key: cord-0952806-qum8cnii
authors: Hoffmann, Magnus A.G.; Kieffer, Collin; Bjorkman, Pamela J.
title: In vitro characterization of engineered red blood cells as viral traps against HIV-1 and SARS-CoV-2
date: 2021-03-10
journal: Mol Ther Methods Clin Dev
DOI: 10.1016/j.omtm.2021.03.003
sha: 8015aa1ec4de619cb4625dfde3dd9ca6d4806faf
doc_id: 952806
cord_uid: qum8cnii

Engineered red blood cells (RBCs) expressing viral receptors could be used therapeutically as viral traps as RBCs lack nuclei and other organelles required for viral replication. However, expression of viral receptors on RBCs is difficult to achieve since mature erythrocytes lack the cellular machinery to synthesize proteins. Here we show that the combination of a powerful erythroid-specific expression system and transgene codon optimization yields high expression levels of the HIV-1 receptors CD4 and CCR5, as well as a CD4-glycophorin A (CD4-GpA) fusion protein in erythroid progenitor cells, which efficiently differentiated into enucleated RBCs. HIV-1 efficiently entered RBCs that co-expressed CD4 and CCR5, but viral entry was not required for neutralization as CD4 or CD4-GpA expression in the absence of CCR5 was sufficient to potently neutralize HIV-1 and prevent infection of CD4+ T-cells in vitro due to the formation of high-avidity interactions with trimeric HIV-1 Env spikes on virions. To facilitate continuous large-scale production of RBC viral traps, we generated erythroblast cell lines stably expressing CD4-GpA or ACE2-GpA fusion proteins, which produced potent RBC viral traps against HIV-1 and SARS-CoV-2. Our in vitro results suggest that this approach warrants further investigation as a potential treatment against acute and chronic viral infections.

enucleated RBCs. During the erythroid differentiation process, transgene expression is 48 restricted through transcriptional silencing 8 , translational control mechanisms 9 and 49 degradation of proteins that are not normally present in RBCs 10 . 50

One strategy to overcome the latter problem is to generate chimeric proteins by fusing the 51 extracellular domain of a non-erythroid protein to a protein that is abundantly expressed in 52

RBCs, such as glycophorin A (GpA) 11, 12 . However, this approach is limited to single-pass 53 transmembrane proteins, prevents localization of viral receptors to their natural membrane 54 subdomains, and might not achieve sufficiently high receptor levels to effectively entrap the 55 virus. In the case of a potential HIV-1 therapeutic, additional strategies are required to 56 J o u r n a l P r e -p r o o f

To evaluate the efficacy of RBC viral traps against HIV-1, we generated RBCs that expressed 101 CD4+/-CCR5 or CD4-GpA+/-CCR5 (Fig. 2a) and used the β-lactamase (BlaM) fusion 102 assay 17 to evaluate if HIV-1 can enter RBC viral traps through attachment of HIV-1 Env 103 spikes to the receptors presented on the RBC surface and subsequent fusion of the viral and 104 RBC membranes. RBCs were incubated with a CCR5-tropic HIV-1 YU2 pseudovirus carrying 105 a BlaM-Vpr fusion protein that enters cells upon infection. When infected cells are exposed 106 to the fluorescence resonance energy transfer (FRET) substrate CCF2-AM, BlaM cleaves the 107 β-lactam ring in CCF2-AM resulting in a shift of its emission spectrum from green (520 nm) 108

to blue (447 nm) 17 . Whereas viral entry events were ≤0.3% in control RBCs and CD4-RBCs, 109 entry was detected in 6.1% of CD4-CCR5-RBCs suggesting that RBC viral traps that present 110 both receptors can entrap the virus. (Fig. 2b; Fig. S3a ). Since only 1/3 of these RBCs 111 expressed both receptors (Fig. 1e) , this corresponds to infection of almost 20% of CD4-112 CCR5-RBCs. CCR5 expression on enucleated RBCs was slightly higher than on nucleated 113 cells; thus it is unlikely that HIV-1 preferentially entered the small number of remaining 114 nucleated cells (Fig. S3b) . Higher rates of viral entry were observed for RBCs that co-115 expressed CD4 and the alternative HIV-1 co-receptor CXCR4 after incubation with a 116 CXCR4-tropic HIV-1 HxBc2 pseudovirus ( Fig. 2c; Fig. S4a ). However, lower frequencies of 117 viral entry were detected for RBCs that co-expressed the CD4-GpA fusion protein and CCR5 118 or CXCR4 (Fig. 2b,c) , and addition of the CD4 D3D4 domains to CD4-GpA did not improve 119 viral entry efficiency (Fig. S4b ). Unlike CD4, GpA does not localize to lipid raft 120 subdomains 18 , thus we speculate that these low rates of viral entry resulted from a lack of co-121 localization between CD4-GpA and the CCR5 and CXCR4 co-receptors. 

