Bacterial strains used in this study are listed in Table 1 . F. tularensis LVS was grown on chocolate II agar plates and incubated at 37°C with 5% CO₂ for 2–3 days. Bacteria from the agar plates were then used to inoculate either Tryptic Soy Broth supplemented with 0.1% L-cysteine (TSBc) or Chamberlain’s Chemically Defined Medium (CDM) (Chamberlain, 1965). Cultures were incubated at 37°C with agitation and grown to stationary phase. When indicated, media were supplemented with 3-4% potassium chloride (KCl). Escherichia coli was grown on Luria Bertani (LB) agar plates and incubated at 37°C for 1 day. Broth cultures were grown to stationary phase by inoculating LB broth with E. coli bacteria and incubating overnight at 37°C with agitation. For bacterial strains harboring plasmids, the media was supplemented with kanamycin at 10 μg/ml for F. tularensis LVS or ampicillin at 100 μg/ml for E. coli.

Human erythrocytes were isolated from whole blood obtained from donors by West Liberty University’s Medical Lab Science Department. PBS was added to an equal volume of the blood sample and underlaid with Ficoll to separate erythrocytes from white blood cells using density gradient centrifugation. The plasma was removed, and erythrocytes were washed in McCoy’s 5A medium supplemented with 10% human AB serum and 25 mM HEPES buffer (red blood cell media). Erythrocytes were counted using a hemocytometer and resuspended in red blood cell media to achieve the desired concentration (Horzempa et al., 2011).

Fab2 and Fc fragments were obtained from unconjugated rabbit IgG polyclonal antibody reactive against human Band 3 (Invitrogen, REF: PA5-80030; Novus, REF: NBP1-70433) and human Glycophorin A (CD235a, Thermo Scientific CAT # 85882). The antibodies were subject to cysteine protease digestion (Genovis FragIT, Pr. No: A2-Fr2-005; Genovis FabRICATOR Immobilized CAT # A0-FR6-010) following the manufacturers protocol to generate Fc fragments and Fab2 fragments. To isolate Fab2 fragments, Fc fragments were captured using IgG-Fc magnetic affinity beads and the Fab2 fragments were eluted (CaptureSelect IgG-Fc Magnetic Agarose Beads, Thermo Scientific). The immobilized Fc fragments produced from antibody digestion were also isolated according to the instructions of the manufacturer and were used as controls. Fab2 and Fc fragment concentrations were quantified utilizing the QuickStart Bradford Protein Assay following the manufacturers protocol. Isolated Fab2 and Fc fragments were stored at -20°C until further analysis could be conducted.

All studies using mice were approved by the West Liberty University Animal Care and Use committee and were conducted using ethical standards (OLAW assurance number D21-01113). WB/Re (WB) +/+ and nb/+ inbred mice (JAX stock #000453), 6–8 weeks old, were purchased from The Jackson Laboratory (Bar Harbor, ME). These mice contain a spontaneous mutation in the Ankyrin-1 gene (Ank1) that results in the introduction of a premature stop codon and subsequent truncation of the Ankyrin-1 protein found in erythrocytes (Birkenmeier et al., 2003). The nb mutation was maintained in the heterozygous state in the WB/Re strain. Homozygotes (WB/Re nb/nb) for this study were obtained by crossing WB/Re heterozygotes and were identified by their light orange color and small size compared to wild type and heterozygotes (The Jackson Laboratory). Confirmation of ankyrin deficiency in murine erythrocytes was obtained via western blotting (rabbit anti-Ankyrin 1, Invitrogen CAT # PA5-63372, 1:1000) and Sanger sequencing of the amplicons generated using the AnkF and AnkR primers (data not shown). Controls consisted of normal WB/Re +/+ mice. In this study, 6 WT mice and 3 nb/nb mice were used. Murine blood was obtained by neonatal decapitation with veterinary-grade isoflurane euthanasia from neonatal mice 2–3 days old (Seman et al., 2020). Approximately 150–200 microliters of blood could be obtained from neonates. EDTA (0.5 mM) was used as an anticoagulant. Murine erythrocytes were then purified as previously described (Horzempa et al., 2011).

