Second-site changes affect viability of amphotropic/ecotropic chimeric enveloped murine leukemia viruses

Chimeras were previously generated between the ecotropic (Moloney-MuLV) and amphotropic (4070A) SU and TM proteins of murine leukemia virus (MuLV). After passage in D17 cells, three chimeras with junctions in the C terminus of SU (AE5, AE6, and AE7), showed improved kinetics of viral spreading, suggesting that they had adapted. Sequencing of the viruses derived from the D17 cell lines revealed second-site changes within the env gene. Changes were detected in the receptor binding domain, the proline-rich region, the C terminus of SU, and the ectodomain of TM. Second-site changes were subcloned into the parental DNA, singly and in combination, and tested for viability. All viruses had maintained their original cloned mutations and junctions. Reconstruction and passage of AE7 or AE6 virus with single point mutations recovered the additional second-site changes identified in the parental population. The AE5 isolate required changes in the VRA, the VRC, the VRB-hinge region, and the C terminus of SU for efficient infection. Passage of virus, including the parental 4070A, in D17 cells resulted in a predominant G100R mutation within the receptor binding domain. Viruses were subjected to titer determination in three cell types, NIH 3T3, canine D17, and 293T. AE6 viruses with changes in the proline-rich region initially adapted for growth on D17 cells could infect all cell types tested. AE6-based chimeras with additional mutations in the C terminus of SU could infect D17 and 293T cells. Infection of NIH 3T3 cells was dependent on the proline-rich mutation. AE7-based chimeras encoding L538Q and G100R were impaired in infecting NIH 3T3 and 293T cells.     

Effects of point mutations in the readthrough domain of the beet western yellows virus minor capsid protein on virus accumulation in planta and on transmission by aphids

Point mutations were introduced into or near five conserved sequence motifs of the readthrough domain of the beet western yellows virus minor capsid protein P74. The mutant virus was tested for its ability to accumulate efficiently in agroinfected plants and to be transmitted by its aphid vector, Myzus persicae. The stability of the mutants in the agroinfected and aphid-infected plants was followed by sequence analysis of the progeny virus. Only the mutation Y201D was found to strongly inhibit virus accumulation in planta following agroinfection, but high accumulation levels were restored by reversion or pseudoreversion at this site. Four of the five mutants were poorly aphid transmissible, but in three cases successful transmission was restored by pseudoreversion or second-site mutations. The same second-site mutations in the nonconserved motif PVT(32-34) were shown to compensate for two distinct primary mutations (R24A and E59A/D60A), one on each side of the PVT sequence. In the latter case, a second-site mutation in the PVT motif restored the ability of the virus to move from the hemocoel through the accessory salivary gland following microinjection of mutant virus into the aphid hemocoel but did not permit virus movement across the epithelium separating the intestine from the hemocoel. Successful movement of the mutant virus across both barriers was accompanied by conversion of A59 to E or T, indicating that distinct features of the readthrough domain in this region operate at different stages of the transmission process.     

Characterization of human immunodeficiency virus type 1 matrix revertants: effects on virus assembly, Gag processing, and Env incorporation into virions.

The matrix protein of human immunodeficiency virus type 1 (HIV-1) has been postulated to serve a variety of functions in the virus life cycle. Previously, we introduced a large number of mutations into the HIV-1 matrix and determined the effects on virus replication. These studies identified domains involved in virus assembly and release and envelope glycoprotein incorporation into virions. Here we describe the identification and characterization of viral revertants containing second-site changes in the matrix which compensate for the effects of four of the original mutations on matrix function. Specifically, mutations at matrix residues 4 and 6 severely impaired virus assembly and release; substitutions at residues 4 and 6 reversed the phenotype of the amino acid 4 change while second-site mutations at matrix positions 10, 69, and 97 partially or fully reversed the phenotype of the amino acid 6 substitution. A mutation at matrix residue 62 reversed the effect of a position 34 change which blocks envelope glycoprotein incorporation into virions, and substitutions at residues 27 and 51 reversed the phenotype of a position 86 mutation which redirects virus assembly to the cytoplasm. In addition to determining the effects of the compensatory changes in the context of the original mutations, we also introduced and analyzed the second-site changes alone in the context of the wild-type molecular clone. The data presented here define potential intermolecular and intramolecular interactions which occur in the matrix during the virus life cycle and have implications for our understanding of the relationship between matrix structure and function.     