We assessed the therapeutic potential of RBC viral traps using a modified version of the 126 HIV-1 TZM-bl neutralization assay 19 (Fig. 3a) . After incubating RBCs with HIV-1 YU2 127 pseudovirus, samples were centrifuged to remove RBCs and virions that attached to or 128 infected RBCs. Supernatants containing free virions that had not been captured by RBCs 129 were transferred to 96-well plates and TZM-bl cells were added to measure infectivity. In 130 three independent assays, CD4-GpA-RBCs neutralized HIV-1 YU2 most potently at an average 131 half-maximal inhibitory concentration (IC 50 ) of 1.9x10 6 RBCs/mL ( Fig. 3b ; Table 1 ). This 132 concentration is equivalent to 0.04% of the RBC concentration of human blood (~5x10 9 133 RBCs/mL), suggesting that it would be feasible to achieve therapeutic concentrations in vivo. 134

CD4-GpA-RBCs were ~3-fold more potent than CD4-RBCs, likely due to higher expression 135 levels ( Fig. 2a; Fig.S5 ). While CCR5 co-expression had no impact on the potency of CD4-136

GpA-RBCs, co-expression of CCR5 lowered the neutralization activity of CD4-CCR5-RBCs 137 by almost 3-fold in comparison to CD4-RBCs ( Fig. 3b We previously showed that virus-like nanoparticles presenting clusters of CD4 (CD4-VLPs) 143 that formed high-avidity interactions with trimeric HIV-1 Env spikes on virions potently 144 neutralized a diverse panel of HIV-1 strains and prevented viral escape in vitro 20 . To confirm 145 that RBC viral traps can also form high-avidity interactions with Env, we evaluated 146 neutralization against a mutant HIV-1 YU2 Env G471R pseudovirus that was resistant to 147 monomeric soluble CD4, but was sensitive to CD4-VLPs 20 . CD4-GpA-RBCs potently 148 J o u r n a l P r e -p r o o f neutralized the HIV-1 YU2 G471R pseudovirus (IC 50 = 1.0x10 7 RBCs/mL) (Fig. 3c) , 149

suggesting that RBC viral traps and CD4-VLPs would be similarly effective in preventing 150 viral escape through formation of high-avidity interactions with HIV-1 Env spikes. 151

The ability of RBC viral traps to protect HIV-1 target cells from infection was evaluated by 153 co-culturing control RBCs or CD4-GpA-RBCs with Rev-A3R5 CD4+ T-cells 21 , a reporter 154 cell line that expresses luciferase upon HIV-1 infection (Fig. 4a) . RBCs, CD4+ T-cells, and 155

HIV-1 pseudovirus were co-incubated at RBC to T-cell ratios of 2:1 and 5:1 overnight under 156 shaking conditions. The pseudovirus was removed by centrifugation and the cells were re-157 suspended in Rev-A3R5 CD4+ T-cell media to permit outgrowth of CD4+ T-cells. After 36 158 hours, luminescence was measured to determine if the presence of RBC viral traps prevented 159 infection of CD4+ T-cells. While control RBCs had no effect, CD4-GpA-RBCs lowered 160 infection rates by 50% and 70%, respectively, demonstrating that RBC viral traps can 161 effectively prevent infection of HIV-1 target cells at RBC:T-cell ratios that are ~1,000-fold 162 lower than typically found in human blood (~5,000:1) 22 (Fig. 4b) . Since HIV-1 did not 163 efficiently enter CD4-GpA-RBCs (Fig. 2b) , these findings also suggest that high-avidity 164 binding of HIV-1 virions to RBC viral traps is sufficient to prevent attached virions from 165 infecting target cells. 166