Erythrocytes isolated from blood samples (as described above) were diluted to a concentration of 1 x 10⁷ cells/ml in red blood cell media and aliquoted into a 96-well V-bottom dish. For experiments involving purified recombinant Band 3 (rBand 3), the bacteria and rBand 3 (10 μg/ml) were incubated for 1 hour at 37 °C and 5% CO₂, after which the bacteria/rBand 3 mixture and erythrocytes were combined and incubated. For assays using Glycophorin A (GPA) and Band 3 Fab2 fragments, erythrocytes were treated with GPA Fab2 fragments (0.2 mg/ml) or Band 3 Fab fragments (2 mg/ml) and incubated for 1 hour at 37°C with 5% CO_2. CDM was inoculated with either F. tularensis LVS, F. tularensis LVSΔgcvT, or F. tularensis LVSΔgcvT/pMB1:gcvT and incubated at 37°C with agitation for at least 16 hours prior to the start of an experiment. The CDM growth medium was used as it promotes erythrocyte invasion while also producing minimal background levels of viable extracellular bacteria after gentamicin treatment (Haggerty, 2025). Bacteria were diluted to a targeted multiplicity of infection (MOI) of 12.5 (1.25 x 10⁸ CFU/mL) in red blood cell media (Schmitt et al., 2017) and were incubated at 37°C with 5% CO₂ for at least 15 minutes. The actual MOI of the experiment was determined by serial diluting and plating for CFU. Bacteria and erythrocytes were incubated for 2–3 hours at 37°C with 5% CO_2. After incubation, erythrocytes were pelleted and washed with PBS. Gentamicin (25 μg/ml in PBS) was pre-warmed and added to all experimental wells. The plate was incubated at 37°C with 5% CO₂ for 45 minutes to 1 hour. Following incubation, the plate was centrifuged at 100 x g for 5 minutes, and the supernatant was discarded. Wells were washed with PBS and erythrocytes were lysed with 0.02% sodium dodecyl sulfate (SDS). Lysates were plated onto chocolate II agar and incubated at 37°C with 5% CO₂ for 3–5 days (Horzempa et al., 2011; Schmitt et al., 2017) to determine CFU. Wells lacking erythrocytes (-RBC) were used to determine the efficacy of the gentamicin in killing F. tularensis LVS which was negligible for all experiments. The average CFU from -RBC wells was considered to be background and this value was subtracted from the experimental groups in each experiment.

Primers and plasmids used in this study are listed in Table 1 . Coding sequence for the cytoplasmic domain of the human Band 3 gene, SLC4A1, was amplified using cDNA from the donor plasmid pDONR221_SLC4A1 (RESOLUTE Consortium & Giulio Superti-Furga; Addgene plasmid # 132180, RRID: Addgene_132180) using primer sets cdB3F and cdB3R. PCR products were excised from the gel and subject to purification using the Monarch DNA Gel Extraction Kit (New England Biolabs). The purified PCR product and plasmid pKHEG were digested with NdeI. The plasmid vector pKHEG was simultaneously incubated with rSAP to prevent self-ligation and then heat inactivated at 65°C for 10 minutes. The digested plasmid vector and insert were ligated using T4 ligase at 15°C for 16 hours and subsequently heat-inactivated at 65°C for 10 minutes. Competent NEB 5α E. coli cells were transformed with the pCDB3 plasmid according to the protocol of the manufacturer. Plasmids were isolated from cultures utilizing the Monarch Plasmid Mini Prep Kit (New England Biolabs) and screened via restriction digest using the enzyme NdeI to confirm the plasmids possessed the desired insert. A subsequent restriction digest using the enzyme KpnI was performed to ensure the insert was in the same direction as the FGRp promotor (Horzempa et al., 2008). Plasmids possessing the desired product were then electroporated into F. tularensis LVS as previously described (Horzempa et al., 2011).

The generation of the pCDB3-emGFP plasmid was performed using a similar approach to the construction of pCDB3 with the following exceptions. Here, the same fragment of SLC4A1 was amplified from pDONR221_SLC4A1 using the primer set cdb3-gfpF and cdb3-gfpR leading to the addition of coding sequence for a 3’ linker peptide. The amplicon produced was excised from the gel and purified using the Monarch DNA Gel Extraction kit (New England Biolabs). The purified amplicon and pKHEG were digested using NdeI and the plasmid was also incubated with rSAP. Reactions were subsequently heat-inactivated at 65 °C for 10 minutes. The plasmid and amplicon restriction fragments were ligated using T4 ligase and then used to transform competent E. coli NEB 5α. Plasmids confirmed to contain the desired product in the appropriate orientation were then electroporated into F. tularensis LVS as previously described (Horzempa et al., 2008).

