The outline to generate huNSGs/IBMI-CD133 is shown in Figure 2 . All animal experiments were performed in accordance with the guidelines established by the animal research committee of Tokushima University. All huNSGs/IBMI-CD133 used in this study are listed in Table 1 .

Preparation for CD133+ cells CD133+ cells from umbilical cord blood cells contain both CD34+ and CD34- cell populations, and thus are more primitive cells (36, 37). It has been shown that CD133 can be a marker for CD34+ and CD34- cells (33, 34). It is difficult to judge whether CD133+ selection has a relevant advantage over CD34+ selection (16, 38–40). It appears that there is not a large difference in the cell expansion including CD34+ cells and the differentiation of human lymphocyte cells in an individual mouse, although CD133+ cells have been suggested to display slightly higher plasticity relative to CD34+ cells (16, 17). Based on these previous studies, we decided to use CD133+ cells in our experiments because of the possibility of better cell differentiation and primitivity of CD133+ cells. CD133+ cells for transplantation were isolated from UCB, obtained from the Japanese Red Cross Kinki Cord Blood Bank (Osaka, Japan) for research purposes, similarly to as described previously (35). Around 100–150 mL of UCB with CPDA (citrate-phosphate-dextrose-adenine) was diluted with PBS (-) containing 2 mM EDTA up to 200 mL. After incubation for 15 min at room temperature, the diluted UCB samples were centrifuged at 500 × g for 15 min at 20°C (slow acceleration and deceleration). Buffy coat samples were collected in one tube, PBS (-) containing 2 mM EDTA was added to make a total volume of 30 mL, and the samples were layered on 20 mL of Ficoll-Paque PLUS (GE Healthcare, USA). After centrifugation at 500 × g for 40 min at 20°C (slow acceleration and deceleration), a layer containing mononuclear cells was collected, PBS (-) containing 2 mM EDTA was added to make a total volume of 50 mL, and centrifuged at 400 × g for 10 min at 20°C (slow acceleration and deceleration). The mononuclear cells were washed twice by the addition of 30 mL of PBS (-) containing 2 mM EDTA and centrifugation at 200 × g for 10 min at 20°C. The cells (around 15–30 × 10⁷) were resuspended in 500 μL of CELLBANKER 1 (ZENOGEN PHARMA, Japan) and stored at -80°C until use. From our experience, the cells could be stored for at least 6 months to 1 year. Subsequent preparation of CD133+ cells was performed on the day of transplantation into mice. The stored mononuclear cells were immediately thawed in a water bath at 37°C and transferred to a conical tube. Around 10 mL of the thawing medium [RPMI-1640 containing 20% heat-inactivated (hi) FBS, 100 U/mL Heparin Sodium MOCHIDA (MOCHIDA PHARMACEUTICAL, Japan) and 100 U/mL DNase I (FUJIFILM Wako Pure Chemical, Japan)] was slowly added to the mononuclear cells. After centrifugation at 500 × g for 5 min at 4°C, RPMI-1640 containing 2% hi FBS was added to the mononuclear cells and then centrifuged at 500 × g for 5 min at 4°C to wash the cells. The mononuclear cells were resuspended in 350 μL of PBS (-), and CD133+ cells were isolated from the mononuclear cells using a CD133 MicroBead Kit (Miltenyi Biotec, Germany) in accordance with the manufacturer’s instruction.