To generate a renewable and cost-effective source of RBC viral traps, we engineered the 169 immortalized BEL-A erythroblast cell line 23 to stably express high levels of CD4-GpA (Fig.  170 5a). The BEL-A / CD4-GpA cells efficiently differentiated into enucleated RBCs, as >50% of 171 CD71-expressing cells did not stain for the nuclear marker DRAQ5 (Fig. 5b) . After 172 differentiation, CD71+/DRAQ5-RBCs were purified using fluorescence-activated cell 173 sorting (FACS). The majority of RBCs still expressed CD4-GpA (Fig. 5c ) and potently 174 neutralized HIV-1 YU2 in vitro (IC 50 = 2.1x10 7 RBCs/mL) (Fig. 5d ). Independent replicates of 175 in vitro differentiation of BEL-A / CD4-GpA cells achieved comparable yields of RBC viral 176 traps (Fig. 5b,c; Fig. S6 ) suggesting that engineered erythroblast cell lines could be used to 177 continuously produce potent RBC viral traps against HIV-1. However, overall production 178 yields would also depend on the quality of the RBCs as the viability of BEL-A cells 179 decreases to ~80% at the end of differentiation 24 and cells could also get damaged during the 180 purification process. To ensure complete removal of nucleated cells for in vivo studies, the 181 RBC viral traps could be further purified using leukoreduction filters and/or gamma 182

To evaluate if RBC viral traps could be effective against other viruses, we generated a BEL-184 A cell line that continuously produces RBC viral traps against SARS-CoV-2, the virus that 185 caused the ongoing COVID-19 pandemic 25 . BEL-A cells were transduced to stably express a 186 chimeric ACE2-GpA protein containing the extracellular domain of the SARS-CoV-2 187 receptor ACE2 25 fused to GpA (Fig. 6a) . Differentiation efficiency and transgene expression 188 on sorted CD71+/DRAQ5-RBCs was comparable to the BEL-A / CD4-GpA cell line (Fig.  189 6b,c). Importantly, lentivirus-based SARS-CoV-2 pseudovirus 26 was highly susceptible to 190 ACE2-GpA-RBC neutralization (IC 50 = 7x10 5 RBCs/mL) (Fig. 6d) suggesting that RBC viral 191 traps have the potential to be effective anti-viral agents against a range of viruses. HIV-1 pseudovirus entered engineered RBCs more efficiently when CCR5 and CXCR4 were 222 co-expressed with wild-type CD4 rather than chimeric CD4-GpA, thus demonstrating that 223 protein modifications that have been used to enhance RBC surface expression 11,12 can affect 224 the functionality of the therapeutic protein. A lack of co-localization of CD4-GpA and co-225 receptors could be the cause of the low entry rates, as CD4 and CCR5 have been shown to 226 co-localize in lipid raft microdomains 13,14 and GpA is not typically associated with lipid 227 rafts 18 . It is also possible that substitution of the membrane-proximal extracellular, 228 transmembrane, or cytoplasmic domains of CD4 interfered with the ability of CD4-GpA to 229 initiate the interaction between HIV-1 Env and co-receptors. 230

Expression of CD4 or the CD4-GpA fusion protein in the absence of CCR5 was sufficient to 231 potently neutralize HIV-1 in vitro due to formation of high-avidity interactions between 232 clusters of CD4 or CD4-GpA on the RBC surface and trimeric HIV-1 Env spikes on virions. 233

suggesting that viral attachment to RBC viral traps effectively prevents HIV-1 virions from 235 infecting target cells. We previously showed that such high-avidity interactions enhanced the 236 potency of CD4-VLPs by >10,000-fold in comparison to conventional CD4-based inhibitors 237 such as soluble CD4 and CD4-Ig, and that HIV-1 was unable to escape against CD4-VLPs in 238 vitro 20 . In contrast to CD4-VLPs that have short in vivo half-lives, RBC viral traps could 239 persist in vivo for months, implying the RBC approach has the potential to provide sustained 240 control of HIV-1 infection. RBC viral traps neutralized HIV-1 in vitro at 2,500-fold lower 241

concentrations than the concentration of total RBCs in human blood and reduced HIV-1 242 infection of CD4+ T-cells by 70% at an RBC to T-cell ratio of 5:1. Given that RBCs 243 outnumber CD4+ T-cells by ~5,000:1 in the blood 22 and CD4+ T-cell lines are more 244 permissive than natural CD4+ T-cells 36 , these results suggest that therapeutic concentrations 245 of RBC viral traps could be achieved in vivo. 246