Western blotting was carried out similarly as previously described (Haggerty et al., 2024). Briefly, bacterial cultures were grown to stationary phase and normalized to the same OD₆₀₀ value to account for differences in bacterial growth. Bacterial lysates were generated by adding an equal volume of 2x Laemmli buffer supplemented with 5% 2-mercaptoethanol and boiling at 95°C for 5 minutes. Samples were electrophoresed in a 4-15% Mini-PROTEAN TGX gel (BioRad Laboratories) and were subsequently electroblotted onto a nitrocellulose membrane. This membrane was blocked with PBS containing 0.5% casein, 0.5% bovine serum albumin, 100 mg/L phenol red, and 0.02% sodium azide pH 7.4 for 30 minutes. The nitrocellulose membrane was then incubated with the primary antibody (rabbit anti-Band 3, Invitrogen CAT # PA5-80030, 1:1000; rabbit anti-emGFP, Invitrogen CAT # A-11122, 1:1000, rabbit anti-Ankyrin 1, Invitrogen CAT # PA5-63372, 1:1000) overnight and shaking at room temperature. The membrane was then washed 3 times with PBS and subsequently incubated with the secondary antibody (goat anti-rabbit alkaline phosphatase, Invitrogen CAT # 31340, 1:1000) for 1 hour shaking at room temperature. The membrane was washed 2 times with PBS and once with 0.05 M Tris buffer, pH 8.0 and then developed by incubating with Fast Red and Naphthol AS-MX phosphate disodium salt dissolved in Tris buffer until red protein bands appeared (Haggerty et al., 2024).

Specimens were imaged using an Olympus IX73 microscope with an ORCA-Flash4.0 LT+ Digital CC11440-42U CMOS camera (Hamamatsu) and a 100x NA 1.45 phase objective. F. tularensis LVS strains expressing Band 3-emGFP were inoculated from chocolate II agar plates into TSBc or CDM supplemented with kanamycin at 10 μg/ml at 37°C for 12–16 hours with agitation. Bacterial cultures were added to 1% agarose pads and upon which a cover slip was placed. Similar exposure times were used for all images, and brightness and contrast was adjusted uniformly across all images.

Freeze dried Dynabeads (M-270 Epoxy; Invitrogen) were equilibrated by resuspending 5 mg of these beads (~3.3 x 10⁸ beads/ml) in Buffer A (0.1 M sodium phosphate, pH 7.4) and incubating for 10 minutes at room temperature with rotation. Beads were placed on a magnet and the supernatant was discarded. The beads were resuspended in Buffer A one more time before being coupled to the antibody. Equilibrated beads were resuspended in an equal volume of Buffer A and ligand (rabbit anti-Band 3 antibody, 100 μg, Invitrogen CAT # PA5–80030 or rabbit anti-PdpC antibody, 100 μg, BEI resources, NR-4379) and vortexed. The same volume of Buffer B (3.0 M ammonium sulfate) was then added, followed by end-over-end rotation for 16–24 hours at 37°C. Beads were separated by a magnet and the supernatant was removed. Beads were washed 4 times with Buffer E1 (PBS) and then resuspended in E1 to achieve the desired concentration (~2.9 x 10⁸ beads/ml). Overnight cultures of F. tularensis LVS, LVS/pCDB3, LVS pdpC-null were normalized to an OD₆₀₀ of 3.0 and sonicated 3 times by 15 seconds to generate lysates. Lysates were added to the antibody conjugated beads and incubated with end-over-end rotation for 2 hours at 4°C. Beads were placed on a magnet and the supernatant was removed. Beads were washed 3 times with Buffer E1 and resuspended in 2x Laemmli buffer with 5% 2-mercaptoethanol. Beads were boiled at 95°C for 5 minutes and placed on a magnet to separate the beads. The supernatant was collected and stored at -20°C.

