Coreceptor Function of Mutant Human CD4 Molecules without Affinity to gp120 of Human Immunodeficiency Virus*

Despite extensive mutational studies on the human CD4 molecule and its affinity to human immunodeficiency virus (HIV) envelope glycoprotein gp120, coreceptor functions of such mutant molecules have only been examined by indirect measurement of their affinity to class II major histocompatibility complex (MHC) molecules. In this report, coreceptor functions of mutant human CD4 molecules, which have no or reduced affinity to an HIV envelope protein, gp120, were assessed in a murine T cell receptor/class II MHC recognition system. The substitution of human C" β strand with the murine homologous segment resulted in the loss of the coreceptor function as well as in the complete loss of gp120 binding capacity, corroborating the consensus that Phe-43 in C" β strand plays crucial roles in both situations. However, simultaneous replacement of the C′-C" loop along with the C" β strand by homologous murine segments rescued the coreceptor function, whereas gp120 binding capacity remained negative. Further analysis indicated that insertion of lysine between Gly-41 and Ser-42 can partially compensate for the coreceptor function lost by the Phe-43 → Val mutation. Although the coreceptor function of these mutant CD4 molecules in a human T cell recognition system is yet to be determined, these observations necessitate a re-evaluation of the role played by Phe-43 in coreceptor function. Examination of the sensitivities of the mutant CD4 molecules expressed on HeLa cells to infection by a T cell-tropic HIV-1 strain indicated that only those mutants that had completely lost gp120 binding capacity were resistant to the infection. All mutants having whole C" substitution, irrespective of additional substitutions or their coreceptor functions, were resistant to the infection.

The most serious consequence of the human immunodeficiency virus type 1 (HIV-1) 1 infection is the acquired immuno-deficiency syndrome characterized by severe deficiency of CD4 ϩ T cell functions. Such deficiency of CD4 ϩ T cells stems from the affinity of an HIV envelope protein, gp120, to the human CD4 molecule (1). Several etiological mechanisms causing the functional deficiency of CD4 ϩ T cells are involved; 1) targeted infection of CD4 ϩ cells initiated by the binding of the virus to CD4, 2) cell fusion induced by binding of HIV viruses resulting in apoptosis of CD4 ϩ T cells, and 3) inhibition of the specific recognition of antigenic peptide/class II major histocompatibility complexes (MHCs) by soluble gp120. Extensive mutational analyses on the human CD4 molecule have identified a CЈ-CЉ ridge of the D1 domain as the gp120 binding area (2)(3)(4). The murine CD4 molecule has no affinity to gp120, although it shares 55% sequence homology of the extracellular portion with its human counterpart. Segmental as well as single amino acid substitutions of the human CD4 molecule with murine sequences confirmed the crucial role played by an aromatic ring of Phe-43 in gp120 binding (5).
Several reports have claimed that the loss of gp120 binding ability of the human CD4 molecule is unavoidably accompanied by the loss of its affinity to class II MHC molecules, which is essential for its coreceptor function (6,7). Nevertheless, these studies employed adhesion assays that did not involve specific recognition of MHC molecules by T cell receptor (TCR). The CD4 molecule plays an important role as a coreceptor upon specific recognition of a class II MHC molecule complexed with an antigenic peptide by the TCR. The CD4 molecule has been shown to perform its coreceptor function by forming a TCR⅐MHC⅐CD4 ternary complex contacting with a highly conserved portion of the ␤2 domain of the target class II MHC molecule (8 -10) and probably with TCR⅐CD3 complex laterally (11). The formation of the ternary complex is believed to raise the total avidity of the TCR⅐MHC interaction and to trigger cytoplasmic signal transduction by bringing a tyrosine kinase p56lck associated with the cytoplasmic portion of the CD4 into the vicinity of the TCR⅐CD3 complex for phosphorylation of CD3 chains (12)(13)(14). In this report, we have examined coreceptor functions of the mutant human CD4 molecules. The coreceptor function was evaluated as the ability of the CD4 molecule to assist the recognition of murine class II MHC antigen (I-A k ) by murine allo-specific T cell hybridoma. The study confirmed that the mutations causing the loss of gp120 binding, including the substitution of Phe-43 to Val, simultaneously resulted in the loss of the coreceptor function. Further studies, however, revealed that the basis for the roles played by Phe-43 in gp120 binding and in the coreceptor function assessed in the murine system are different.

