Abstract

   This report demonstrates a novel class of innate immune cells
   designated “variable immunoreceptor–expressing myeloids” (VIREMs).
   Using single-cell transcriptomics and genome-wide epigenetic profiling,
   we establish that VIREMs are myeloid cells unrelated to lymphocytes. We
   visualize the phenotype of B-VIREMs that are capable of genetically
   recombining and expressing antibody genes, the exclusive hallmark
   function of B lymphocytes. These cells, designated B-VIREMs, display
   monoclonal antibody cell surface signatures and regularly circulate in
   the blood of healthy individuals. Single-cell data reveal clonal
   expansion of circulating B-VIREMs as a dynamic response to disease
   stimuli. Live-cell imaging models suggest that B-VIREMs load their own
   Fc receptors with endogenous antibodies during vesicle transport to the
   cell surface. A first cloned B-VIREM–derived antibody (Vab1)
   specifically binds stomatin, a ubiquitous scaffold protein that is
   strictly expressed intracellularly, allowing Vab1-bearing macrophages
   to phagocytose cell debris without requiring prior opsonization. Our
   results suggest important antigen-specific tissue maintenance
   functionalities in these innate immune cells.
     __________________________________________________________________

   Immunorecognition revisited: smart phagocytes provide a sophisticated
   link of innate and adaptive functions in the immune system.

INTRODUCTION

   The immune system of vertebrates comprises two independent functional
   systems that are strictly separated with respect to their recognition
   mechanisms and their biological effector functions: an innate and an
   adaptive arm characterized by very distinct immunorecognition molecules
   for the detection of infectious agents ([50]1, [51]2), malignant cell
   growth ([52]3, [53]4), and tissue damage ([54]5, [55]6) or the
   surveillance of tissue homeostasis and immune tolerance ([56]7, [57]8).
   Specifically, the innate system comprises cells of myeloid
   differentiation that express a very limited set of genome-encoded,
   invariant pattern recognition receptors (PRRs) like the Toll-like
   receptors, CD14 or MD-2 ([58]9). These receptors are readily available
   to recognize pathogen-associated molecular patterns or
   damage-associated molecular patterns as triggers for innate immune
   activation ([59]10, [60]11). In contrast, the phylogenetically younger
   adaptive immune system comprises lymphocytes that use somatic
   recombination of genome-encoded building blocks to generate, in a
   stochastic fashion, active genes coding for immunoglobulins (Igs) and T
   cell receptors (TCRs). Subsequently, the variable immunoreceptors are
   further improved in their antigen-binding efficiencies through
   antigen-driven hypermutations within the productive variable region
   genes. This enables the adaptive system to recognize and discern
   antigens with very high molecular resolution and affinity ([61]2). In
   contrast to the innate immune system, this adaptive response is slow to
   develop, usually taking several weeks to respond effectively to a
   stimulus ([62]12, [63]13).

   Over the past years, circumstantial evidence of partial mRNA coding for
   single-variable Ig genes has suggested that Ig expression by
   nonlymphocytic cells may occur in sample preparations of neoplastic
   cells ([64]14), myeloid leukemia cell lines ([65]15, [66]16), and
   murine macrophages ([67]17). Our group has reported the first
   full-length Ig heavy- and light-chain antibody sequences obtained from
   a single tumor-associated macrophage ([68]18). However, beyond of the
   functional genomics level, the phenotype of the cells, which we
   designate “variable immunoreceptor–expressing myeloids” (VIREMs), has
   not been demonstrated. In addition, their progeny, the specificities of
   their variable immunoreceptors, and their biological functions remain
   entirely elusive so far.

   Phagocytosis is a major effector function of the innate immune system
   allowing monocytes/macrophages, granulocytes, dendritic cells, and
   osteoclasts to engulf and digest particulate matter like bacteria, cell
   fragments, and microparticles ([69]19, [70]20). If target surfaces are
   decorated (opsonized) with antibodies, then they become “visible” to
   phagocytes through attachment and cross-linking of their cell surface
   Fc γ receptors (FcγRs) generating the specific trigger for
   opsonophagocytosis ([71]21, [72]22). In contrast, confronting
   phagocytes with increasing concentrations of nonimmune human serum are
   known to block their FcγR-mediated effector functions ([73]23).

   Here, we provide the first phenotypic evidence for B-VIREM cells
   expressing monoclonal antibodies, both morphologically and
   functionally, and characterize them at the molecular level. Genome-wide
   methylation studies and single-cell transcriptomics demonstrate the
   myeloid origin of B-VIREM with very close kinship to
   monocytes/macrophages while being developmentally distant from
   lymphocytes. B-VIREM cells expressing monoclonal antibodies can be
   visualized distinctively in the peripheral blood of healthy donors and
   patients, in part showing marked clonal expansions potentially related
   to health conditions. Furthermore, we identify the specific antigen of
   the first cloned B-VIREM–derived monoclonal antibody and provide a
   model for the functionality of endogenous antibody expression in
   myeloid cells simultaneously expressing Fc receptors (FcRs) to allow
   for an antigen-specific yet opsonization-independent way of innate
   phagocytosis. Our work establishes B-VIREM cells as a class of
   monocytes/macrophages expanding our fundamental understanding of
   distinct immune mechanisms in health and disease.

RESULTS

Visualization and characterization B-VIREM in PBMCs

   Normal peripheral blood monocytes show a polyclonal Ig cell surface
   signature due to simultaneous binding of both kappa- and lambda-bearing
   antibodies through FcRs, resulting in mixed fluorescence (fig. S1A).
   Immunostaining of peripheral blood mononuclear cells (PBMCs) identified
   monocytes and B cells by CD14^+ and cyCD79a positivity, respectively,
   and also determined the presence of antibodies by staining for kappa or
   lambda light chains ([74]Fig. 1A). [75]Figure 1B shows a
   superimposition of representative fluorescent images to distinguish
   regular CD14^+ monocytes (blue) from B-VIREM. The latter show
   fluorescence staining in cyan (mix of blue and green) and magenta (mix
   of blue and red), which identifies these CD14^+ monocytes as kappa- or
   lambda-expressing B-VIREM cells, respectively. Multicolor
   immunostaining consistently identified B-VIREM in PBMCs at frequencies
   of approximately 1.73 ± 0.48% among the CD14^+ cells (fig. S1B and
   table S1). [76]Figure 1C and fig. S1C show all B-VIREM cells identified
   in this study and allow for estimation of their frequencies and an
   assessment of the appropriate in-sample controls for sensitivity and
   specificity. Under conditions of high overexposure of [77]Fig. 1B,
   B-VIREM cells stain positive for CD79A, an accessory protein for BCR
   signaling (fig. S1D). These results establish that B-VIREM cells are
   immunocytochemically distinct from B cells and regular CD14^+ monocytes
   and are regularly found circulating in peripheral blood of healthy
   individuals.