Second-site suppression of a nonfunctional mutation within the Leishmania donovani inosine-guanosine transporter

LdNT2 is a member of the equilibrative nucleoside transporter family, which possesses several conserved residues located mainly within transmembrane domains. One of these residues, Asp(389) within LdNT2, was shown previously to be critical for transporter function without affecting ligand affinity or plasma membrane targeting. To further delineate the role of Asp(389) in LdNT2 function, second-site suppressors of the ldnt2-D389N null mutation were selected in yeast deficient in purine nucleoside transport and incapable of purine biosynthesis. A library of random mutants within the ldnt2-D389N background was screened in yeast for restoration of growth on inosine. Twelve different clones were obtained, each containing secondary mutations enabling inosine transport. One mutation, N175I, occurred in four clones and conferred augmented inosine transport capability compared with LdNT2 in yeast. N175I was subsequently introduced into an ldnt2-D389N construct tagged with green fluorescent protein and transfected into a Deltaldnt1/Deltaldnt2 Leishmania donovani knockout. GFP-N175I/D389N significantly suppressed the D389N phenotype and targeted properly to the plasma membrane and flagellum. Most interestingly, N175I increased the inosine K(m) by 10-fold within the D389N background relative to wild type GFP-LdNT2. Additional substitutions introduced at Asn(175) established that only large, nonpolar amino acids suppressed the D389N phenotype, indicating that suppression by Asn(175) has a specific size and charge requirement. Because multiple suppressor mutations alleviate the constraint imparted by the D389N mutation, these data suggest that Asp(389) is a conformationally sensitive residue. To impart spatial information to the clustering of second-site mutations, a three-dimensional model was constructed based upon members of the major facilitator superfamily using threading analysis. The model indicates that Asn(175) and Asp(389) lie in close proximity and that the second-site suppressor mutations cluster to one region of the transporter.     

Secondary mutations M36I and A71V in the human immunodeficiency virus type 1 protease can provide an advantage for the emergence of the primary mutation D30N

Development of resistance mutations in enzymatic targets of human immunodeficiency virus 1 (HIV-1) hampers the ability to provide adequate therapy. Of special interest is the effect mutations outside the active site of HIV-1 protease have on inhibitor binding and virus viability. We engineered protease mutants containing the active site mutation D30N alone and with the nonactive site polymorphisms M36I and/or A71V. We determined the K(i) values for the inhibitors nelfinavir, ritonavir, indinavir, KNI272, and AG1776 as well as the catalytic efficiency of the mutants. Single and double mutation combinations exhibited a decrease in catalytic efficiency, while the triple mutant displayed catalytic efficiency greater than that of the wild type. Variants containing M36I or A71V alone did not display a significant change in binding affinities to the inhibitors tested. The variant containing mutation D30N displayed a 2-6-fold increase in K(i) for all inhibitors tested, with nelfinavir showing the greatest increase. The double mutants containing a combination of mutations D30N, M36I, and A71V displayed -0.5-fold to +6-fold changes in the K(i) of all inhibitors tested, with ritonavir and nelfinavir most affected. Only the triple mutant showed a significant increase (>10-fold) in K(i) for inhibitor nelfinavir, ritonavir, or AG-1776 displaying 22-, 19-, or 15-fold increases, respectively. Our study shows that the M36I and A71V mutations provide a greater level of inhibitor cross-resistance combined with active site mutation D30N. M36I and A71V, when present as natural polymorphisms, could aid the virus in developing active site mutations to escape inhibitor binding while maintaining catalytic efficiency.     

Characterization of Two Second-Site Mutations Preventing Wild Type Protein Aggregation Caused by a Dominant Negative PMA1 Mutant

The correct biogenesis and localization of Pma1 at the plasma membrane is essential for yeast growth. A subset of PMA1 mutations behave as dominant negative because they produce aberrantly folded proteins that form protein aggregates, which in turn provoke the aggregation of the wild type protein. One approach to understand this dominant negative effect is to identify second-site mutations able to suppress the dominant lethal phenotype caused by those mutant alleles. We isolated and characterized two intragenic second-site suppressors of the PMA1-D378T dominant negative mutation. We present here the analysis of these new mutations that are located along the amino-terminal half of the protein and include a missense mutation, L151F, and an in-frame 12bp deletion that eliminates four residues from Cys409 to Ala412. The results show that the suppressor mutations disrupt the interaction between the mutant and wild type enzymes, and this enables the wild type Pma1 to reach the plasma membrane.     

