An ultrasensitive planar array p24 Gag ELISA to detect individual HIV-1 viral particles and infected cells

: Human Immunodeficiency (HIV-1) antiretroviral therapy (ART) has halted the development of curative strategies. Measuring HIV persistence is complex due to the low frequency of cells containing virus in vivo . Most 15 of the commercially available assays to date measure nucleic acid. These assays have 16 the advantage of being highly sensitive and allow for the analysis of sequence diversity, 17 intactness of the HIV genome or evaluation of diverse RNA species. However, these 18 assays are limited in evaluating translational competent viral reservoirs. In here, we 19 developed an ultrasensitive p24 ELISA that uses the Simoa TM planar array technology 20 that can detect as low as a single HIV-1 particle and a single HIV-1 infected cell. 21 Furthermore, the assay is optimized to measure very low levels of p24 in different 22 biological fluids without a major loss of sensitivity or reproducibility. Our results 23 demonstrate that the ‘homebrew’ planar p24 ELISA immunoassay is a broadly applicable new tool to evaluate HIV persistence in diverse biological fluids. 25


INTRODUCTION 26
Human immunodeficiency virus-1 (HIV-1) persistence is responsible for the increase in  Fig. 2F). This result suggest that the background IV 174 units associated with the matrix could be a strong driver in the sensitivity of the planar 175 array assay. One of the potential explanations of the high background could be due to 176 non-specific binding of matrix proteins, the detector antibody, or the streptavidin HRP. To 177 address whether the background associated with the matrixes could be reduced, and the 178 sensitivity of the assay increased, we introduced a blocking step using sample diluent 179 supplemented with 5% milk either prior to or after incubation of the matrix. We decided to 180 use milk as our previous results indicated that breast milk does not interfere with the 181 assay ( Fig. 2 and Extended Data Fig. 2). Albeit we did not appreciate a background 182 reduction when the step was added prior to the matrix (Fig. 5A), we did observe a 183 significant background reduction with different plasma matrixes when the blocking step 184 was performed after the matrix incubation (Fig. 5B). We next addressed whether this 185 reduction in matrix background could improve the sensitivity of the assay. We prepared 186 the p24 standard in either sample diluent or K3EDTA plasma and performed the assay 187 introducing the blocking step after incubation with the matrix with either sample diluent 188 alone or with sample diluent supplemented with 5% milk. The additional blocking step did 189 not interfere with the assay using sample diluent as matrix for the standard (Fig. 5C-D). 190 Interestingly, adding the extra blocking step with milk improved the LOD 3.5 fold. The 191 improvement in the sensitivity of the assay was more evident for K3EDTA plasma,192 reducing the LOD 13-fold to the low fg/ml, without interference in the reproducibility of the 193 assay ( Fig. 5E-F). In conclusion, the addition of a blocking step after matrix incubation 194 can improve the sensitivity of the homebrew planar SP-X p24 ELISA. 195 196