Fig. 1. Phenotyping of antibody-expressing cells in the peripheral blood of
healthy individuals identifies B-VIREM cells.

   [78]Fig. 1.
   [79]Open in a new tab

   (A) Immunofluorescence staining of peripheral blood mononuclear cells
   (PBMCs) using antibodies directed against CD14, cyCD79a, and Igκ and
   Igλ light chains presented in individual fluorescence channels. (B)
   Merged image of (A); boxed areas relate to images of interest and are
   presented in more detail; arrow positions as in (A). (C) Columns relate
   to the four fluorescent biomarkers used. Rightmost column taken to
   boxed areas in (B); rows 1 and 2 show antibody-expressing B-VIREMs
   highlighted by arrows and dashed lined across. Rows 3 and 4, CD14^+
   monocytic cells; row 5, cyCD79a^+Igλ^+ B cells and cyCD79a^+Igκ^+ B
   cell; row 6, CD14 ^+ monocyte and cyCD79a^+Igκ^+ B cells. Scale bars,
   20 μm.

   Leukocyte populations and B-VIREM cells were sorted from peripheral
   blood samples of healthy individuals and analyzed for their lineage
   affiliation using a genome-wide 850k methylome array defining
   differentiation-dependent intragenic and intergenic methylation
   patterns ([80]Fig. 2A). This analysis reveals that B-VIREM cells
   differentiate within the myeloid lineage showing closest kinship to
   monocytes followed by granulocytes and are developmentally not related
   to lymphocytes.

Fig. 2. Single-cell mRNA analyses reveal B-VIREM cell populations in PBMCs.

   [81]Fig. 2.
   [82]Open in a new tab

   (A) Methylome analyses of fluorescence-activated cell sorting–sorted
   leukocytes from six healthy donors. A total of 10,652 DMRs from all
   possible pairwise comparisons between four cell populations (B cells, T
   cells, granulocytes, and monocytes) were used to cluster blood cell
   populations demonstrating that B-VIREMs are of closest kinship to
   monocytes. (B) Single-cell transcriptomic data of PBMCs (n = 4) were
   classified by V(D)J profiling. Monocytes/macrophages and B-VIREMs were
   integrated and subclustered using uniform manifold approximation and
   projection (UMAP) from clusters 0 to 3 (classical monocytes) and
   clusters 15, 16, and 17 (potential B-VIREM). Feature plots illustrate
   the expression of myeloid markers and Ig heavy- and light-chain
   expression. (C) Pathway enrichment analysis of B-VIREM cluster 15
   compared to monocyte cluster 0. The top 20 differentially expressed
   Gene Ontology (GO) terms (P < 0.001) are shown. MHC, major
   histocompatibility complex. (D) Left: Single-cell analysis of B-VIREM
   in healthy controls (HCs) and in patients with coronavirus disease 2019
   (COVID-19) from early-recovery stage (ERS) and late-recovery stage
   (LRS) in technical quadruplicates. Right: Ig heavy-chain clonotypes of
   antibody-expressing B-VIREM cells. The dark pie sections represent
   expanded clonotypes with numbers specifying factor of expansion. Total
   number of B-VIREM cells in each cell group: HC, 431; ERS, 1400; LRS,
   667. For distribution of clonotypes, see table S2.

   Publicly available single-cell data of PBMCs from healthy donors
   allowed to cluster the RNA sequencing [single-cell RNA (scRNA)]
   profiles of all monocytic cells in the datasets. Subsequent V(D)J
   profiling of the scRNA profiles revealed three putative B-VIREM
   clusters (clusters 15, 16, and 17; [83]Fig. 2B). Sample bias led us to
   exclude clusters 4, 16, and 17 from further analyses (fig. S2A) and
   further analyze cluster 15 accounting for approximately 0.8% of the
   cells. Gene Ontology analysis revealed that, together with classical
   innate pathways like phagocytosis and innate immune response, B-VIREMs
   showed marked enrichment of prominent adaptive immune pathways when
   compared to classical monocytes (clusters 0) ([84]Fig. 2C). For
   example, the B cell receptor (BCR) signaling pathway was found to be
   most prominently activated, and alternatively spliced mRNA species for
   soluble and for membrane-bound Ig were identified in cluster 15,
   suggesting BCR expression. In line with this, B-VIREM and B cells
   express comparable ratios of membrane Ig, while plasma cells
   predominantly express soluble Ig (fig. S2B). In addition, different Ig
   isotypes appear to be expressed by B-VIREMs (fig. S2C). B-VIREMs show
   an approximately threefold higher rate of cells with unproductive Ig
   rearrangements compared to circulating B cells (fig. S2D). Together,
   the immunocytochemical phenotype, the transcriptional profile and
   pathway analyses, and the genome-wide methylome analyses
   unidirectionally point to the myeloid nature of B-VIREMs.

   To address the question about potential dynamic changes of blood-borne
   B-VIREM in different health conditions, we analyzed the V(D)J
   single-cell datasets of a study of 14 PBMC samples (in technical
   quadruplicates) from patients with coronavirus disease 2019 (COVID-19)
   and from healthy individuals [healthy control (HC)]. The COVID-19
   samples were taken in their early-recovery stage (ERS) and
   late-recovery stage (LRS), i.e., during the clinical virus-positive
   phase and the polymerase chain reaction (PCR)–negative post–COVID-19
   phase, respectively ([85]Fig. 2D) ([86]24). From the HC, ERS, and LRS
   samples, we extracted full-length sequences of corresponding heavy and
   light chains of Igs coding for 116, 131, and 133 bona fide intact
   antibody clonotypes, respectively (table S2). The pie charts in
   [87]Fig. 2D show that, of the total of 380 clonotypes, 29 showed
   different degrees of clonal expansions. These were most pronounced in
   the ERS stage with up to 362 cells in one sample expressing the
   identical antibody, while the HC group showed significantly fewer and
   lower-degree clonotype expansions (ERS versus HC, P = 0.033, Wilcoxon
   test). When clustering the B-VIREM clones, we observed two clusters not
   discriminatory for the HC, ERS, or LRS, while the expanded clusters
   coincided with the health conditions ([88]Fig. 2D, left).