EXPERIMENTAL PROCEDURES
Mutant CD4 cDNA and Plasmids-Mutagenesis of human CD4 using single stranded DNA and synthetic oligonucleotides with mutant sequences was performed according to Kunkel's method as described previously (5). Table I enlists mutant human CD4 molecules studied in this report with their amino acid sequences. The wild and mutant human CD4 cDNAs were cloned into expression vectors pEneoAf Ϫ and pMkitneo. The plasmids were purified by equilibrium centrifugation in CsCl-ethidium bromide gradients and used for transfection.
Transfection-Purified plasmid (40 g) was transfected into parental 4QBW cells (5 ϫ 10 6 ) by electroporation. Stable transfectants were selected by culturing in medium containing 1 mg/ml G418, and transfectants were cloned by limiting dilution at 0.3 cells/well and screened by staining with fluorescein isothiocyanate (FITC)-labeled OKT4 antibody. All mutants with a Phe-43 3 Val substitution lost the reactivity to Leu3a and mutant 18 (Arg-59 3 Glu) did not react with OKT4A (Table II). None of the mutants gained reactivity to GK1.5 (data not shown).
Cells and Culture-The murine T cell hybridoma 4QBW was generated by fusion between a murine CD4 ϩ T cell clone 4Q11 and murine thymoma BW5147. 4Q11 was derived from a TCR-transgenic mouse expressing TCR ␣␤ chains of an allogeneic I-A k -specific T cell clone QM11 (15). TA3 was an I-A k -positive B cell hybridoma TA3 (H-2 d/a ) originally produced by Glimcher (16). The cells were maintained in Dulbecco's modified essential medium supplemented with 5% heatinactivated fetal calf serum, nonessential amino acids, sodium pyruvate, 2-mercaptoethanol, and 10 mM HEPES buffer.
Antibodies-Antibody inhibition studies were performed using monoclonal antibodies (mAbs) to I-A k ␤ chain (10-2-6, mouse IgG2b), to the mouse CD4 (GK1.5, rat IgG2b) (17), and to human CD4 (OKT4 and OKT4A; mouse IgG2a, and Leu3a; mouse IgG1) (18,19). TCR expression of mutant CD4 transfectant 4QBW cell lines was determined using biotinylated mAb specific for the idiotype of QM11 (ID11, mouse IgG1) (15) followed by addition of phycoerythrin (PE)-conjugated avidin. PElabeled anti-mouse CD4 (RM4 -5, rat IgG2a) and unconjugated Leu3a were purchased from Pharmingen (San Diego, CA) and Becton Dickin-son Immunocytometry Systems (San Jose, CA). FITC-labeled as well as nonlabeled OKT4 and OKT4A were purchased from Ortho Diagnostic Systems (Raritan, NJ). FITC-labeled goat anti-mouse IgG Fc portion (Organon Teknika, PA) was used for secondary staining. Biotin-conjugated gp120 was prepared in our laboratory and used with PE-avidin as a secondary reagent. For negative controls, cells were stained with the FITC-labeled goat antibody to mouse IgG Fc portion alone or PE-avidin alone. The gp120 binding capacity of mutant CD4 molecules was determined as relative binding index (RBI) as described previously (5). Briefly, BW5147 cells expressing transfected mutant CD4 were stained with FITC-OKT4 and biotin-gp120ϩPE-avidin separately. RBI was calculated as: where M e and M c stand for mean fluorescence values of experimental and control staining, respectively, FL1 for staining with OKT4, and FL2 for staining with gp120. The staining of wild type CD4 was included in every set of experiments for the calculation. IL-2 Assay and Stimulation of Transfectant T Hybridoma-In the coreceptor function assay, responder cells (2 ϫ 10 4 ) were cocultured with 5 ϫ 10 4 TA3 cells in a total volume of 200 l on 96-well U-bottomed microtiter plates. After 14 h of incubation, culture supernatants were harvested. The harvested supernatants were serially diluted with fresh culture medium, and the concentration of murine IL-2 was determined by enzyme-linked immunosorbent assay (ELISA). For murine IL-2 ELISA, capturing mAb (JES6 -1A12, rat IgG2a), biotinylated detecting mAb (JES6 -5H4, rat IgG2b), and peroxidase-labeled streptavidin (Sigma) were used. As a coloring substrate 3,3Ј,5,5Ј-tetramethylbenzidine was used.