IBMI protocol Female 5-week-old NSG mice were purchased from The Jackson Laboratory Japan, Inc. (Kanagawa, Japan). The mice were maintained under germ-free sterilized conditions. The NSG mice were irradiated with 2.5 Gy radiation one day before transplantation, and were injected with CD133+ cells (more than 5.0 x 10⁴) using the IBMI method, as described previously (21, 22, 35) ( Figures 2 , 3 ; Supplemental Movie ). The mice were anesthetized using medetomidine hydrochloride (0.75 mg/kg, Nihon Zenyaku Kogyo, Japan), midazolam (4.0 mg/kg, Sando, Japan), and butorphanol tartrate (5.0 mg/kg, Meiji Seika Pharma, Japan). After shaving the leg hairs of the anesthetized mice ( Figure 3A ), to make a hole to transplant CD133+ cells into the bone marrow, the position of the joint was checked ( Figure 3B ) and the shaved leg was disinfected with 70% ethanol ( Figure 3C ). The tibia was punctured while rotating a 27G × 3/4” needle (Terumo, Japan) ( Figure 3D ). If the needle enters bone marrow properly, the needle does not move or sway even if the hand is released. The needle was then slowly removed from the tibia without moving the hole. To transplant CD133+ cells into the bone marrow, cells were resuspended in RPMI1640 without FBS and drawn into a syringe with a needle (MS-50 micro syringe, Ito Corporation, Japan) ( Figure 3E ). After insertion of the syringe into the pre-punctured hole, the cells (5 μL/mouse) were slowly injected into the bone marrow over about 1 min while rotating the handle of the syringe. About 30 sec after transferring the cells, the syringe was slowly pulled out. The hole was pinched with fingers for around 30 sec, and disinfected with iodine ( Figure 3F ). The mice were administered an anesthetic antagonist (antisedan, Nippon Zenyaku Kogyo, Japan) and returned to the cage. Since the teeth of mice transplanted with huCD133+ cells often grow longer preventing proper feeding, caution should be taken to observe their teeth and cut them if necessary.

As indicated in Figure 2 , peripheral blood samples of huNSGs/IBMI-CD133 were periodically collected from the tail vein into EDTA-treated hematocrit glass capillary tubes (Marienfeld Superior, Lauda-Königshofen, Germany). When necessary, samples were treated with β-propiolactone and ultraviolet irradiation to inactivate viruses, as previously described (41). Briefly, 10 μL of whole blood samples were mixed with 1 uL each of 1M Tris-HCl (pH 7.50) and 10% β-Propiolactone, and then exposed to UV light (spectral power 254nm, 39W) (Atlantic Ultraviolet, USA) for 10 min at 30 cm distance. To stain cell surface markers, PE anti-mouse CD45 [30-F11], APC/Cy7 anti-human CD45 [HI30], APC anti-human CD3 [HIT3a], PerCP/Cy5.5 anti-human CD8a [HIT8a], FITC anti-human CD4 [RPA-T4], and PE/Cy7 anti-human CD19 [HIB19] (Biolegend, USA) antibodies were used. Whole blood samples (10 μL) were mixed with 50 μL of PBS containing 3% hi FBS and appropriate antibodies (0.5 μL each), and incubated at room temperature for 30 min. To determine absolute cell numbers in the blood samples, AccuCount Ultra Rainbow Fluorescent Particles (ACRFP-100-3, Spherotech, USA) were used in accordance with the manufacturer’s instruction. Before FACS analysis, erythrocytes in the samples were lysed by treating with RBC lysis buffer (144 mM NH₄Cl and 17 mM Tris [pH 7.65]) for at least 5 min at room temperature. The stained cells were analyzed using BD FACSVerse and FACSuite software (Becton Dickinson, USA).

HIV-1 proviral clones NL4-3 (42) and its R5-tropic derivative NL-A4Y1 were used in this study. NL-A4Y1 clone was generated by introducing the env gene of gtu+A4CI1 clone (43), derived from a clinical isolate of subtype B, into the corresponding region of NL4-3 through overlapping PCR. NL-A4Y1-ΔNef clone, a Nef-negative frameshift mutant, was constructed using the XhoI site in the nef gene, as described previously (44, 45).

A human monolayer cell line 293T (46) and a luciferase reporter cell line TZM-bl (47, 48) were maintained in Eagles’s minimal essential medium (EMEM) containing 10% hi FBS. A human lymphocyte cell line MT4/CCR5 (MT4 cells stably expressing CCR5) was maintained in RPMI-1640 medium containing 10% hi FBS and 200 μg/mL hygromycin (Merck, Germany). Human PBMCs were purchased from Cellular Technology Ltd. (USA). Cryopreserved PBMCs were thawed in accordance with the manufacturer’s instruction and used for virus replication assays.

Virus stocks for infection were prepared from 293T cells transfected with proviral clones by the calcium-phosphate coprecipitation method, as described previously (42, 49). Virion-associated reverse transcriptase (RT) activity was measured as described previously (50, 51). Prior to infection, PBMCs were stimulated with RPMI 1640 medium containing 10% hi FBS, 50 units/ml of recombinant human IL-2 (Bio-Rad, USA), and 2 μg/ml of phytohemagglutinin L (PHA-L) (Roche Diagnostics GmbH, Germany) for 3 days. MT4/R5 and stimulated PBMCs were spin-infected (52) with equal amounts of viruses. Infected PBMCs were cultured in the presence of IL-2 throughout the experiment. Viral growth was monitored by RT activity in the culture supernatants.