Self-arming of FcR CD32A by endogenous antibody expression

   We have coexpressed an FcγRIIA (CD32) fused to orange fluorescent
   protein (OFP; CD32-OFP) with an IgG-Fc domain fused to the C terminus
   of enhanced green fluorescent protein (eGFP) (eGFP-Fc) in human
   embryonic kidney–293 (HEK293) cells to investigate whether B-VIREM
   cells will self-arm their membrane FcRs with endogenously expressed Ig
   as suggested by the immunocytochemical images of the cells ([89]Fig.
   3A). [90]Figure 3B shows cellular sections with vesicular transport of
   both molecules being simultaneously shuttled to the cell surface. Live
   imaging of several independent experiments with double transfectants
   demonstrated that around 20% of the vesicles in transit to the cell
   membrane were double positive ([91]Fig. 3B, movies S1 and S2, and fig.
   S3D). Expression of CD32 and the functionality to bind to Ig-Fc domains
   were effective in HEK293 + CD32-OFP cells (fig. S3A). In
   double-transfected HEK293 cells (CD32-OFP^+/eGFP-Fc^+), the soluble
   eGFP-Fc fusion protein was not displaced from the surface FcR by
   polyclonal human IgG even at normal serum concentration (fig. S3B). In
   contrast, when CD32-OFP–transfected HEK293 cells were offered the
   eGFP-Fc fusion protein externally, its binding to the surface was
   efficiently competed for by increasing concentrations of polyclonal IgG
   as shown by diminishing green fluorescence staining (fig. S3B). These
   results demonstrate that coexpression of Ig and CD32 in cells
   simultaneously expressing both molecules allows efficient self-arming
   of the membrane FcR with monoclonal antibody and protects from
   displacement by external Ig. The results also indicate that, in the
   extravascular space, where IgG concentrations are approximately 13-fold
   lower, monoclonal antibody surface signatures will be stable on
   B-VIREMs (fig. S3C) ([92]23).

Fig. 3. Simultaneously expressed Ig and FcR colocalize during intracellular
vesicle transport and on the cell surface.

   [93]Fig. 3.
   [94]Open in a new tab

   (A) Fluorescence microscopy of human embryonic kidney–293 (HEK293)
   cells transfected with CD32–orange fluorescent protein (OFP) (top row),
   enhanced green fluorescent protein (eGFP)–Fc (middle row), and CD32-OFP
   + eGFP-Fc (bottom row) as indicated. Nuclear staining of ells
   previously grown on coverslips was done with 300 μM
   4′,6-diamidino-2-phenylindole (DAPI). Single and merged fluorescence
   images as indicated. Scale bars, 50 μm. (B) Snapshot of the live-cell
   imaging video sequence of HEK293 cotransfected with CD32-OFP and
   eGFP-Fc. Double-fluorescent vesicles are highlighted by arrowheads in
   the cytoplasm, in the vicinity to the cell membrane and on the surface.
   Video sequences show intracellular transport of both proteins with 20%
   of mixed-fluorescence vesicles (movies S1 and S2; see also the
   Supplementary Materials). Scale bars, 5 μm.

Identification of stomatin as the specific antigen of Vab1

   As specific immunorecognition appears to be a key role in B-VIREMs, we
   have expressed the first myeloid-derived antibody reported ([95]18) and
   studied its specificity to approach the cells’ biological functions. On
   the basis of a total of 24 immunoprecipitation experiments from human
   cell line extracts, PBMCs, and human erythrocytes followed by protein
   mass spectrometry (MS), we have identified stomatin, a strictly
   intracellular protein as the antigen specific for Vab1 in UniProt
   (table S3) ([96]25, [97]26). Western blot analysis of erythrocyte
   membrane ghost preparations immunoprecipitated by Vab1 and the
   commercially available monoclonal anti-stomatin antibody clone E5
   demonstrate a protein band of 35 kDa corresponding to the molecular
   mass of monomeric stomatin and a high–molecular weight complex
   ([98]Fig. 4A and fig. S4). Vab1 also identifies a recombinant
   stomatin–glutathione S-transferase fusion protein and depletes it from
   solution, while control IgG does not ([99]Fig. 4B). As assessed by MS,
   72 ± 5% of the proteins precipitated with both monoclonal antibodies
   were identical, and two-thirds are known to complex with stomatin in
   erythrocytes (table S4). Together, the data obtained by independent
   methods demonstrate that the monoclonal B-VIREM antibody Vab1 is
   specific for the intracellular scaffold protein stomatin.

Fig. 4. B-VIREM antibody identifies human stomatin as specific target by
Western blotting.

   [100]Fig. 4.
   [101]Open in a new tab

   (A) Antigens were immunoprecipitated from erythrocyte membrane ghost
   preparations using Vab1, murine monoclonal anti-stomatin, and murine
   antibody isotype control as indicated. Left: Staining with murine
   anti-stomatin and horseradish peroxidase (HRP)–coupled rabbit
   anti-mouse antibody; the 35-kDa band corresponds to monomeric human
   stomatin; red arrows mark high–molecular weight moieties suggesting
   membrane complexes containing stomatin. Right: Staining with
   HRP-coupled rabbit anti-mouse antibody to detect unspecific binding;
   heavy and light chains detected by secondary antibodies are indicated.
   (B) Recombinant stomatin–glutathione S-transferase (GST) fusion protein
   (predicted molecular mass of 62 kDa) immunoprecipitated with Vab1 and
   human isotype control antibody followed by staining with murine
   anti-stomatin and HRP-coupled rabbit anti-mouse antibody. Left:
   Recombinant stomatin–GST (65 ng of stomatin). Right: Different amounts
   of stomatin were used for precipitations [lanes 1, 3, 4, and 6:
   stomatin (9 μg/ml); lanes 2 and 5: stomatin (4.5 μg/ml)]. Lanes 1 to 3,
   immunoprecipitates; lanes 4 to 6, supernatants of immune-depleted
   stomatin samples; lane 6, stomatin not precipitated from the solution
   by the human IgG1κ isotype control antibody.