HIV Infection Analysis-Sensitivity of mutant CD4 molecules to HIV infection was assessed on CXCR4-positive HeLa cells expressing mutant CD4 generated by calcium phosphate transfection and G418 selection. The expression of the transfected CD4 was determined by staining with FITC-OKT4. A T cell-tropic HIV-1SF2 had been molecularly cloned as previously reported (20). HeLa transfectants were treated with 2 g/ml polybrene (Sigma) for 30 min at 37°C and exposed to 1 ml of culture supernatant containing viruses (p24 antigen at 5 ng/ml) for 2 h at 37°C. Cells were washed and cultured. The levels of p24 core antigen in the culture fluids were determined by ELISA (Abbott, Wiesbaden-Delkenheim, Germany) at 4-day intervals.

Developing an Assay System for Evaluating the Coreceptor Function of Mutant Human CD4
Molecules-Several reports have indicated that the human CD4 molecule could substitute the murine CD4 in restoring the class II MHC-restricted specific recognition of murine T cells. (21)(22)(23). Therefore, attempts were made to develop a murine recognition system in which the coreceptor function of the human CD4 could be assessed. Human wild type CD4 cDNA, in expression vector pMkit, was transfected into a murine CD4 ϩ T cell hybridoma, 4QBW, specific for an allogeneic class II MHC antigen. 4QBW recognized I-A k in a CD4-dependent manner and secreted interleukin-2 (IL-2) in response to the recognition. Several human CD4-expressing clones, simultaneously expressing murine CD4 and TCR in comparable amounts with those of the mother line 4QBW, were isolated. Fig. 1 shows the expression of TCR, murine, and human CD4 by 4Q/wild, which is one of such transfectant clones. 4Q/wild as well as 4QBW produced IL-2 when cocultured with a B cell hybridoma TA3, which expressed I-A k . As shown in Fig. 2A, these responses were inhibited by a monoclonal antibody to I-A k , 10-2-16, confirming the specificity of the response. The efficiencies of the recognition by 4Q/wild and that by 4QBW appeared comparable as judged from the inhibition curves of the responses by graded concentrations of the antibody. Fig. 2B shows the inhibitory effect of GK1.5, an anti-murine CD4 monoclonal antibody, on the recognition of I-A k by 4QBW or by 4Q/wild. The complete inhibition of the IL-2 production response of 4QBW by GK1.5 confirmed the coreceptor dependence of the recognition by TCR of this hybridoma. On the contrary, the recognition of I-A k by 4Q/wild was only partially inhibitable by this antibody. Therefore, in the presence of an excessive amount of GK1.5 (1:10 4 ), graded amounts of monoclonal antibody to D1 domain of human CD4, OKT4A, or Leu3a were added to the mixed cultures of 4Q/wild and TA3. As shown in Fig. 2C, the GK1.5-resistant component of the response of 4Q/wild was inhibitable by these anti-human CD4 antibodies, indicating that wild type human CD4 molecules compensated for the coreceptor function of murine CD4 molecules when they were blocked. These anti-human CD4 antibodies did not have a detectable inhibitory effect on the response in the absence of GK1.5. Similar experiments were performed on three other independently cloned human CD4 transfectant hybridomas with essentially the same results. These results confirmed the earlier reports that the human CD4 could function as a coreceptor in the recognition of murine class II MHC molecules by murine TCR (21)(22)(23). Therefore, this assay system was employed for evaluating the coreceptor function of mutant human CD4 molecules.
Coreceptor Function of Mutant CD4 Molecules-Mutants 1-5 and 18 were examined for their coreceptor function. Mutants 1-5 have segmental replacement of the CЈ ␤ strand, the CЈ-CЉ loop, the CЉ ␤ strand, the CЉ-D loop, and the D ␤ strand, respectively, by murine homologous segments. Mutant CD4 cDNA cloned in an expression vector, pMkit, was transfected into 4QBW, and at least three independent clones with reasonable expression levels of the transgenic CD4, endogenous CD4, and TCR were established for each mutant (data not shown). Antigen-induced IL-2 production responses of these clones were examined, and representative results are summarized in Fig. 3. In these graphs, the relative amount of IL-2 produced by a given transfectant is expressed as the percentage of the amount produced by the same clone in response to antigenic stimulation by TA3 in the absence of inhibitory antibodies (see Fig. 3 legend).