Multiple correcting COL17A1 mutations in patients with revertant mosaicism of epidermolysis bullosa

Revertant mosaicism by somatic reversion of inherited mutations has been described for a number of genetic diseases. Several mechanisms can underlie this reversion process, such as gene conversion, crossing-over, true back mutation, and second-site mutation. Here, we report the occurrence of multiple corrections in two unrelated probands with revertant mosaicism of non-Herlitz junctional epidermolysis bullosa, an autosomal recessive genodermatosis due to mutations in the COL17A1 gene. Immunofluorescence microscopy and laser dissection microscopy, followed by DNA and RNA analysis, were performed on skin biopsy specimens. In patient 1, a true back mutation, 3781T-->C, was identified in the specimen from the arm, and a second-site mutation, 4463-1G-->A, which compensated for the frameshift caused by the inherited 4424-5insC mutation, was identified in the 3' splice site of exon 55 in a specimen from the middle finger. Patient 2 showed--besides two distinct gene conversion events in specimens from the arm and hand sites, both of which corrected the 1706delA mutation--a second-site mutation (3782G-->C) in an ankle specimen, which prevented the premature ending of the protein by the 3781C-->T nonsense mutation (R1226X). Thus, both inherited mutations, paternal as well as maternal, reverted at least once by different reversion events in distinct cell clusters in the described patients. The occurrence of multiple correcting mutations within the same patient indicates that in vivo reversion is less unusual than was generally thought. Furthermore, in the male patient, mosaic patterns of type XVII collagen-positive keratinocytes were present in clinically unaffected and affected skin. This latter observation makes it likely that reversion may be overlooked and may happen more often than expected.     

Effects of point mutations in the major capsid protein of beet western yellows virus on capsid formation,virus accumulation,and aphid transmission

Point mutations were introduced into the major capsid protein (P3) of cloned infectious cDNA of the polerovirus beet western yellows virus (BWYV) by manipulation of cloned infectious cDNA. Seven mutations targeted sites on the S domain predicted to lie on the capsid surface. An eighth mutation eliminated two arginine residues in the R domain, which is thought to extend into the capsid interior. The effects of the mutations on virus capsid formation, virus accumulation in protoplasts and plants, and aphid transmission were tested. All of the mutants replicated in protoplasts. The S-domain mutant W166R failed to protect viral RNA from RNase attack, suggesting that this particular mutation interfered with stable capsid formation. The R-domain mutant R7A/R8A protected approximately 90% of the viral RNA strand from RNase, suggesting that lower positive-charge density in the mutant capsid interior interfered with stable packaging of the complete strand into virions. Neither of these mutants systemically infected plants. The six remaining mutants properly packaged viral RNA and could invade Nicotiana clevelandii systemically following agroinfection. Mutant Q121E/N122D was poorly transmitted by aphids, implicating one or both targeted residues in virus-vector interactions. Successful transmission of mutant D172N was accompanied either by reversion to the wild type or by appearance of a second-site mutation, N137D. This finding indicates that D172 is also important for transmission but that the D172N transmission defect can be compensated for by a "reverse" substitution at another site. The results have been used to evaluate possible structural models for the BWYV capsid.     

Combinatorial reshaping of a lipase structure for thermostability: Additive role of surface stabilizing single point mutations

Thermostable lipases are of high priority for industrial applications. In the present study, targeted improvement of the thermostability of a lipase from metagenomic origin was examined by using a combinatorial protein engineering approach exploring additive effects of single amino acid substitutions. A variant (LipR5) was generated after combination of two thermostabilizing mutations (R214C & N355K). Thermostability of the variant enzyme was analyzed by half-life measurement and circular dichroism (CD). To assess whether catalytic properties were affected by mutation, the optimal reaction conditions were determined. The protein LipR5, displayed optimum activity at 50 °C and pH 8.0. It showed two fold enhancement in thermostability (at 60 °C) as compared to LipR3 (R214C) and nearly 168 fold enhancement as compared to parent enzyme (LipR1). Circular dichroism and fluorescence study suggest that the protein structure had become more rigid and stable to denaturation. Study of 3D model suggested that Lys355 was involved in formation of a Hydrogen bond with OE1 of Glu284. Lys355 was also making salt bridge with OE2 of Glu284.     