Mediation of phagocytosis by the monoclonal B-VIREM antibody Vab1

   We first confirmed that phagocyte cell line HL-60 predominantly
   expresses FcγRIIA (CD32) (fig. S5A). These cells can be successfully
   stripped and reloaded with polyclonal serum or monoclonal antibodies
   (fig. S5B). Opsonophagocytosis by HL-60 cells is efficient for
   serum-coated latex microbeads, while plain beads or bovine serum
   albumin (BSA)–loaded particles are not engulfed ([102]Fig. 5A). HL-60
   phagocytes preloaded with polyclonal anti-stomatin or with monoclonal
   Vab1 efficiently phagocytosed latex beads coated with a truncated
   recombinant stomatin lacking the N-terminal 55 amino acids ([103]Fig.
   5B). In contrast, the commercial monoclonal antibody clone E5 directed
   against the N-terminal stomatin domain does not provoke phagocytosis of
   the truncated stomatin demonstrating that the Vab1-specific epitope is
   not located in the N-terminal domain. HL-60 armed with the monoclonal
   anti–carcinoembryonic antigen (CEA) T84.1 antibody phagocytosed
   CEA-coated beads, while none of the other antibody-bearing cells did.
   Similarly, CD14^+ cell preparations from five different healthy
   individuals previously loaded with anti-stomatin engulf stomatin-coated
   beads, while unloaded cells do not (fig. S5C). This confirms that the
   cells armed via their FcR do not require additional opsonization of
   antigen-positive surfaces. When using carboxyfluorescein diacetate
   succinimidyl ester–labeled erythrocyte membrane ghosts as a
   physiological phagocytosis substrate, Vab1-loaded HL-60 showed a
   2.5-fold higher activity over unloaded cells ([104]Fig. 5C). In
   contrast, phagocytosis of intact erythrocytes by anti-stomatin–loaded
   cells is not successful (fig. S5D).

Fig. 5. Phagocytosis of latex microbeads and erythrocyte membrane ghosts by
plain and antibody-loaded myeloid cells.

   [105]Fig. 5.
   [106]Open in a new tab

   Each experiment used 10^5 HL-60 phagocytes co-incubated overnight (o/n)
   at 37°C with latex beads (yellow-green fluorescence, 1-μm diameter) or
   with freshly prepared carboxyfluorescein diacetate succinimidyl
   ester–labeled erythrocyte ghosts. Results are presented as %
   phagocytosing cells and given as means with SD. (A) Coating of beads
   o/n at 4°C: PBS (uncoated), BSA (900 nM BSA), and serum (20 μg human
   serum). Plain or preloaded HL-60 phagocytes were used as indicated. *P
   < 0.01 (B) Coating of beads o/n at 4°C: PBS (uncoated), CEA (900 nM
   CEA), and truncated stomatin (900 nM recombinant stomatin lacking the
   55 N-terminal amino acids). Preloading of HL-60 phagocytes with each
   antibody (2 μg/10^5 cells) as indicated; *P < 0.05. (C) Phagocytosis of
   erythrocyte membrane ghosts by HL60 cells. Left: Images of HL-60 cells
   preloaded with Vab-1 (2 μg/10^5 cells, top) or unloaded (PBS, bottom).
   Cells are counterstained with DAPI; magnification, ×40. Scale bars, 20
   μm. Right: Results from three independent experiments; *P < 0.05.

DISCUSSION

   A central, long-standing paradigm in immunology posits that the
   strategies and mechanisms of immunorecognition used by the innate
   system and the adaptive system are fundamentally different and mutually
   exclusive, leading to a strict functional separation of the two arms.
   Specifically, the phylogenetically older innate immune system relies on
   genome-encoded preconfigured and invariant PRRs. In contrast, the
   younger adaptive system generates enormous repertoires of variable
   immunoreceptors in the form of antibodies and TCRs by stochastic
   recombination of independent genomic building blocks with subsequent
   antigen-driven clonal selections and refinement of their binding
   properties. Functionally, innate immune cells use a fast-responding
   recognition system with low antigen resolution, while the lymphocytes
   of the adaptive system provide antigen recognition at extremely high
   molecular resolution at the cost of a delayed immune response time for
   their generation and maturation. A very limited number of studies have
   reported partial sequences of isolated recombined Ig gene transcripts
   in nonlymphoid cells and cancer cell lines, mostly based on PCR
   amplification results ([107]15, [108]16, [109]18, [110]27). Using a
   self-developed single-cell technology, we have previously reported the
   first full-length light- and heavy-chain variable region genes coding
   for a bona fide intact monoclonal antibody obtained from a
   tumor-associated macrophage ([111]18). However, phenotype and progeny,
   functionality, and biological significance of these cells have remained
   undetermined until now. Our report establishes a class of innate cells
   designated VIREMs that provide a link between the recognition functions
   of the innate and adaptive arms of the immune system.

   The visualization of antibody-expressing VIREM (B-VIREM) by
   fluorescence microscopy now establishes these cells as a regular
   component in the blood of healthy individuals. The B-VIREM phenotype is
   easily distinguished from antibody-expressing B cells and from regular
   CD14^+ monocytes by its additional monoclonal antibody cell surface
   signature exhibiting either kappa or lambda light-chain expression. We
   found approximately 1.7% B-VIREM cells among the CD14^+ monocytes
   ([112]Fig. 1B; fig. S1, B and C; and table S1). The investigation of
   leukocytes from healthy individuals using intra- and intergenic
   genome-wide methylome analysis that differentiates cell lineages of
   hematopoiesis ([113]28) establishes B-VIREM cells as myeloid leukocytes
   with closest kinship to monocytic cells but unrelated to lymphocytes
   ([114]Fig. 2A). The combined single-cell mRNA and V(D)J profiling data
   from four healthy donors show that B-VIREMs account for well-defined
   clusters among monocytes ([115]Fig. 2B).

   Gene Ontology pathway analyses show key B cell pathways being activated
   in B-VIREM in contrast to regular monocytes ([116]Fig. 2C) consistent
   with the immunofluorescence and molecular data. For example, our data
   support the notion that B-VIREM will use B lineage markers associated
   with the antibody expression function. For example, our earlier data
   had shown that B-VIREM can express CD79A at the mRNA level ([117]18),
   which is now confirmed. Specifically, fig. S1D shows that overexposure
   of the PBMC fluorescence staining in [118]Fig. 1A demonstrates faint
   but discernable positivity in the antibody-expressing CD14^+ monocytes,
   i.e., B-VIREMs. CD79A is an accessory molecule involved in signaling of
   the BCR until now associated with B cells. B-VIREMs express
   alternatively spliced Ig heavy-chain transcripts comprising the
   membrane exon required for the BCR. The proportion of mRNAs coding for
   the membrane and secreted antibody moieties is comparable to that of B
   cells. By comparison, plasma cells only show low expression of
   membrane-bound Ig heavy chain (fig. S2B), as expected.