GFP-Trap Magnetic agarose beads (Chromotek) were equilibrated by resuspending beads in ice-cold dilution buffer (Chromotek) 3 times. Overnight stationary phase broth cultures of F. tularensis strains containing pCDB3-emGFP or pKHEG were normalized to the lowest OD₆₀₀ and sonicated to generate whole cell lysates. Lysates were added to equilibrated beads and incubated for 1 hour at 4°C with end-over-end rotation. The beads were separated with a magnet and the supernatant was removed. Beads were washed 3 times with Wash buffer (Chromotek) and transferred to a new tube after the final wash. The supernatant was removed, and beads were resuspended in 80μl of 2x Laemmli buffer supplemented with 5% 2-mercaptoethanol. To dissociate complexes from the beads, beads were boiled for 5 minutes at 95°C. The beads were separated with a magnet and the supernatant was stored at -20°C.

Digestion reagents were prepared as follows: Trypsin Protease stock solution was prepared by reconstituting 20 µg of the lyophilized material in 20 µL of 0.1% acetic acid. Aliquots of the trypsin stock solution were transferred to individual microcentrifuge tubes and stored at -80°C until needed. Destaining solution for Coomassie blue was prepared by dissolving 80 mg of ammonium bicarbonate in 40 mL of a 50:50 solution of acetonitrile and ultrapure water. Destaining solution for silver-stained gel was prepared by mixing a solution of 50% ultrapure water, 10% acetic acid, and 40% methanol. Digestion buffer was prepared by dissolving 10 mg of ammonium bicarbonate in 5 mL of ultrapure water. Reducing buffer was prepared immediately prior to use by mixing 3.3 µL of TCEP with 30 µL of digestion buffer for each digestion sample. Alkylation buffer was prepared immediately before use while avoiding exposure to light by preparing a 500 mM solution of IAA in ultrapure water with dilution of this stock solution by mixing 7 µL of the stock solution with 28 µL of digestion buffer for each digestion sample. Activated trypsin was prepared immediately prior to use by diluting 5 µL of trypsin stock solution with 45 µL of digestion buffer for each digestion sample.

Protein bands of interest were excised from the gel and stored at -20°C until use. Bands were destained in destaining solution for 30 minutes at 37°C with agitation. The destaining solution was removed and repeated until the stain was removed from the samples. Reducing buffer was added to samples and incubated at 60°C for 10 minutes. Reducing buffer was removed and alkylation buffer was added to samples and incubated in the dark at room temperature for 60 minutes. Following incubation, alkylation buffer was removed, and samples were washed twice with destaining buffer at 37°C for 15 minutes with agitation. Samples were digested by adding acetonitrile and incubating samples for 15 minutes at room temperature. Acetonitrile was removed and samples were allowed to dry for 10 minutes at room temperature. Once samples were dried, activated trypsin and digestion buffer were added to samples and incubated at 30°C overnight. After incubation, activated trypsin and digestion buffer were removed and 1% formic acid solution was incubated with the samples for 5 minutes. Once the formic acid solution was removed, samples were subject to purification using Pierce C18 spin columns (Thermo Fisher Scientific). Samples were stored at -20°C until further analysis could be conducted.

A Thermo Scientific Q-Exactive hybrid Quadropole-Orbitrap mass spectrometer (Thermo Fisher, San Jose, CA) was utilized for all liquid chromatography-mass spectrometry (LC/MS) measurements. A Thermo Scientific Vanquish Ultra-high performance Liquid Chromatograph system (Thermo Fisher, San Jose, CA) was utilized for the separation of the digestion samples with subsequent mass measurement. Liquid chromatography settings are as follows: separation of peptides in digestion samples was conducted utilizing a Phenomenex Aeris Peptide XB-C18 analytical column with an Agilent Eclipse XDB-C18 guard. A buffer system consisting of 0.1% formic acid in ultrapure water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B) was also used. Each digestion sample was injected in triplicate and separation was achieved over a 60-minute separation utilizing the following gradient elution: Initial conditions 5.0% B, 0.0 min to 45.0 min linear gradient to 65.0% B, 45 min to 50 min linear gradient to 95.0% B, 55 min 95.0% B, 55 min to 55.1 min linear gradient to 5.0% B, 55.1 min to 60.0 min 5% B at a flow rate of 250 µL min^-1. The column temperature during the separation was maintained at 40°C with an autosampler temperature of 4°C.