Charged residues in hepatitis C virus NS4B are critical for multiple NS4B functions in RNA replication

The nonstructural 4B (NS4B) protein of hepatitis C virus (HCV) plays a central role in the formation of the HCV replication complex. To gain insight into the role of charged residues for NS4B function in HCV RNA replication, alanine substitutions were engineered in place of 28 charged residues residing in the N- and C-terminal cytoplasmic domains of the NS4B protein of the HCV genotype 1b strain Con1. Eleven single charged-to-alanine mutants were not viable, while the remaining mutants were replication competent, albeit to differing degrees. By selecting revertants, second-site mutations were identified for one of the lethal NS4B mutations. Second-site mutations mapped to NS4B and partially suppressed the lethal replication phenotype. Further analyses showed that three NS4B mutations disrupted the formation of putative replication complexes, one mutation altered the stability of the NS4B protein, and cleavage at the NS4B/5A junction was significantly delayed by another mutation. Individual charged-to-alanine mutations did not affect interactions between the NS4B and NS3-4A proteins. A triple charged-to-alanine mutation produced a temperature-sensitive replication phenotype with no detectable RNA replication at 39°C, demonstrating that conditional mutations can be obtained by altering the charge characteristics of NS4B. Finally, NS4B mutations dispensable for efficient Con1 RNA replication were tested in the context of the chimeric genotype 2a virus, but significant defects in infectious-virus production were not detected. Taken together, these findings highlight the importance of charged residues for multiple NS4B functions in HCV RNA replication, including the formation of a functional replication complex.     

Second-site suppressor mutations assist in studying the function of the 3'' noncoding region of turnip yellow mosaic virus RNA.

The 3' noncoding region of turnip yellow mosaic virus RNA includes an 82-nucleotide-long tRNA-like structure domain and a short upstream region that includes a potential pseudoknot overlapping the coat protein termination codon. Genomic RNAs with point mutations in the 3' noncoding region that result in poor replication in protoplasts and no systemic symptoms in planta were inoculated onto Chinese cabbage plants in an effort to obtain second-site suppressor mutations. Putative second-site suppressor mutations were identified by RNase protection and sequencing and were then introduced into genomic cDNA clones to permit their characterization. A C-57----U mutation in the tRNA-like structure was a strong suppressor of the C-55----A mutation which prevented both systemic infection and in vitro valylation of the viral RNA. Both of these phenotypes were rescued in the double mutant. An A-107----C mutation was a strong second-site suppressor of the U-96----G mutation, permitting the double mutant to establish systemic infection. The C-107 and G-96 mutations are located on opposite strands of one helix of a potential pseudoknot, and the results support a functional role for the pseudoknot structure. A mutation near the 5' end of the genome (G + 92----A), at position -3 relative to the initiation codon of the essential open reading frame 206, was found to be a general potentiator of viral replication, probably as a result of enhanced expression of open reading frame 206. The A + 92 mutation enhanced the replication of mutant TYMC-G96 in protoplasts but was not a sufficiently potent suppressor to permit systemic spread of the A + 92/G-96 double mutant in plants.     

Lambda repressor mutants that are better substrates for RecA-mediated cleavage

RecA-mediated cleavage of the bacteriophage lambda repressor results in inactivation of the protein and leads to induction of the lambda prophage. Here, we report the identification of three mutations in lambda repressor that significantly increase the rate of RecA-mediated cleavage. These mutations were isolated as intragenic second-site suppressors of a mutation (ind-) which prevents cleavage. Purified repressor proteins that contain both the ind- mutation and one of the second-site mutations undergo cleavage at near wild-type rates. Purified repressors that contain the second-site mutations in otherwise wild-type backgrounds undergo RecA-mediated cleavage at significantly faster rates than wild-type, and form dimers more poorly than the wild-type protein. In related experiments, we found that other repressor mutants that dimerize poorly are also better substrates for RecA-mediated cleavage. Conversely, we show that a covalent disulfide-bonded repressor dimer is resistant to cleavage. These results support a model in which repressor monomers are the only substrate in the cleavage reaction.     