   One central question relates to the potential dynamics of B-VIREM in
   response to biological stimuli in health and disease. As a first
   approximation, we used uniform manifold approximation and projection
   (UMAP) clustering of independently published single-cell transcriptomic
   data obtained from peripheral blood samples of healthy individuals (HC)
   or patients in ERS and LRS of COVID-19. The samples show B-VIREM
   clusters that are indiscriminate for the three cohorts, while others
   display a patient- and stage-dependent differential clustering
   ([119]Fig. 2D). Together, the B-VIREMs expressing full-length
   corresponding light- and heavy-chain genes of bona fide intact
   antibodies represent 380 different clonotypes (table S2), 29 of which
   were clonally expanded to different degrees as shown in the pie charts
   in [120]Fig. 2D, e.g., with clonotypes of up to 362 cells expressing
   the same antibody. Both the number and the degree of the expanded
   clonotypes markedly differed in the three groups and were significantly
   pronounced in patients particularly during the ERS of the disease as
   compared the HCs. It appears highly likely that the observed expansions
   of B-VIREM clones are reactions to unknown biological and, in this
   COVID-19 study, disease-related stimuli encountered by these cells
   eventually that caused their dynamic skewing of B-VIREM repertoires
   seen in the blood. Considering that B-VIREMs are regular constituents
   in the blood of healthy individuals, this observation strongly suggests
   an important biological role of these cells in health and disease.

   Ig expression in monocyte-derived cells raises questions about the
   mechanism enabling the monoclonal cell surface signature and the
   physiological function of B-VIREM. To investigate this, we resorted to
   a double-transfectant model of HEK293 cells allowing us to track OFP
   CD32 regularly expressed in monocytes and macrophages ([121]22,
   [122]29, [123]30) without interference by a “cold” endogenous FcR. In
   contrast, the use of the THP-1 macrophage cell line was not feasible,
   because THP-1 would dilute the OFP-CD32 signal due to their endogenous
   FcR CD32, and they were also not stably transfected in our hands.
   Furthermore, we required an adherent cell system for easier live-cell
   imaging studies. Coexpression of the FcR CD32 and its Fc domain ligand
   in the double transfectants lead to coordinated transport of both
   molecules to the cell surface. Specifically, around 20% of vesicles
   display mixed fluorescence due to colocalization of the proteins
   ([124]Fig. 3B and movies S1 and S2). These data explain the monoclonal
   surface signature of the B-VIREM phenotype as a result of CD32 occupied
   by their endogenous antibody Fc domain upon arrival at the surface. It
   is noteworthy that the monoclonal signatures are stable against the
   high concentrations of polyclonal serum IgG encountered in peripheral
   blood and, predictably, in the lower IgG concentration conditions
   within the extravascular space. This stability would be an important
   contributor for efficient opsonization-independent immunorecognition
   and the triggering of biological effector functions in this new class
   of innate immune cells. We hypothesize that the determination of
   specificities of B-VIREM antibodies is a key to approaching the
   functional roles of these cells. After recombinantly expressing the
   first functional B-VIREM–derived antibody (Vab1) recently identified in
   a tumor-associated macrophage ([125]18), we found that Vab1
   specifically recognizes stomatin, a strictly intracellular
   membrane-associated autoantigen that is ubiquitously expressed
   ([126]Fig. 4). The specificity of Vab1 for stomatin was extensively
   confirmed. All commercial stomatin antibodies used in this study
   immunoprecipitated the antigen also as part of a high–molecular weight
   complex. This is consistent with its organization in a 9- to 12-mer
   oligomeric form as part of a multiprotein structure within lipid rafts
   ([127]Fig. 4A and table S4) ([128]25, [129]26, [130]31).

   Phagocytes can identify targets in an antigen-specific fashion only, if
   these targets are opsonized by antibodies, which, in turn, require a
   paradigmatic preexisting B cell response ([131]19). Antibody-decorated
   targets are subsequently bound by the cell’s membrane FcRs triggering
   opsonophagocytosis of the target. Our data strongly suggest a
   profoundly different mechanism for B-VIREM immunorecognition.
   Specifically, the endogenous concomitant expression of a monoclonal
   antibody and FcR leads to stable self-armament even during vesicle
   transport to the cell surface, thereby enabling the cell to exert
   specific binding to antigen-positive targets in an
   opsonization-independent fashion. In the case of Vab1, we demonstrate
   that the antibody-loaded cells will recognize particle matter, i.e.,
   polystyrene and latex beads as well as membrane ghosts of destroyed
   erythrocytes known to be abundant with stomatin ([132]32). For these
   experiments, we used HL-60 for its phagocytic capacity, although the
   cell line shows a more granulocyte-like differentiation ([133]33,
   [134]34).

   Specific immunorecognition of the strictly intracellular stomatin by a
   specific B-VIREM may allow antigen-specific “injury sensing” to trigger
   the clearance of damaged cells ([135]7, [136]8) without the need of
   prior B cell immunity. We speculate that such a B-VIREM function would
   aid in low-threshold restoration of tissue homeostasis and repair. We
   consider B-VIREM a prveiously unknown class of “smart phagocytes”
   capable of linking antigen-specific immunorecognition with innate
   biological effector functions parallel to, and even independent of, the
   lymphoid system ([137]Fig. 6). Future studies need to investigate the
   spectrum of B-VIREM immunorecognition in detail. For example, this
   function may be part of an armory of tools required by complex
   organisms to maintain their bodily integrity at a low immunological
   activity threshold and contribute to the phagocytic code very recently
   reviewed by Cockram and colleagues ([138]35). It is intriguing that,
   very recently, a concept called “normality sensing” has been proposed
   for tissue surveillance as a cellular immune function of γδT cells
   ([139]36). Injury sensing could be a complementary innate function.
   Closely related to monocytes/macrophages, these cells designated
   B-VIREM are predicted to be capable of the antigen-specific
   immunorecognition of the B cell system combined with the instantaneous
   effector capabilities of innate effector functions.