Mass spectrometer operational parameters were optimized utilizing the Tune software for a flow rate of 250 µL min^-1. Experiments were conducted in positive ion mode at +4.5 kV utilizing a mass-to-charge (m/z) scan range of 275 – 2000 (1 microscan). The instrument was operated at a resolving power of 70,000 for all full scan m/z measurements with an AGC target setting of 1 x 10⁶ and an S-lens voltage of 50 V. The inlet capillary temperature was maintained at 250°C with the following nitrogen gas flow rates utilized for the HESI source: sheath gas flow rate was set to 25 arbitrary units (arb), the auxiliary gas flow rate was set to 12 arb with auxiliary gas heater maintained at 112.5 °C. MS² scans were obtained for the top 40 peaks at a resolution setting of 17,500 with an AGC target of 1 x 10⁵ and an isolation window of ± 4.0 m/z. Stepped normalized collision energy (NCE) was set to 28 eV with a dynamic exclusion time of 10.0 seconds. Raw data files from the mass spectrometer were utilized to identify peptides present within the digestion samples utilizing the Proteome Discoverer software suite (Thermo Fisher). Dataset features were identified using exact mass comparisons for retention time (t_R ) resolved mass spectra with default parameters for MS² spectra utilizing Orbitrap m/z detection. Peptide sequence matches were detected in positive ion mode ranging from [M+H]⁺ to [M + 6H]⁺⁶ ions with sequence lengths ranging from 6 to 25. Proteins were identified utilizing the ProteinProspector software (UCSF) and uploading known peptide masses into the MS Fit database.

All data were analyzed using GraphPad Prism software. P-values and tests used to determine statistical significance can be found within figure legends.

We hypothesized that F. tularensis uses red blood cell (RBC) surface protein Band 3 as a receptor to invade erythrocytes. This hypothesis is based on several lines of circumstantial evidence. First, a study by Schmitt et al. found that treating human RBCs with P. guttatus (blue-bellied black snake) venom inhibits F. tularensis LVS invasion (Schmitt et al., 2017). The venom binds to Band 3 on the surface of RBCs and disrupts the RBC’s spectrin filament architecture, suggesting that either Band 3, Spectrin, or both are involved.

To determine the role of Band 3 in invasion of RBCs by F. tularensis, we used similar methodologies to those used to evaluate the role of this host protein during erythrocyte invasion by other microbes (Crandall et al., 1993; Hogh et al., 1994; Clough et al., 1995; Goel et al., 2003; Deng et al., 2012; Baldwin et al., 2014; Almukadi et al., 2019). Firstly, this host membrane protein was physically blocked using anti-Band 3 Fab2 fragments. Fab2 fragments were used because RBC invasion requires intact human serum and therefore whole antibodies against Band 3 would have activated complement and lysed the targeted erythrocytes (Horzempa et al., 2011). Here, RBCs were incubated with anti-Band 3 Fab2 fragments for one hour prior to the addition of F. tularensis LVS bacteria. Blocking Band 3 in this fashion inhibited erythrocyte invasion by F. tularensis ( Figure 1A ). However, incubation with Fc portions of the anti-Band 3 antibody had no effect on RBC invasion ( Figure 1A ). To validate this finding, F. tularensis LVS was incubated with rBand 3 for one hour before adding the erythrocytes to evaluate invasion ( Figure 1B ). Treatment with rBand 3 (10 μg/ml) diminished invasion, as would be expected in a competitive inhibition of bacterial binding with the erythrocyte Band 3 ( Figure 1B ). A similar reduction in erythrocyte invasion was observed when rBand 3 was included at 5 μg/ml and 15 μg/ml (data not shown). These findings support the hypothesis that the erythrocyte Band 3 protein plays a role in RBC invasion by F. tularensis.

