingly, it was shown recently that distal nicks generated by Cas9 nickase (Cas9n) can mediate “CSR-like” intrachromosomal deletions (40), suggesting that distal nicks are indeed capable of inducing DSB formation. However, there are intrinsic differences between Cas9n-generated nicks and uracil-induced nicks. DNA uracil has only one chance to induce strand breakage within a cell cycle because it will be either repaired via BER or replicated over by a DNA polymerase. In contrast, Cas9n will keep nicking the DNA even after the nicks are repaired (e.g., joined by DNA ligase I or III) as long as the guide/protospacer-adjacent motif (PAM) sequences remain intact. In addition, Cas9n unwinds the local duplex upon binding to the guide sequences. Thus, nicking by Cas9n does not directly mirror the nicking mechanism via DNA uracil repair.
MATERIALS AND METHODS
Reagents, cell culture, and class switch recombination assay.CH12F3 cells were cultured in RPMI 1640 medium supplemented with 10% (vol/vol) fetal bovine serum and 50 μM β-mercaptoethanol. To induce CSR, CH12F3 cells were seeded at 5 × 104 cells/ml in culture medium containing 0.5 μg/ml anti-CD40 antibody (Ab) (clone HM40-3; eBioscience), 5 ng/ml IL-4 (R&D Systems), and 0.5 ng/ml TGF-β1 (R&D Systems) and grown for 72 h unless otherwise specified. Cells were stained with a fluorescein isothiocyanate (FITC) (or CF633, equivalent to allophycocyanin [APC])-conjugated anti-mouse IgA rat monoclonal antibody (clone C10-3; BD Biosciences) and analyzed on an LSR II flow cytometer (BD Biosciences). CSR efficiency was determined as the percentage of IgA-positive cells. Rat anti-AID (catalog number 14-5959-82; eBioscience) and mouse anti-MSH2 (catalog number sc-376501; Santa Cruz Biotechnology) monoclonal antibodies were used for Western blotting. Oligonucleotides were ordered from Millipore-Sigma and Integrated DNA Technologies.
Knockout and genetic complementation of the UNG2 and MSH2 genes.Annealed oligonucleotides were cloned into pSpCas9(BB)-2A-Puro (Addgene plasmid 62988) to make CRISPR vectors as described previously (41). The targeting sequence for the MSH2 gene is 5′-GTCAAAGACATGTTGGAGTT-3′, and the one for the UNG2 gene is 5′-GCTGGAAGCAGCAGCTGTGC-3′. CRISPR vector-transfected cells were allowed to recover for 48 h before being cloned in 96-well plates by limiting dilution. Cell clones were screened by Western blotting (MSH2) or an activity assay (UNG). Candidate clones were further analyzed by PCR cloning and Sanger sequencing to confirm the frameshift deletions on each allele.
A bicistronic expression vector (pEF-UNG2-2A-MSH2-ires-neo) where the UNG2 and MSH2 cDNAs were separated by a self-cleaving 2A peptide was constructed to complement the UNG2 and MSH2 deficiencies simultaneously in UNG2−/− MSH2−/− cells. G418-resistant stable transfectants were screened for UNG expression by a UNG activity assay and for MSH2 expression by Western blotting.
Uracil-DNA glycosylase assay.One million cells were lysed in 50 μl of a solution containing 20 mM Tris-Cl (pH 7.5), 100 mM NaCl, 10 mM EDTA, and 0.5% Igepal CA-630 by freezing and thawing twice. After centrifugation for 10 min at 21,000 × g, the supernatant (cell extract) was collected. A uracil-DNA glycosylase assay was carried out in a 10-μl reaction mixture containing 10 mM Tris-HCl (pH 7.4), 50 mM NaCl, 5 pmol of the fluorescein-labeled oligonucleotide substrate, and 2 μl of the cell extract. The reaction mixture was incubated for 10 min at 37°C before sodium hydroxide was added to a final concentration of 0.1 M and incubated for 5 min at 95°C. The reaction mixture was mixed with an equal volume of formamide and heated for 5 min at 95°C before being resolved on a 10% denaturing polyacrylamide gel. Gel images were acquired on a Typhoon 9500 phosphorimager (GE Life Technologies).
AID footprint analyses.For AID footprint analyses on cell pools, cells were seeded at 5 × 104 cells/ml in medium containing CIT and grown for either 1 or 4 weeks with subculturing (1:5) when the cell density reached 1 × 106 cells/ml. Switch regions were PCR amplified with the Phusion high-fidelity DNA polymerase (New England Biolabs) from genomic DNA extracted from cell pools, cloned, and sequenced.
For AID footprint analyses on individual molecules, cells were seeded at 5 × 104 cells/ml in medium containing CIT and grown for 24 h before being collected and washed twice with medium without cytokines (CIT withdrawal). Cells were seeded into the wells of 96-well plates at 50 cells per plate and examined visually under a microscope. After 10 days, clones were picked for PCR (by Phusion) and Sanger sequencing. PCR products showing double peaks on the chromatogram were cloned into a plasmid and transformed into E. coli. At least 12 colonies from each cloning were sequenced to separate the template- and nontemplate-strand-derived molecules coexisting in the PCR product (which is responsible for the appearance of double peaks on the chromatogram).
ACKNOWLEDGMENTS
This work is supported by NIH grants R21AI126359 and R01AI139039 to K.Y.
FOOTNOTES
- Received 26 March 2020.
- Returned for modification 14 May 2020.
- Accepted 1 June 2020.
- Accepted manuscript posted online 8 June 2020.
- Copyright © 2020 American Society for Microbiology.