Fig. 6. Schematic representation of antigen-specific immunorecognition and
phagocytosis in myeloid cells.

   [140]Fig. 6.
   [141]Open in a new tab

   (Left) In a B-VIREM cell, the endogenous antibody and FcR are
   coexpressed and colocalize during transport to the cell surface leading
   to armed FcR. The specific antigen is directly recognized without
   requiring prior opsonization leading to direct phagocytosis. (Right)
   Only FcRs are expressed, and receptors transit to the cell membrane. To
   recognize immunogenic particles, opsonization with B cell–dependent Ig
   is required to cluster FcRs and leads to activation of phagocytosis of
   the targeted particles.

MATERIALS AND METHODS

Patient samples

   Tissue samples were obtained from the University Medicine Mannheim
   Biobank archive with written informed consent from the patients. Human
   buffy coats were provided by the German Red Cross Blood Donor Service
   Baden-Wuerttemberg Hessen. All blood specimens were anonymized. The
   study was approved by the local Ethics Committee II at the Medical
   Faculty Mannheim of Heidelberg University (2017-548 N-MA and 2014-562
   N-MA).

Cell culture and PBMC isolation

   Human HL-60 (RRID:CVCL_0002) were cultured in RPMI (Merck, Burlington,
   USA) supplemented with 10% fetal bovine serum (FBS) (FBS.S0615;
   Bio&SELL) or X-VIVO-15 (Lonza). HEK293 cells (RRID:CVCL 0045) were
   cultured in Dulbecco’s modified Eagle’s medium (Thermo Fisher
   Scientific) supplemented with 10% FBS. Cells were cultured at 37°C, 5%
   pCO[2], and >95% relative humidity and passaged two to three times per
   week to a maximum of 25 times. No used cell line is listed as “commonly
   misidentified cell line” by the International Cell Line Authentication
   Committee. In addition, cell lines were authenticated using the
   Multiplexion Multiplex Cell Authentication and Contamination tests
   ([142]37, [143]38). PBMC isolation was performed by gradient
   centrifugation (Ficoll-Paque PLUS) of human buffy coats. Erythrocytes
   were lysed by hypotonic shock, and the cell preparations were routinely
   counted in a Sysmex XP-300 hematocytometer (Sysmex).

Immunofluorescence

   PBMCs were isolated from heparinized peripheral blood samples of
   healthy volunteers, and ~5 × 10^5 PBMCs were used for each staining.
   The cells were prepared for cytoplasmic staining with FIX&PERM
   (Nordic-MUbio) according to the manufacturer’s protocol. For staining
   of the permeabilized cells, the following antibodies were used:
   CD14-BV421 (clone M5E2; BD Biosciences), the multiMix dual-color
   reagent (kappa light chain–fluorescein isothiocyanate +
   lambda-phycoerythrin; Agilent-Dako), and cyCD79a-allophycocyanin (clone
   HM57; Agilent-Dako). The stained cells were mixed with heat-inactivated
   human serum at a 1:1 ratio, and cytospins (40g for 10 min) were
   prepared and analyzed on the Zeiss Axio Observer Z1.

Single-cell sequencing and analysis

   We obtained single-cell gene expression and V(D)J immune profiling
   datasets for two HC samples (denoted HC1 and HC2) with four technical
   replicates each ([144]24). In addition, two PBMC data samples from
   healthy human donors (denoted 10X_10k and 10X_20k were downloaded from
   10X Genomics ([145]https://support.10xgenomics.com). To address
   technical artifacts of microfluidics-based systems and to reduce the
   risk of potential multiplet cell bias, we used four different detection
   tools: Scrublet ([146]39), DoubletDetection ([147]40), DoubletFinder
   ([148]41), and SCDS ([149]42). We excluded all cells, which were
   concordantly predicted as doublets by at least three prediction tools.

   We used the Seurat R package (version 4.0.3) to analyze single-cell
   gene expression datasets ([150]43). For quality control, we eliminated
   genes, which were detected in less than three cells and cells with less
   than 200 genes with nonzero counts. We further excluded all cells
   having more than 20% of mitochondrial gene counts.

   Ig-positive cells were determined by integrating the V(D)J immune
   profiling datasets. In detail, we filtered for high-confidence,
   full-length, and productive Ig contigs and identified unique cells as
   V(D)J-positive cells if they expressed at least one Ig heavy chain
   (IGH) and at least one IG light chain (IGkappa or IGlambda). The four
   datasets were integrated and subsequently clustered using the UMAP
   method. Monocyte and macrophage clusters were identified using the
   expression profile of the markers CD14, LYZ, S100A8, and S100A9. A
   subset of all V(D)J-positive cells was extracted and reclustered in a
   separate analysis. Here, potential B-VIREM cells were identified using
   the expression level of the four monocytic markers and the formation of
   a distinct cluster. Last, monocytes and macrophages as well as B-VIREM
   were integrated and reclustered. Pathway enrichment analysis was
   performed using the TopGO R package ([151]44), and cell type prediction
   was done with celldex ([152]45).

Clonotype expansion analysis

   For the B-VIREM clonotype analysis, we accessed the complete
   single-cell gene expression and V(D)J immune profiling dataset from Wen
   et al. ([153]24). For every patient group (ERS, LRS, and HC), we
   integrated the respective patient datasets (in quadruplicates) and
   identified B-VIREM as described above. Within these B-VIREMs,
   clonotypes for both heavy and light chains were identified by their
   full-length amino acid sequence according to their V(D)J data.

Analysis of IGHM splice variants

   To distinguish the secreted and membrane-bound IGHM transcript variants
   in B-VIREM cells, we investigated the expression of specific IGHM
   splice forms on the single-cell level. The DNA sequences of the IGHM
   exon 4, the CHS region, and the M1 region were downloaded from
   [154]imgt.org ([155]46). The sequences of exon 4 and the CHS region
   representing the secreted form as well as exon 4 and M1 region
   expressing the membrane-bound variant were joined. In addition, the M1
   region and the adjacent intron were combined to represent the pre-mRNA.
   Cellranger mkref was used to build a custom reference package. FASTQ
   raw reads were mapped on the custom reference by using the CellRanger
   count command. In the resulting bam file, reads overlapping the
   interface of exon 4 and CHS region and of exon 4 and M1 region at
   position 330 were counted with the SAMtools view command. For the
   intron–M1 region, all reads covering position 490 were considered.
   Barcodes to identify specific cells were extracted from the CR tag of
   the reads header and summarized per sample.