Band 3 interacts with several integral membrane proteins such as Glycophorin A (GPA) and ankyrin to form complexes with the spectrin cytoskeleton (van den Akker et al., 2010). Therefore, given the data that Band 3 may contribute to erythrocyte invasion, it is possible that other proteins that complex with this host protein are important for red blood cell invasion by F. tularensis. Furthermore, GPA has been identified as a receptor for Babesia divergens and Plasmodium falciparum facilitating red blood cell invasion (Lobo, 2005). Therefore, we sought to determine if Glycophorin A plays a role in erythrocyte invasion by F. tularensis.

To study the potential role of GPA in F. tularensis erythrocyte invasion, we inhibited interactions with this host protein using Fab2 fragments and carried out a gentamicin protection assay. Surprisingly, erythrocytes treated with anti-GPA Fab2 fragments resulted in a significant increase in invasion by F. tularensis LVS compared to the untreated erythrocytes ( Figure 2 ) suggesting that this host protein may be antagonizing bacterial entry.

Engulfment of bacteria or invasion of host cells is typically mediated by the host cytoskeleton (Adam et al., 1995; Clemens et al., 2005). Therefore, F. tularensis likely manipulates the erythrocyte spectrin cytoskeleton to invade these host cells. This hypothesis is supported by the previous finding that spectrin impairment inhibits red blood cell invasion by F. tularensis (Schmitt et al., 2017). Band 3 is anchored to the spectrin cytoskeleton in two distinct complexes: the Band 3/Ankyrin complex and the 4.1R complex (Salomao et al., 2008; Mankelow et al., 2012). In the 4.1R complex, Band 3 is attached to the spectrin cytoskeleton via small actin bundles and the 4.1 protein. Previous data indicated that actin was dispensable for erythrocyte invasion (Schmitt et al., 2017). In the other complex, Band 3 interacts with Ankryin-1 and other accessory proteins to attach to the spectrin cytoskeleton. Given the potential role of Band 3 and Spectrin, here we sought to investigate if Ankyrin-1 is required for erythrocyte invasion by F. tularensis. Erythrocytes were isolated from the whole blood of Ankyrin-1-deficient mice (nb/nb) or isogenic wild-type mice and were subjected to a gentamicin protection assay. Here, Ankyrin-deficient erythrocytes (nb/nb) exhibited decreased invasion by F. tularensis compared to red blood cells isolated from wild-type mice ( Figure 3 ). We did not observe a noticeable loss in the number of erythrocytes from the (nb/nb) mice compared to the wild-type animals during the duration of these in vitro studies, suggesting that cell lysis was not likely responsible for the disparity in erythrocyte invasion observed here (data not shown). These data suggest that erythroid Ankyrin-1 is required for red blood cell invasion by F. tularensis.

Given the potential role of Band 3 in RBC invasion, we sought to identify F. tularensis bacterial proteins that interact with this host protein to further elucidate the mechanism of RBC entry. Further, given that Ankyrin-1 is required for invasion, and that this protein binds to the cytoplasmic domain of Band 3, we reasoned that this domain may be a target of F. tularensis effectors. To identify potential binding partners between F. tularensis and the cytoplasmic domain of Band 3, we recombinantly expressed the cytoplasmic domain of Band 3 (CBD3) (encoded in pCDB3) in F. tularensis LVS. Western blotting was used to confirm expression of the recombinant Band 3 protein in LVS ( Figure 4A ). When probed with an anti-Band 3 polyclonal antibody, two bands were observed at approximately 45 kD and 70 kD for LVS/pCDB3, which matched similar banding patterns observed with the red blood cell control ( Figure 4A ). No bands were observed in LVS bacteria harboring the parent vector (pKHEG), indicating the bands observed in LVS/pCDB3 were due to the expression of the cytoplasmic domain of Band 3.

Next, co-immunoprecipitation using the recombinant cytoplasmic domain of Band 3 as “the bait” was performed to identify any potential F. tularensis binding partners from bacterial lysates. Here, an anti-Band 3 polyclonal antibody was conjugated to magnetic agarose beads and co-incubated with LVS/pCDB3 lysates for at least 2 hours. Beads were then washed twice with PBS, and protein complexes were eluted by boiling beads in 2x Laemmli buffer supplemented with 5% 2-mercaptoethanol and boiling at 95°C for 5 minutes. Following sample elution and SDS-PAGE, imperial staining and silver staining revealed four potential F. tularensis interacting proteins of approximately 10, 15, 30, and 60 kD ( Figure 4B ).