The transfectants expressing mutant CD4 molecules showed variable sensitivities to the blocking effect of an excessive amount of GK1.5, whereas the response of the mother hybridoma line 4QBW was almost completely inhibitable. The responses of mutants 1 and 5 (4Q/M1 and 4Q/M5) showed GK1.5resistant components comparable to those of 4Q/wild, which were inhibitable by additional anti-human CD4 antibodies. Therefore, it was concluded that the coreceptor function of these mutant molecules was not impaired by these mutations. Mutants 2 and 4 showed somewhat reduced coreceptor activities. Mutants 3 and 18 did not show significant GK1.5-resistant responses. The same experiments were repeated with 10 (mutant 3) and 3 (mutant 18) independent clones with essentially the same results. Based on these observations, it was concluded that mutants 3 and 18 were nonfunctional as a coreceptor. A salient correlation between the coreceptor function and gp120 binding ability (Table II and Ref. 5) of these mutant CD4 molecules was observed.
Because mutant 3 had three amino acid substitutions within the CЉ strand, contributions of the individual substitutions were studied. As shown in Fig. 4, Phe-43 was the crucial amino acid for the coreceptor function (mutant 6) and other two amino acid substitutions had no obvious effects (mutant 7 and 8). Phe-43 had also been shown to be the key amino acid in gp120 binding (Table II and Ref. 5). The results of the two experiments apparently confirmed the general correlation between gp120 binding and affinity to class II MHC observed with mutant CD4 molecules (6,7).
Functional Mutants with No gp120 Binding-Because the wild type human CD4 molecule functioned as a coreceptor in the murine recognition system, it was rather puzzling that the substitution of the CЉ strand of a functional human CD4 by the Wild type 100  1. Staining profiles of 4QBW and its human CD4 transfectant 4Q/wild. Human and murine CD4 were stained with FITC-labeled OKT4 and RM4 -5, respectively. I-A k -specific TCR was stained with idiotypic antibody ID11 (15). See legend of Table II for descriptions  of OKT4, OKT4A, and Leu3a. sequence of a fully functional murine CD4 resulted in a functionless molecule. Thus, we speculated that this substitution might have generated a certain structural incompatibility between the murine CЉ strand and surrounding structures of the human CD4 molecule, causing certain local distortion of the ternary structure. Therefore, three additional mutants were produced to test this possibility. In mutant 19, the CЈ-CЉ loop along with the CЉ strand were replaced by the murine counterpart. Mutant 20 has murine sequences in the CЉ strand and the CЉ-D loop. In mutant 21, CЉ and the two flanking loops were replaced by murine counterparts altogether.
The gp120 binding capacities of the three mutants were examined by cell surface staining assay on murine thymoma cell BW5147 transfected by the respective mutant CD4 cDNA FIG. 2. Compatibility of human CD4 as a coreceptor in murine recognition system. 4QBW and 4Q/wild were stimulated with an I-A k -positive B cell line, TA3. The amount of secreted IL-2 in 14-h culture supernatant was determined by ELISA. The amount of IL-2 is expressed as the percentage of the amount produced by the respective cell line stimulated in the absence of inhibitory antibodies. A, IL-2 production responses of 4QBW (squares) and 4Q/wild (triangles). The cell lines were cocultured with TA3 in the presence or absence of serially diluted ascites of anti-I-A k mAb (10-2-6). Average amounts of IL-2 produced in the absence of the antibody were 14 and 9.5 ng/ml for 4QBW and 4Q/wild, respectively. B, failure of anti-mouse CD4 antibody to completely inhibit the specific recognition of 4Q/wild. Serial dilutions of an ascites of GK1.5 (anti-murine CD4 mAb) were added to mixed cultures of 4QBW (squares) or 4Q/wild (triangles) and TA3. C, sensitivity of the GK1.5-resistant component of response of 4Q/wild to antibodies to human CD4. 4Q/wild was cultured with TA3 in the presence of an excessive amount of GK1.5 and serially diluted with Leu3a or OKT4A. These antibodies completely blocked the GK1.5-resistant component of the response of 4Q/wild. (see "Experimental Procedures" and Ref. 5). All three mutant molecules had completely lost gp120 binding ability as indicated by the relative binding index in Table II. The IL-2 production assays were carried out on 3 to 10 transfectant clones for each mutant, and representative results are shown in Fig.  5. Mutants 19 and 21 had coreceptor activities comparable to that of the wild type human CD4 despite their loss of affinity to gp120. Because these three mutants had lost the Leu3a epitope as mutant 3 did, their GK1.5-resistant IL-2 production was not inhibitable by this antibody. Mutant 20 did not function as a coreceptor. These results suggest a requirement of compatibility between CЈ-CЉ loop and CЉ ␤ strand for the coreceptor function.