Analysis of productive and unproductive V(D)J antibody chains

   We evaluated the BCR-V(D)J data from Wen et al. ([156]24) and extracted
   all antibody chains that were positive for all of the binary attributes
   “is_cell,” “full_length,” “productive,” and “high_confidence.” Antibody
   chains that missed at least one criterion were considered as
   unproductive. Then, all cells having at least one productive heavy
   chain (IGH) and one productive light chain (IGL or IGK) were filtered.
   All cells missing this condition were regarded as cells expressing
   nonproductive antibodies.

Methylome analyses

   Lymphocytes, monocytes, granulocytes, and B-VIREM populations were
   fluorescence-activated cell sorting–sorted from peripheral blood
   samples of six healthy donors using the BD FACSMelody cell sorter (BD
   Biosciences). Lymphocytes and monocyte populations were sorted from
   PBMCs isolated by density gradient centrifugation (Ficoll-Paque PLUS),
   and granulocytes were sorted from the pellet sediment after Ficoll
   gradient centrifugation. Forward scatter (FSC) and sideward scatter
   (SSC) singlets were gated, and consecutive gating steps were applied to
   generate highly pure cell populations. The following sorting strategy
   and markers were used: B cells, CD14^−CD19^+; T cells, CD14^−CD2^+;
   monocytes, CD19^−CD14^+; B-VIREMs, CD19^− CD14^+Igκ^+Igλ^− or
   CD19^−CD14^+Igλ^+Igκ^−; and granulocytes, CD14^−CD15^+. After sorting,
   DNA was isolated using the QIAmp DNA Micro Kit (Qiagen) following the
   manufacturer’s protocol. The samples were then subjected to total
   methylome analysis using the Infinium Methylation EPIC BeadChip 850K
   array (Illumina). The work was carried out by the DKFZ Genomics and
   Proteomics Core Facility of the German Cancer Research Center,
   Heidelberg.

Illumina infinium human methylation EPIC 850k bead array analysis

   Raw intensity files were obtained using minfi package [Minfi 1.34.0;
   ([157]47)] to calculate methylation ratios and the log[2] methylation
   ratios (Beta values and M values, respectively). The data were
   normalized using Illumina preprocessing method implemented in minfi.
   Several quality control measures were applied to check for arrays with
   low quality. Median methylated and unmethylated signals were calculated
   for each array; detection P < 0.05 and badSampleCutoff = 10.5 were used
   to remove low-quality arrays. Probes that failed in one or more samples
   or contained an annotated single-nucleotide polymorphism at the
   single-base extension or CpG sites were removed (95,170 probes
   removed). Minfi 1.34.0 was used. Bump hunting method was applied to
   identify differentially methylated regions (DMRs) in EPIC 850k array
   [Bumphunter 1.30.0; ([158]48)]. A maximum of 30,000 candidate bumps
   were used to determine the cutoff when finding DMRs (2.1 used with M
   values as input). Statistical significance was assigned by
   bootstrapping, and the familywise error rate (FWER)–adjusted P value
   cutoff used for downstream analysis was 0.05. To assign closest kinship
   of B-VIREM cells to normal blood cell populations, mean methylation
   value of each normal blood cell epigenetic signature (10,652 DMRs) for
   B-VIREM/normal blood cells (methylation profile) was retrieved. hclust
   function with ward method in R was used to generate the cluster
   dendrogram analysis.

Cloning

   The eGFP–open reading frame was PCR-amplified with primers including
   endonuclease cleavage sites for Eco RV
   (5′-agcttgcgcgatatcagtgagcaagggcgaggag-3′) and Nco I
   (5′-tattgcaccatggacttgtacagctcgtccatgc-3′) and removing start and stop
   codons using the pEGFP-N1 (Clontech, Takara Bio) plasmid as template.
   For subcloning, the pFUSE-hIgG1-Fc2 (pfuse-hg1fc2; InvivoGen) plasmid
   and the purified amplicon were both digested with Eco RV and Nco I for
   2 hours at 37°C, following heat inactivation for 20 min at 80°C.
   Ligation was carried out by the Quick Ligation Kit (vector, 50 ng;
   insert, 75 ng; volume, 20 μl) as suggested by the manufacturer [New
   England Biolabs (NEB)]. The ligation mix was used to transform OneShot
   TOP10 cells (Thermo Fisher Scientific), and bacteria were plated on
   agar plates containing zeocin (50 μg/ml) and incubated overnight at
   37°C. Several clones were picked, expanded, and isolated with the
   QIAprep Spin Miniprep Kit (Qiagen). Recombinant clones were identified
   by double restriction with Eco RI and Bgl II, and the integrity of
   their inserts was verified by DNA sequencing (Eurofins Genomics).

Transfection and selection

   For transfection of HEK293 cells, endotoxin-free plasmid DNA was
   linearized using Ase I (pEGFP-N1), Not I (eGFP-Fc), or Mfe I (pCD32-OFP
   (pCMV3-CD32a-OFPSpark, SinoBiological) and purified. HEK293 cells were
   transfected using the SF Cell Line 4D-NucleofectorTM X Kit L and the
   4D-NucleofectorTM X Unit (Lonza) according to the manufacturer’s cell
   line protocol. Twenty-four hours after transfection, the medium was
   exchanged to medium supplemented with geneticin (1 mg/ml; pEGFP-N1),
   zeocin (100 μg/ml; eGFP-Fc), or hygromycin B (50 μg/ml; pCD32-OFP).
   Transfected HEK293 cells were analyzed by microscopy (TCS SP8 DLS Leica
   inverted Microscope DMi9 CS Bino DLS; Leica Microsystems). Live-cell
   imaging of transfected HEK293 cells was performed using a Leica TCS SP8
   confocal microscope equipped with HCPL APO 20×/0.75 IMM CORR objective
   with 488- and 561-nm lasers.