The polyclonal antibody used to immobilize the cytoplasmic portion of Band 3 could have potentially interfered with the binding of bacterial proteins in the co-immunoprecipitation depicted in Figure 4B . Therefore, as an alternative approach, coding sequence for the cytoplasmic domain of Band 3 (CDB3) was fused to emerald green fluorescent protein (CDB3-EmGFP; encoded in pCDB3-emGFP) and this chimeric protein was recombinantly expressed in F. tularensis LVS. In this iteration, the EmGFP domain was targeted in a co-immunoprecipitation assay to identify F. tularensis LVS protein binding partners to the cytoplasmic domain of Band 3. Western blotting was used to determine expression of the CDB3-EmGFP fusion. When probed with an anti-Band 3 polyclonal antibody, bands were observed between from ~50 kD to above 75 kD with additional bands below the 50 kD marker for F. tularensis LVS/pCDB3-emGFP, while lysates from F. tularensis LVS alone produced no detectible bands ( Figure 5A ). When probed with an anti-GFP monoclonal antibody, bands were also observed above 50 kD to above 75 kD as well as a band near 27 kD ( Figure 5B ). However, F. tularensis LVS alone again produced no detectible bands ( Figure 5B ). These data indicate that the observed bands were due to the expression of the cytoplasmic domain of Band 3 linked to emGFP.

We next sought to confirm expression of CDB3-emGFP using fluorescence microscopy. F. tularensis LVS cells producing recombinant CBD3-emGFP were grown in Chamberlain’s Defined Medium (CDM) to stationary phase at 37 °C with agitation, then added to 1% agarose pads in PBS to visualize the bacteria. In LVS/pCDB3-emGFP, we observed robust fluorescence in cells expressing cdb3-emgfp, compared to wild type bacteria which elicited no detectible fluorescence ( Supplementary Figure S1 in the Supplementary Material ). Fluorescence was exhibited uniformly throughout the cytoplasm of LVS/pCDB3-emGFP, suggesting expression of CDB3-emGFP does not result in localization of the resulting fusion protein to a specific bacterial site nor did expression of this recombinant protein produce evidence of inclusion bodies which suggests that this protein did not misfold.

Next, potential CBD3-EmGFP protein complexes were isolated using magnetic beads conjugated to anti-GFP nanobodies (ChromoTek GFP-Trap). These beads were co-incubated with LVS/pCDB3-emGFP lysates for at least 2 hours at 37 °C. Samples were eluted and analyzed using SDS-PAGE. Proteins were visualized by staining using imperial ( Figure 6A ) and silver stains ( Figure 6B ). Staining revealed potential F. tularensis bacterial proteins binding to the cytoplasmic domain of Band 3 ( Figure 6 ). To ensure that these bacterial proteins were not directly binding to EmGFP, we incubated magnetic beads conjugated to anti-GFP nanobodies with bacterial lysates containing the parent vector expressing only EmGFP (LVS/pKHEG). No bands were observed following SDS-PAGE and staining of lysates from this control strain, indicating the bacterial proteins observed were interacting with the cytoplasmic domain of Band 3 and not the EmGFP portion of this chimeric protein (data not shown).

To identify the F. tularensis LVS bacterial proteins potentially interacting with the cytoplasmic domain of Band 3, bands of interest were excised from the stained gels and subject to in-gel proteomic digestion. This digested material was analyzed using liquid chromatography-mass spectrometry to identify peptide masses. These peptide masses were subsequently analyzed using the ProteinProspector software MS-Fit database (UCSF) by crossing samples with known F. tularensis proteins. The Molecular Weight Search Engine (MOWSE) score was used to determine how well observed peptide masses match with peptide masses from known proteins (Damodaran et al., 2007). A MOWSE score higher than 50 indicates a statistically significant match between observed peptide masses and peptide masses of a particular protein ( Table 2 ). Some notable proteins that were identified that fulfill this criterion include the outer membrane vesicle-associated lipase protein, FtlA ( Table 2 ). FtlA has been found to contribute to attachment and entry into host cells and pathogenesis in a mouse model of tularemia (Chen et al., 2017). A conserved outer membrane protein of F. tularensis, known as LpnA (Tul4) was also identified ( Table 2 ). This protein has been shown to stimulate toll-like receptor 2 (TLR2) in human macrophages but does not affect F. tularensis virulence in mice (Forestal et al., 2008). The glycine cleavage protein T (GcvT), which is a part of the glycine cleavage system in F. tularensis was identified as a potential interactor with Band 3 ( Table 2 ). This protein has been shown to contribute to pathogenesis in murine models as well as intracellular replication in serine limiting environments ( Table 2 ) (Brown et al., 2014).