To investigate possible contributions of the length of the murine CЈ-CЉ loop and the charged amino acid within the loop in compensating the function, Gly-41 was replaced by Lys (mutant 22) or Lys was inserted between Gly-41 and Ser-42 (mutant 23) along with the Phe-43 3 Val substitution (Fig. 6). The insertion of Lys between positions 41 and 42 (mutant 23) partially restored the coreceptor function, whereas mutant 22 (Gly-41 3 Lys, Phe-43 3 Val) had marginal activity. The results indicated that both the length of the loop and lysine did contribute to, but were not sufficient for, the restoration of the coreceptor function.
Sensitivity of Mutant CD4 Molecules to HIV Infection-To examine the sensitivity of the mutant CD4 molecules to HIV infection, the mutant genes were transfected into a chemokine receptor (CXCR4)-positive human cell line, HeLa. The transfected cells were exposed to a T cell-tropic HIV-1 strain, SF2, and cultured. Fig. 7 shows the relative amounts of HIV core antigen p24 in the culture supernatants determined on indicated culture days. The transfectant-expressing wild type human CD4 was sensitive to the infection by SF2. The cells expressing Mutants 2 and 4 were also infectable despite their significant loss of gp120 binding ability, indicating that the residual weak binding activities were sufficient for HIV infection. This conclusion was enforced by another mutant with residual gp120 binding activity, i.e. mutant 6 (Phe-43 3 Val), which also showed sensitivity to the infection (data not shown). On the contrary, mutant 3, which had completely lost the binding activity, did not confer the sensitivity to HIV infection to the transfectant. Mutants 19 and 21, as expected from the total loss of gp120 binding capacity, did not succumb to the infection either.

DISCUSSION
The consensus constructed through extensive mutational analyses of the CD4 molecule has been that all mutations resulting in the loss of gp120 binding capacity were unavoidably accompanied by the loss of their affinity to the class II MHC molecule and consequently the loss of the coreceptor function (6,7). The present report largely confirmed the observations by showing that mutations on the human CD4 molecule causing the loss of the affinity to gp120 simultaneously resulted in the loss of the coreceptor function and that Phe-43 played crucial roles in both situations. However, more careful analysis revealed that the bases for the roles played by Phe-43 in the two functions of the CD4 molecule might be different.
Chen et al. (24) reported that a synthetic mimetic of the  Table I. Coreceptor function was treated with 4 and 10 independent transfectant clones of mutants 22 and 23, respectively, and the figure shows representative results. ␤-turn formed by amino acid residues 40 -43 of the human CD4 was capable of inhibiting the binding of gp120 to human CD4, indicating that this portion of the molecule has direct involvement in binding to gp120. In our previous reports, it was shown that the presence of an aromatic side chain at position 43 was essential for gp120 binding and, therefore, it was speculated that electrons might be involved in the binding (5). It is known that the murine CD4 molecule does not function as a coreceptor in the human TCR/class II MHC recognition system. On the other hand, the fact that the wild type human CD4 molecule functioned as a coreceptor in the murine recognition system indicates that differences in amino acid sequences between human and murine CD4 molecules are not critical in the other direction. Therefore, the loss of the coreceptor function in the murine system, resulting from the substitution of the CЉ strand of a functional human CD4 by the sequence of a fully functional murine CD4, suggests that this substitution generated a certain structural incompatibility between the murine CЉ strand and surrounding structures of the human CD4 molecule, and thereby, generated a local distortion of the ternary structure. The CЈ-CЉ loop of the human CD4 consists of only two amino acids forming a hairpin loop (␤-turn) (Fig. 8). It was speculated that the maintenance of such a structure might somehow be dependent on surrounding structures (25). The homologous loop of the murine CD4 molecule consists of three amino acids with two glycines at both ends making the loop more flexible than the human counterpart. Thus the structure of the murine CЈ-CЉ loop is likely to be much less dependent on surrounding structures. If this is the case, the combination of murine CЉ and human CЈ-CЉ loop might not reconstitute the right ternary structure. The results obtained in this report are consistent with this model.