Validation of the homebrew Simoa planar p24 ELISA 197
Finally, we wanted to validate whether the assay can detect both viral particles as well as 198 HIV-infected primary CD4T cells. First, viral stocks of the JR-CSF viral strain of HIV-1 199 were quantified using qPCR. A serial dilution of the viral stock was made from 10 7 to 1 200 HIV copies/ml in assay diluent and the levels of p24 were quantified using the optimized 201 homebrew Simoa planar p24 ELISA. There was a strong linear correlation between the 202 levels of HIV p24 and the levels of HIV RNA with detection above the LOD between 10-203 100 HIV copes per ml (Fig. 6A). To further investigate the limit of detection of HIV RNA 204 copies, we evaluated a range of concentrations spanning 1 to 70 copes/ml. We could 205 clearly detect HIV p24 above the LOD and 3xLOD with as low as 60 HIV RNA copies/ml 206 (Fig. 6B). Since each virion contains 2 HIV RNA copies, our results suggest we could 207 detect as low as 30 virions/ml or 1.5 viral particles in 50 µl of assay diluent, similar to our 208 previous estimation of 1.18+/-0.98 using purified protein (Fig. 1B). 209 Next, we wanted to explore whether this assay could be compatible using cell extracts. 210 This will also allow to measure translational competent virus directly in cells. We first 211 evaluated whether the assay was compatible with a cell lysis buffer containing Nonidet 212 P-40 (NP40), EDTA, Tris and NaCl (NETN). We diluted the HIV p24 standard directly in 213 NETN using a range of concentrations spanning from 100 pg/ml to 6.4 fg/ml in triplicates 214 and a blank control. Interestingly, the NETN did not affect the sensitivity or performance 215 of the assay (Fig. 6C). Next, we infected primary CD4T cells with the replication 216 competent viral stain of HIV-1 JR-CSF and infection was evaluated using flow cytometry 217 ( Fig. 6D). A 7-fold serial dilution of infected cells was done in uninfected primary CD4T 218 cells. A subset of cells in each dilution were stained and 100,000 alive cells were collected 219 and the percentage of infection was evaluated using flow cytometry. The rest of the cells 220 were lysed and the levels of p24 were evaluated using the homebrew Simoa planar p24 221 ELISA. We calculated the equivalent infected cells per million CD4T cells using the 222 percentage of infection in the infected culture and plotted that versus each of the analyses 223 ( Fig. 6D). Flow cytometry allows the detection above the LOD between 501 and 71 224 infected cells per million (Fig 6E). On the other hand, cell lysates equivalent to 500,000 225 cells in 50 µl allows the detection of at least 1.5 infected cells per million CD4T cell (Fig.  226   6F). Importantly, cell extracts of uninfected cells did not show p24 levels above the LOD, 227 indicating a low cross-reactivity of the assay with cellular proteins (Fig. 6F). We then 228 performed serial dilutions of the cell extracts in NETN buffer. We could easily detect as 229 low as 10 and 71 infected cells per million when using extracts of only 50,000 cells or 230 5,000 respectively (Fig. 6F). Finally, we normalized the levels of p24 to the total amount 231 of protein calculated using a bicinchoninic acid (BCA) protein assay. We observed a 232 strong correlation between the levels of p24 protein and the levels of infected cells at the 233 three dilutions with detection of 1.5 infected cells above uninfected cells using cell extracts 234 of either 500,000 or 50,000 cells (Fig. 6G). 235 In conclusion, we validated the homebrew Simoa planar p24 ELISA and demonstrated 236 that the assay can detect as low as a single HIV-1 viral particle or an individual infected 237 CD4T cell. This assay has several advantages over the previous one. First, this new assay allows 252 the detection of HIV p24 directly in different biological matrixes without appreciable loss 253 of sensitivity or reproducibility, including breast milk, CSF and cell lysates. Second, it uses 254 less volume and it does not require further manipulation of the sample. Third, it allows 255 13 higher accuracy of quantification by using the same matrix to prepare the standard than 256 the sample. As such, this assay could be used to evaluate cure strategies such as "shock-257 and-kill" to measure HIV-1 Gag expression in diverse biological fluids and cell lysates 258 after LRA administration using limited amount of sample. Our study also revealed that the 259 anticoagulant used to obtain plasma can have a major influence in the sensitivity of the 260 assay. Based on our studies, K2EDTA plasma will be the recommended if no further 261 manipulation of the sample such as dilution or precipitation can be performed 19 . 262 This assay can also be performed directly using tissue culture media. not tested and further development will be required. The capture antibody used 280 recognizes HIV-2 and other detector antibodies could be used to specifically detect HIV-281 2. Finally, the assay loses sensitivity with those matrixes that have a high background 282 probably due to non-specific binding of either the detector antibody or the streptavidin 283 HRP to the matrix. Interestingly, adding an extra blocking step can improve the sensitivity 284 of the assay in some of those instances. However, only 5% milk was tested and other 285 blocking reagents or concentration of the blocking reagent may also improve the 286 sensitivity of the assay. Further optimization to mitigate the matrix effect will be required 287 on those instances in which the matrix may interfere with the assay and the additional 288 blocking step does not improve it. In spite of these caveats, here we demonstrate that the 289 homebrew planar SP-X p24 ELISA could be a potential new tool and assay that can be 290