Erythroblast cell lines that stably express therapeutic proteins represent a renewable and more 247 cost-effective source for large-scale manufacturing of genetically-engineered RBCs than 248 CD34+ HSCs. BEL-A cell lines that stably expressed CD4-GpA and ACE2-GpA efficiently 249 differentiated into potent RBC viral traps against HIV-1 and the pandemic SARS-CoV-2 250 virus, respectively, suggesting that RBC viral traps could be effective treatments against a 251 diverse range of viruses. RBC viral traps could become a rapid-response treatment strategy 252

for future viral outbreaks, as erythroblast cell lines could be rapidly developed once a host 253 receptor for a pandemic virus has been identified. 254

In vivo studies will be required to evaluate the safety and efficacy of RBC viral traps and a 255 number of potential issues need to be addressed. First, it has been shown that reticulocytes 256 generated by in vitro differentiation mature in vivo into biconcave erythrocytes 37 , but it needs 257 to be determined if surface expression of viral receptors is affected by this final maturation 258 step in vivo. Second, the half-life of genetically-modified RBCs expressing chimeric VHH-259

GpA/Kell proteins was comparable to control RBCs following intravenous injection in 

Human cord blood or mobilized peripheral blood CD34+ HSCs (StemCell Technologies) 277

were differentiated into enucleated RBCs using a modified version of a previously-described 278 protocol 12 . Briefly, CD34+ HSCs were cultured in expansion medium (100 ng/mL rhFlt3, 279 100 ng/mL rhSCF, 20 ng/mL rhIL-6, 20 ng/mL rhIL-3, and 100 nM dexamethasone in 280 StemSpan II medium) at a density of 10 5 cells/mL for 4 days. Cells were then placed in 281 differentiation 1-2 medium (2% human AB plasma, 3% human AB serum, 3 U/mL heparin, 282 10 ng/mL rhSCF, 1 ng/mL rhIL-3, and 3 U/mL erythropoietin in StemSpan II medium) at a 283 density of 10 5 cells/mL for 3 days and at 2 x 10 5 cells/mL for an additional 3 days. The cells 284

were then passaged into differentiation 3 medium (2% human AB plasma, 3% human AB 285 serum, 3 U/mL heparin, 10 ng/mL rhSCF, and 1 U/mL erythropoietin in StemSpan II 286 medium) at a density of 2 x 10 5 cells/mL for 4 days. To induce RBC maturation, cells were 287 cultured in differentiation 4 medium (2% human AB plasma, 3% human AB serum, 3 U/mL 288 heparin, 0.1 U/mL erythropoietin, and 200 µg/mL holo-transferrin in StemSpan II medium) at 289 a density of 10 6 cells/mL for 4 days, and in differentiation 5 medium (2% human AB plasma, 290 3% human AB serum, 3 U/mL heparin, and 200 µg/mL holo-transferrin in StemSpan II 291 medium) at a density of 5 x 10 6 cells/mL for an additional 3 days. For morphological 292 analysis, cells were spun onto glass slides by cytocentrifugation, stained with May-293

Grünwald-Giemsa reagents (Sigma-Aldrich), and examined under an LSM800 laser scanning 294 confocal microscope (Zeiss). were seeded at a density of 10 6 cells/mL in 12-well plates in the presence of 10 µg/mL 317 polybrene. 20 µL of concentrated lentiviral vector was added per well and plates were spun 318 for 1.5 hours at 850 x g at 30°C. Plates were then incubated for 3 hours at 37°C before 319 passaging the transduced cells into differentiation 3 medium. For cells that were co-320 transduced to express two transgenes, 20 µL of each lentiviral vector was added per well. To 321 generate large numbers of engineered RBCs for neutralization assays, two transductions steps 322 were performed on days 10 and 14 of the differentiation protocol. 