Another possibility, nevertheless, can also be envisioned. There might be certain differences in the basis of CD4/class II MHC interaction between human and murine molecular pairs, i.e. Phe-43 playing a decisive role in the human pair, whereas, in the murine pair, the interaction being largely determined by the charged side chains of the CЈ-CЉ loop. In fact, the insertion of lysine between Gly-41 and Ser-42 (mutant 23) partially restored the coreceptor function lost in the Phe-43 3 Val mutation. However, the restoration by this insertion was merely partial. The difference between mutants 19 and 23 indicates the involvement of not only the lysine residue but also glycine at position 42, suggesting a requirement for more flexibility at this position. Thus it appears that three factors, i.e. lysine in the middle of the CЈ-CЉ loop, the length of the loop, and serine at position 42, contributed to the restoration of the coreceptor function. Structural analyses are necessary to elucidate the actual structural changes introduced by these mutations. More importantly, coreceptor functions of the mutant molecules have to be assessed in a human TCR/class II MHC recognition system to see whether or not the observations are reproducible in the human system. It may be argued that they wouldn't, because the murine CD4 molecule fails to function in the human system mainly because of the substitution of Phe-43 to Val. However, it is also possible that the failure observed in such experiments was due to the putative structural change resulting from the incompatibility between the amino acid in the 43rd position and surrounding structures rather than to the direct effect.
There has been only a limited number of studies examining the coreceptor function of mutant CD4 molecules. The majority of the studies employed adhesion assays in which binding between CD4-expressing transfectants and class II MHC antigenexpressing cells was measured. We have not been able to develop reliable adhesion assay systems mainly because of the subjectiveness of the assay and the highly variable background of the adhesions. Although the assay system employed in the present report does not allow us to extract pure information on the affinity between MHC class II molecules and mutant CD4 molecules, it provides more comprehensive information concerning the coreceptor function of the CD4 molecule. The CD4 molecule has been shown to perform its coreceptor function by forming a TCR⅐MHC⅐CD4 ternary complex in contact with a  8. The C-C؆ ridge of murine and human CD4. The human structure is drawn according to x-ray crystallography (3,4). The murine structure was drawn by arbitrary alignment to the human structure. Shaded ribbons indicate ␤ strands.
highly conserved portion of the ␤2 domain of the target class II MHC molecule (8 -10) and probably laterally with the TCR⅐CD3 complex (11). The formation of the ternary complex is believed to raise the total avidity of the TCR⅐MHC interaction and to trigger cytoplasmic signal transduction by bringing a tyrosine kinase p56lck associated with the cytoplasmic portion of the CD4 into the vicinity of the TCR⅐CD3 complex for phosphorylation of the chain (12)(13)(14). In one series of studies, the ability of mutant CD4 molecules to assist the recognition of class I MHC molecules by class I-restricted TCR was studied (26,27). In this situation, TCR⅐MHC⅐CD4 ternary complexes should not have been formed. The significance and actual existence of TCR-independent interactions between coreceptor molecules and MHC molecules in physiological cellular interactions is not clear. 2 Therefore, it is essential to evaluate the function of coreceptor molecules in systems involving specific recognition of MHC molecules by TCR. We had to compromise and utilize a murine recognition system for evaluating coreceptor functions of the mutant human CD4 molecules, because our repeated long-lasting efforts to develop human systems have not been successful so far.
The focused insults on CD4 ϩ T cells by HIV would be avoided if CD4 molecules did not have affinity to gp120. Attempts to block the binding of gp120 to CD4 molecules in vivo by soluble CD4 molecules or peptides were hampered by the development of a CD4-independent variant of HIV-1 in patients and by tactical difficulties in maintaining high levels of soluble inhibitors in vivo. The advancement of the gene targeting technology has opened a path to a novel strategy for the reconstitution of the HIV-resistant immune system. If CD4 structural genes of innocent hematopoietic stem cells of an HIV-infected patient were replaced (knock-in) by a CD4 gene encoding a gp120 nonbinding yet functional mutant molecule, the progeny mature CD4 ϩ T cells arising from the engineered stem cells would be resistant to T cell-tropic HIV-1 infection. Although this strategy may not eradicate viruses in the patient's body, particularly those of the CD4-independent variants, it is expected to reconstitute a functional CD4 ϩ T cell subpopulation and to correct the most serious consequence, i.e. immunodeficiency. Needless to say, this approach requires the developments of a technology for massive propagation of hematopoietic stem cells in vitro and an efficient method for replacing a gene in somatic cells. The recent development of human embryonic stem (ES) cell lines (28) might provide another less difficult approach. Replacement of genes in ES cells is a routine technology now. In addition, the manipulated ES cells can differentiate in vitro into cells of hematopoietic lineage, including lymphoid precursors, but are devoid of mature immunocompetent cells. Such cells may well be taken by patients with severe immunodeficiency without causing graft versus host reactions. Efforts are being made along these lines.