Reagents 295
The following matrixes were obtained from Innovative Research: Pooled Human Plasma 296

Generation of capture and detector antibodies 313
The conjugation of the capture and detection antibody with homebrew reagents was 314 performed by following manufacturers protocol. Brielfy, capture and detection antibodies 315 were generated through a series of washes in Amicon filters to exchange the antibody 316 buffer to the assay conjugation buffer as indicated in the Simoa Planar Array Homebrew 317 Starter kit. Briefly, the concentration of each antibody was measured using Nanodrop 318 Spectrophotometer and the concentration adjusted to 1 mg/mL in conjugation buffer. The 319 capture antibody was first tagged using Sulfo-SMCC for 30 minutes then 3.65 µL of 320 Ultrapure Tris-HCl buffer was added to stop the reaction. Then, the Simoa Planar Array 321 Homebrew Tag 1 was added for 30 minutes at room temperature. The detection antibody 322 was incubated with NHS-PEG4-Biotin for 30 minutes at room temperature. Both capture 323 and detection antibodies were then purified to remove excess Sulfo-SMCC, Tag 1, and 324 Biotin through a series of washes with an Amicon filter. The concentration of each 325 antibody were measured using Nanodrop Spectrophotometer and the concentration was 326 adjusted to 0.25 mg/mL in the Simoa Planar Array Homebrew Diluent A. The capture and 327 detection antibodies are then stored at 4ºC for at least one month until ready to use. 328

Planar array assay procedure 329
Prior to performing the assay, the 25X wash buffer was diluted to 1X with ddH20. The Peroxide and 3 mL SuperSignal Luminol Enhancer and 50 µL was added to the wells. 349 The plate was immediately read on the SP-X Imager. 350 In some instances, an additional blocking step was added. After incubation of the plate 351 with the matrix and washing four times, 50 µL of 5% non-fat dry milk in sample diluent 352 (filtered through a 0.22 µM) was added to the wells and incubated for 30 minutes on a 353 shaker at 515 rpm. The plate was then washed four times and patted dry to remove 354 excess wash buffer from wells and the assay continues as described above. 355

Generation of HIV infected primary CD4T cells and cell lysates 387
Total CD4 T cells were isolated from HIV-1 negative blood donors using magnetic 388

SP-X Imager and analysis 443
The plate was scanned and analyzed using the Quanterix SP-X Imager. The data file was 444 analyzed using the Quanterix SP-X Analysis Software where a standard curve using 5 445 parametric logarithmic (5PL) curve fit, %CV, LOD, LLOQ, and R 2 were created.  50,000 or 5,000 cells were quantified using a standard curve using a range of 552 concentrations from 100 pg/ml to 6.4 fg/ml prepared in NETN buffer. Unpaired t test was 553 used to calculate p values relative to uninfected (*<0.05,**<0.01;***<0.001****<0.0001). 554 (G) Correlation between the levels of p24 per µg of total protein and the equivalent 555 infected cells per million CD4 was calculated using Pearson correlation. 556 Figure 1 Detection of single viral HIV particle using the homebrew Simoa planar array p24 ELISA. (A) Simoa Planar Homebrew overview. Created with Biorender.com. (B) Standard curves of 12 independent experiments using a range of concentrations from 100 pg/ml to 6.4 fg/ml of p24 prepared in homebrew sample diluent. Each standard was done in technical triplicates. Data has been transformed by subtracting the IV units from the matrix (0) and calculating the log10 from both the IV units and the p24 concentration. Low limit of detection (LOD) for each experiment is provided as table. (C) % of the coe cient of variation (%CV) intra-assay for each concentration for the 12 independent experiments. (D) % of the coe cient of variation (%CV) inter-assay for each concentration. (E) Capture and detector antibody pair batch effect on the LOD. (F) Correlation of the time after capture and detector antibody pair preparation for Lot#A and the LOD, calculated using Pearson correlation. Figure 2