Preparation of erythrocyte ghosts and cell/tissue lysates

   We prepared erythrocyte membrane ghosts as reported previously
   ([159]49). In short, after Ficoll gradient centrifugation, erythrocytes
   were resuspended in 50 ml of washing buffer (isotonic NaCl 0.9% in 10
   mM tris-HCl and 0.1 mM EGTA) and centrifuged at 1000g for 10 min at
   15°C. After washing twice, erythrocytes were lysed in 37 ml of ice-cold
   sodium Hepes buffer [2.5 mM Hepes (pH 7.4) adjusted with NaOH and 0.1
   mM EGTA] and centrifuged at 40,000g for 70 min at 4°C. The preparations
   were washed twice in 50 ml of phosphate-buffered saline (PBS) and
   centrifuged at 1000g for 10 min at 15°C, and the erythrocyte ghosts
   were resuspended in 1 ml of PBS and filtered through a 40-μm filter.

Generation of recombinant antibodies

   The B-VIREM antibody Vab1 was recombinantly expressed on the basis of
   cloned variable heavy chain (VH) and variable light chain (VL)
   sequences ([160]18) in HEK293 cells as a IgG1/κ full-length Ig and
   provided as lyophilized powder (Syd Labs Inc.).

Immunoprecipitation

   Total protein lysates [n = 24: PBMCs (n = 10), erythrocytes (n = 6),
   HepG2 (n = 2), colon tumor tissue (n = 2), normal colon tissue (n = 2),
   and esophagus tumor tissue and normal esophagus tissue (both n = 1)]
   were prepared in radioimmunoprecipitation assay buffer. To desalt,
   buffer was exchanged by 30 min of centrifugation through Vivaspin 6
   column (molecular weight cutoff, 5 kDa; GE Healthcare). To reduce
   nonspecific protein interactions, the lysates were subsequently
   precleared on Pierce Spin columns (Thermo Fisher Scientific) with an
   agarose matrix for 30 min at 4°C. Immunoprecipitations of stomatin were
   performed using 10 to 20 μg of Vab1 or the murine monoclonal
   anti-stomatin antibody (clone E5, Santa Cruz Biotechnology).
   Recombinant human IgG1κ clone (Bio-Rad) and purified murine IgG2bκ
   isotype control (BD Pharmingen) were used as control antibodies.
   Antibodies were diluted to concentrations of 50 μg/ml and incubated
   with protein A–coated magnetic Dynabeads (Thermo Fisher Scientific) for
   30 min at room temperature and were subsequently washed with PBS. The
   desalted and precleared lysates were added to the beads at
   concentrations of 0.33 to 1.25 mg of protein per sample and incubated
   overnight at 4°C under constant movement. Beads were recovered and were
   washed three times in PBS/0.02% Tween 20 (v/v). The precipitated
   proteins were eluted in Laemmli buffer.

Western blotting

   Primary antibodies (rabbit anti-GFP) and murine monoclonal anti-human
   stomatin were used at indicated concentrations overnight at 4°C;
   secondary horseradish peroxidase–conjugated antibodies were incubated
   at indicated concentrations for 2 hours at room temperature.
   Immunoblots were developed using the Amersham ECL Prime Western
   Blotting Detection Reagent (GE Healthcare) by the ChemiDoc XRS+ Imager
   (Bio-Rad).

Protein preparation and MS

   Bands of interest were excised from the SDS–polyacrylamide gel
   electrophoresis separations and prepared for MS analyses. In short, the
   dried peptide samples were redissolved in 0.1% trifluoroacetic acid and
   loaded onto a C18 precolumn (Acclaim; Thermo Fisher Scientific) using
   an RSLCnano HPLC system (Thermo Fisher Scientific), from which they
   were separated with an aqueous-organic gradient using a flow rate of
   300 nl/min. We used electrospray injection into an LTQ Orbitrap XL mass
   spectrometer (Thermo Fisher Scientific) using a Triversa Automate
   (Advion) as ion source. Each scan cycle consisted of one Fourier
   transform mass spectrometry (FTMS) full scan and up to seven ion trap
   mass spectrometry (ITMS)–dependent tandem MS (MS/MS) scans of the seven
   most intense ions. Dynamic exclusion (30 s), mass width [10 parts per
   million (ppm)], and monoisotopic precursor selection were enabled. All
   analyses were performed in positive ion mode. Extracted MS/MS spectra
   were searched against the human Uniprot database using the PEAKS search
   engine (Bioinformatics Solutions Inc.) accepting common variable
   modifications, one missed cleavage site, a peptide tolerance of ±10
   ppm, and MS/MS tolerance of ±0.5 Da. All protein identification
   experiments were carried out using the corresponding decoy database and
   a false discovery rate of 1%. Further details as described elsewhere
   ([161]50).

Phagocytosis assay

   Green-yellow fluorescent latex beads (10^7 beads) with (1 μm; Merck)
   were coated with 900 nM BSA (NEB), 20 μg of human serum obtained from
   healthy individuals, 20 μg of mouse serum (Merck), 900 nM recombinant
   stomatin (Abcam or ProSpec), or 900 nM CEA (provided by J. E. Shively,
   City of Hope National Cancer Center) overnight at 4°C. Subsequently,
   coated beads were centrifuged, washed twice in PBS, resuspended in PBS
   (final concentration: 2.5 × 10^5 beads/μl), and stored in the dark for
   <7 days at 4°C. HL-60 cells (10^6; undifferentiated), HL-60 dimethyl
   sulfoxide (DMSO) (3 day differentiation with 1.3% DMSO), or primary
   CD14^+–magnetic-activated cell sorting (Miltenyi Biotec) isolated
   monocytes were coated with 20 μg of polyclonal rabbit anti-human
   stomatin (Biomol), murine monoclonal anti-human stomatin, and Vab1 or
   anti-CEA antibodies (T84.1- and T84.66-containing cell culture
   supernatants), respectively, for 1 hour at 4°C in PBS. Cells were
   washed twice with PBS. Phagocytosis assay was performed in black
   96-well clear bottom plates (Greiner Bio-One) with serum-free medium,
   beads (7 × 10^6) or erythrocyte ghosts (7 × 10^6), and cells (5 × 10^4)
   added to the plate in final volumes of 180 μl per well followed by
   incubation for 3 hours or overnight at 37°C, 5% pCO[2]. Sytox blue
   (Thermo Fisher Scientific) was added, and phagocytosis was measured by
   flow cytometry using a FACSCanto II (BD Biosciences).

Statistical analysis

   Results represent means ± SD, unless otherwise indicated. Statistical
   significance between groups was determined by unpaired Student’s t test
   or analysis of variance (ANOVA) and Bonferroni or Dunnett’s correction
   for multiple comparisons.

Acknowledgments