Given the previous research conducted on the role of GcvT in F. tularensis pathogenesis, as well as serendipitously having a GcvT deletion mutant on hand, we sought to investigate the potential role of GcvT in erythrocyte invasion. To determine if GcvT is important for erythrocyte invasion, we performed a gentamicin protection assay using the F. tularensis LVS gcvT deletion mutant. Here, human erythrocytes were isolated from whole blood and incubated with the indicated F. tularensis strains at an MOI ~ 12.5 for 2 hours ( Figure 7 ). Following incubation, cells were treated with gentamicin to kill extracellular bacteria, washed with PBS, and then lysed and plated to enumerate CFU ( Figure 7 ). This assay indicated that the gcvT strain exhibited a significant decrease in erythrocyte invasion compared to the wild type bacteria ( Figure 7 ). Complementation of gcvT in the F. tularensis gcvT mutant rescued invasion to wild-type levels indicating that the observed decrease in erythrocyte invasion by the mutant was not due to a polar or pleiotropic effect. Moreover, disk diffusion assays indicated that the gcvT mutant exhibited a similar level of sensitivity to both gentamicin and SDS compared the wild type strain (data not shown). Therefore, the observed decrease in mutant bacteria surviving the gentamicin protection assay was unlikely due to an artifact of this experimental methodology (data not shown). These data suggest that GcvT is important for erythrocyte invasion by F. tularensis likely through an interaction with the cytoplasmic domain of Band 3.

Because PdpC is required for attachment to and invasion of red blood cells, we hypothesized that identifying erythrocyte proteins that interact with PdpC could provide insight into the mechanism of red blood cell invasion (Cantlay et al., 2022). To identify potential erythrocyte binding partners of PdpC, a co-immunoprecipitation assay was performed. Magnetic agarose beads were conjugated to an anti-PdpC polyclonal antibody and co-incubated with F. tularensis LVS lysates for at least 2 hours. Following incubation, isolated lysed human erythrocytes diluted to a desired concentration (1 x 10⁷ cells/ml) were added to the beads and incubated at 37 °C for another two hours. As a control, erythrocyte proteins were also added to antibody-conjugated beads incubated with lysates from a F. tularensis mutant strain lacking PdpC to assess potential non-specific binding. The beads were then washed twice with PBS, and the bound protein complexes were eluted by boiling beads in 2x Laemmli buffer supplemented with 5% 2-mercaptoethanol. Samples were analyzed via SDS-PAGE, and proteins were visualized by silver stain. A single band was observed above 250 kD but was not observed in our LVS pdpC-null mutant ( Figure 8 ). To confirm the band observed was not PdpC, a western blot was conducted using an anti-PdpC polyclonal antibody (data not shown). PdpC was not detected, indicating the band observed was an erythrocyte protein.

To identify the potential erythrocyte protein interacting with PdpC, the single band was excised from the gel and subject to in-gel proteomic digestion to purify the protein. The sample was then analyzed using liquid chromatography-mass spectrometry to identify peptide masses. Peptide masses were subsequently analyzed using the ProteinProspector software MS-Fit database (UCSF) by cross referencing with peptide masses of known human erythrocyte proteins. Human erythrocyte proteins that most closely fit sample peptide masses (MOWSE score >50) were then identified ( Table 3 ). The erythrocyte protein identified to potentially interact with PdpC was Spectrin (alpha chain) (MOWSE score of 3086; Table 3 ). This finding supports previous data suggesting the involvement of spectrin in erythrocyte invasion (Schmitt et al., 2017). This result also indicates that PdpC interacts with the major erythrocyte cytoskeletal protein, Spectrin (alpha chain), potentially mediating erythrocyte invasion through this interaction.