by transferring the cells into primary media (3% human AB serum, 2% FBS, 3 U/mL 408 heparin, 10 ng/mL rhSCF, 1 ng/mL rhIL-3, 3 U/mL erythropoietin, 200 µg/mL holo-409 transferrin, and 1 µg/mL doxycycline in StemSpan II medium) for 3-4 days at a density of 2 x 410 10 5 cells/mL. To induce RBC maturation, cells were moved into tertiary media (3% human 411 AB serum, 2% FBS, 3 U/mL heparin, 3 U/mL erythropoietin, 500 µg/mL holo-transferrin, 412 and 1 U/mL Pen-Strep in StemSpan II medium) for 4 days at a density of 1 x 10 6 cells/mL. 413

Enucleated RBCs were purified by FACS on day 7 of the BEL-A differentiation protocol. 415

Brilliant Violet 421-conjugated anti-human CD71 antibody (BioLegend) and the nuclear 416 stain DRAQ5 (Abcam) were diluted 1:100 and 1:1,000 in PBS+ (PBS supplemented with 2% 417 FBS). Cells were stained at a concentration of 2.5 x 10 7 cells/mL for 30 min at room 418 temperature in the dark. After two washes in PBS+, cells were resuspended in PBS+ at a 419 concentration of 1 x 10 7 cells/mL. Enucleated RBCs were defined as CD71+/DRAQ5-cells 420 and this cell population was purified using a SONY SH800 cell sorter (Sony Biotechnology). Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease 534

Inhibitor. 7566-7571, doi:10.1073/pnas.94.14.7566 (1997) . RBCs (green), and CD4-GpA-CCR5-RBCs (blue). Data points are the mean and SD of 634 duplicate measurements. c, In vitro neutralization assay against mutant HIV-1 YU2 Env G471R 635 

Targeting recombinant 593 thrombomodulin fusion protein to red blood cells provides multifaceted 594 thromboprophylaxis

In vivo binding and clearance of circulating antigen by bispecific heteropolymer-598 mediated binding to primate erythrocyte complement receptor

Designed proteins induce the formation of nanocage-containing 602 extracellular vesicles

Foot-and-Mouth-Disease Virus 2a Oligopeptide Mediated 604 Cleavage of an Artificial Polyprotein

CD4-GpA-RBCs (green), and CD4-GpA-636

Data points are the mean and SD of duplicate measurements

Schematic illustrating 640 the workflow for co-incubation of CD4-GpA-RBCs, Rev-A3R5 CD4+ T-cells, and HIV-1 YU2 641 pseudovirus to assess the ability of RBC viral traps to prevent infection of HIV-1 target cells 642 in vitro. b, Bar chart comparing the ability of control RBCs (black) and CD4-GpA-RBCs 643 (green) to reduce the infection rate of

BEL-A erythroblast cell lines stably express CD4-GpA to produce potent RBC 647 viral traps against HIV-1. a, Flow cytometry measurement of CD4-GpA expression on 648

GpA cells pre-differentiation. b, Flow cytometry analysis of enucleated

Enucleated RBCs expressed CD71 and did not stain 650 for the nuclear dye DRAQ5. c, Flow cytometry analysis of CD4-GpA expression on 651

GpA cells post-sorting on day 8 of differentiation. d, In vitro 652 neutralization assay against HIV-1 YU2 pseudovirus comparing control RBCs (black)

Data points are the mean and SD of duplicate measurements

BEL-A erythroblast cell lines stably express ACE2-GpA to produce potent RBC 657 viral traps against SARS-CoV-2. a, Flow cytometry analysis of ACE2-GpA expression on 658

/ ACE2-GpA cells pre-differentiation. b, Flow cytometry analysis of enucleated 659

Enucleated RBCs expressed CD71 and did not 660 stain for the nuclear dye DRAQ5. c, Flow cytometry measurement of ACE2-GpA expression 661 on CD71+/DRAQ5-BEL-A / ACE2-GpA cells post-sorting on day 8 of differentiation. d, In 662 vitro neutralization assay against lentivirus-based SARS-CoV-2 pseudovirus comparing 663 control RBCs (black) and ACE2-GpA-RBCs (green)