Virus stocks were prepared as described above. The titer of input viruses was determined by calculating TCID₅₀ using infectivity to TZM-bl cells (53). Briefly, virus stocks with 10-fold serial dilution were inoculated into TZM-bl cells, and the cells were lysed and subjected to luciferase assays on day 2 post-inoculation (Promega, USA). TCID₅₀ values were calculated by the Reed-Muench method. Virus stocks were diluted with EMEM or saline, and the diluted viruses (a total volume of 100 μL) were inoculated intraperitoneally into anesthetized IBMI-huNSG mice 6–8 months after the cell transplantation. Peripheral blood samples (around 100 μL) were periodically collected from the tails of mice ( Figure 2 ) and used to monitor virus replication and absolute CD4 cell numbers. Virus replication was assessed by measuring viral RNA copy numbers in the plasma samples. Viral RNA genome in the plasma sample (25 μL) was automatically extracted using QIAamp Viral RNA Mini Kit (QIAGEN, Netherlands) by QIAcube (QIAGEN, Netherlands), and subjected to real-time RT-PCR analysis. The RT-PCR reaction was performed using a specific primer set qPCR Set3 F (caggccatatcacctagaactt)/qPCR Set3 R (gggtggctccttctgataat) and QuantiFast SYBR Green RT-PCR Kit (QIAGEN) in accordance with the manufacturer’s instruction. The molecular clone pNL4-3 (42) digested with EcoRI was used as a standard to quantify viral RNA copy numbers.

The time schedule for the HIV-1 infection experiments using huNSGs/IBMI-CD133 is shown in Figure 2 . NSG mice were transplanted with UCB-derived CD133+ cells by IBMI on the next day after X-ray irradiation ( Figure 3 ; Supplemental Movie ) ( Table 1 ). Blood samples were collected periodically to monitor the development of human hematopoietic cells by FACS ( Figure 4A ). Three months after the transplantation, the human leukocyte population (CD45+ cells) was around 30% of the whole lymphocyte cell population and increased to approximately 40% at 6–8 months post-transplantation. The percentage of human B cells (CD19+ cells) was higher than that of human T cells (CD3+ cells) at 3 months post-transplantation (more than 80% vs less than 10%, respectively) ( Figure 4B ). While the percentage of human B cells in the human lymphocyte population was gradually decreased, the percentage of human T cells was significantly increased (~50%) up to a similar level to that of human B cells at 6–8 months after the transplantation ( Figure 4B ). This result is in good agreement with previous reports that human B cells develop earlier than human T cells in non-BLT humanized mice (17, 54). The percentage of CD4+ cells in the human T cell population was higher than that of CD8+ cells at 6–8 months post-transplantation ( Figure 4C ) and the absolute number of CD4+ cells was increased up to around 200 cells/uL blood. Consistent with a previous report (35), stable T cell numbers and B-cell-to-T-cell ratio were maintained in huNSGs/IBMI-CD133 for a long time (6–8 months after the transplantation).

Using twelve huNSGs/IBMI-CD133 ( Table 1 ), we assessed HIV-1 replication in individual mice. X4-tropic HIV-1 (NL4-3), R5-tropic HIV-1 (NL-A4Y1), and its nef-deletion mutant (NL-A4Y1ΔNef) were used for the infection experiments in this study. The NL-A4Y1 clone carries an env gene derived from a clinical isolate (subtype B). HuNSGs/IBMI-CD133 6–8 months after the transplantation were used for HIV-1 infection, because the percentage of T cells in the human lymphocyte population was higher than that of B cells and the CD4+ cell number was stably maintained ( Figure 4 ). Virus samples prepared from 293T cells transfected with proviral clones were intraperitoneally inoculated into huNSGs/IBMI-CD133. Blood samples were periodically collected, and the HIV-1 replication and the CD4+ cell number were measured using quantitative RT-PCR and FACS, respectively. Of the twelve huNSGs/IBMI-CD133 infected, three died in a short period of time, within 7 weeks, from rapid increases in CD4+ cells (n=2) or HIV-1 (n=1) ( Table 1 ). In our experience, approximately 6% of mice transplanted with CD133+ cells have died of unidentified causes after transplantation, possibly because of some infections and/or stresses caused by transplantation such as irradiation, anesthesia, blood collection, or graft versus host disease. As shown in Figure 5 , the X4-tropic HIV-1 (NL4-3) viremia peaked at 5 to 10 weeks post-inoculation accompanied by a rapid reduction in the CD4+ cell number upon or just after the peak. On the other hand, upon the inoculation of the same dose of viruses (10,000TCID₅₀), the plasma viral load with R5-tropic HIV-1 (NL-A4Y1) reached a peak at 2 to 3 weeks post-inoculation ( Figures 6A, B ). The viremia was maintained 4 to 6 months after the virus-inoculation with a concomitant gradual decrease in the CD4+ cell number, repeatedly increasing and decreasing ( Figures 6A, C ). Viral replication kinetics in mice inoculated with a lower dose of viruses (1,000TCID₅₀) were similar to those in mice inoculated with a 10,000TCID₅₀ dose of viruses ( Figure 6A ). The peak viremia and replication kinetics in the plasma of huNSGs/IBMI-CD133 individuals were not related to the percentage of human leukocyte reconstitution and the CD4+ cell number at the HIV-1 inoculation time point ( Table 1 ; Figures 5 , 6 ). Since p24-positive cells have been detected in lymphoid organs (23, 24), various sources of viruses may affect the plasma viral load. Considering the stable maintenance of the CD4+ cell number for at least 10 months after transplantation (our unpublished results) and the gradual decrease in the CD4+ cell number associated with fluctuations in viral growth ( Figure 6 ), it is reasonably assumed that this decrease was not due to a spontaneous decrease in CD4+ cells in infected huNSGs/IBMI-CD133. These results show that HIV-1 replicates and persists well for a long time in huNSGs/IBMI-CD133.

It has been shown that Nef-deleted or -inactivated HIV-1 displays an attenuated phenotype in humanized BLT mice, such as the decrease in plasma viral load and the weak decline in CD4+ cells (55–57). Using our hu/NSGs/IBMI-CD133 model, we investigated the effect of Nef deletion on HIV-1 replication in vivo. Our Nef-negative R5-tropic HIV-1 (NL-A4Y1ΔNef) was newly constructed in this study by introducing a frameshift mutation into the viral nef gene. First, we compared the growth ability of NL-A4Y1ΔNef and its parental NL-A4Y1 in human lymphocyte cells. As shown in Figure 7 , these two clones appeared to exhibit similar growth kinetics in an MT4/R5 cell line, although NL-A4Y1ΔNef reached a peak slightly later than NL-A4Y1. Nef-deleted HIV-1 mutants have been reported to be attenuated in their ability to replicate in stimulated PBMCs compared to those of their parental clones (58, 59). To examine the effect of Nef-deletion on the viral replication ability, virus samples (NL-A4Y1 and NL-A4Y1ΔNef) were inoculated into human PBMCs stimulated with PHA-L. In two different preparations of PBMCs (PBMC1 and PBMC2 in Figure 7 ), NL-A4Y1ΔNef grew poorly compared to NL-A4Y1. The degree of decrease in the viral replication ability of the Nef-deleted clone appeared to be different in different cell types (MT4/R5 cells vs PBMCs) and in distinct preparations of PBMCs (PBMC1 vs PBMC2). This may be due to the difference in cellular physiological states and/or some specific cellular factors such as CD4 and SERINC3/5 (60, 61). Next, NL-A4Y1ΔNef virus samples were intraperitoneally inoculated into huNSGs/IBMI-CD133. As shown in Figure 8 , the plasma viral load by the Nef-deleted clone reached a peak later than 8 weeks post-inoculation. In huNSGs/IBMI-CD133 inoculated with the same dose of viruses (10,000TCID₅₀), NL-A4Y1ΔNef virus exhibited delayed growth kinetics compared to that for the parental NL-A4Y1 ( Figures 6A , 8A, C ). This may reflect the lowered replication ability of Nef-deleted HIV-1 in PBMCs, although further experiments need to be done. The activity/function of Nef in HIV-1 replication in vivo has been investigated using humanized BLT and non-BLT mice (55–57, 62). HuNSGs/IBMI-CD133 also may be useful for investigating the activity/function of the four accessory proteins of HIV-1 in vivo.