Stepwise improvements in catalytic effectiveness: independence and interdependence in combinations of point mutations of a sluggish triosephosphate isomerase

Second-site suppressor changes that improve the catalytic potency of a sluggish mutant of the enzyme triosephosphate isomerase have been examined both individually and in combination. Each of the second-site mutations increases the specific catalytic activity of a triosephosphate isomerase in which the catalytic base, glutamate-165, has been changed to aspartate. These second-site suppressors are G10S, S96P, S96T, E97D, V167D, and G233R. Not one of these changes enhances the value of kcat/Km for the wild-type enzyme, which is consistent with the knowledge that the reaction catalyzed by the wild-type enzyme is already diffusion-controlled. Indeed, two of the changes, S96P and V167D, are catalytically deleterious to the wild-type isomerase. When pairs of second-site suppressors are combined with the primary lesion E165D, six pairs show additive independence while the effects of eight other pairs are less than additive. The sites fall into two clusters: pairs within a cluster always interfere with one another and do not produce additive improvements in catalytic activity, whereas combinations of changes from different clusters tend to be additive in their effects. No combination of second-site suppressor mutations behaves synergistically, though there seems to be no a priori reason to exclude this possibility. Since the catalytic potency of each of the six second-site suppressor mutants can be further improved by the introduction of (at least) one of the other five changes, it is evident that none of the double mutants lies at a local catalytic maximum. In these cases, therefore, the opportunity exists for at least two "steps" of monotonic catalytic improvement along each of six different "paths" in protein space.     

Helical stalk segments S4 and S5 of the plasma membrane H+-ATPase from Saccharomyces cerevisiae are optimized to impact catalytic site environment

The stalk segments of P-type ion-translocating enzymes are presumed to play important roles in energy coupling. In this work, stalk segments S4 and S5 of the yeast H(+)-ATPase were examined for helical character, optimal length, and segment orientation by a combination of proline substitution, insertion/deletion mutagenesis, and second-site suppressor analyses. The substitution of various residues for helix-disrupting proline in both S4 (L353P,L353G; A354P; and G371P) and S5 (D676P and I684P) resulted in highly defective or inactive enzymes supporting the importance of helical character and/or the maintenance of essential interactions. The contiguous helical nature of transmembrane segment M5 and stalk element S5 was explored and found to be favorable, although not essential. The deletion or addition of one or more amino acids at positions Ala(354) in S4 and Asp(676) in S5, which were intended to either rotate helical faces or extend/reduce the length of helical segments, resulted in enzyme destabilization that abolished most enzyme assembly. Second-site suppressor mutations were obtained to primary site mutations G371A (S4) and D676G (S5) and were analyzed with a molecular structure model of the H(+)-ATPase. Primary site mutations were predicted to alter the site of phosphorylation either directly or indirectly. The suppressor mutations either directly changed packing around the primary site or altered the environment of the site of phosphorylation. Overall, these data support the view that stalk segments S4 and S5 of the H(+)-ATPase are helical elements that are optimized for length and interactions with other stalk elements and can influence the phosphorylation domain.     

Effects of cAMP-binding site mutations on intradomain cross-communication in the regulatory subunit of cAMP-dependent protein kinase I

Each protomer of the regulatory subunit dimer of cAMP-dependent protein kinase contains two tandem and homologous cAMP-binding domains, A and B, and cooperative cAMP binding to these two sites promotes holoenzyme dissociation. Several amino acid residues in the type I regulatory subunit, predicted to lie in close proximity to each bound cyclic nucleotide based on affinity labeling and model building, were replaced using recombinant techniques. The mutations included replacement of 1) Glu-200, predicted to hydrogen bond to the 2'-OH of cAMP bound to site A, with Asp, 2) Tyr-371, the site of affinity labeling with 8-N3-cAMP in site B, with Trp, and 3) Phe-247, the position in site A that is homologous to Tyr-371 in site B, with Tyr. Each mutation caused an approximate 2-fold increase in both the Ka(cAMP) and Kd(cAMP); however, the off-rates for cAMP and the characteristic pattern of affinity labeling with 8-N3-cAMP differed markedly for each mutant protein. Furthermore, these mutations affect the cAMP binding properties not only of the site containing the mutation, but of the adjacent nonmutated site as well, thus confirming that extensive cross-communication occurs between the two cAMP-binding domains. Photoaffinity labeling of the native R-subunit results in the covalent modification of two residues, Trp-260 and Tyr-371, by 8-N3-cAMP bound to sites A and B, respectively, with a stoichiometry of 1 mol of 8-N3-cAMP incorporated per mol of R-monomer (Bubis, J., and Taylor, S. S. (1987) Biochemistry 26, 3478-3486). A stoichiometry of 1 mol of 8-N3-cAMP incorporated per R-monomer was observed for each mutant regulatory subunit as well, even when 2 mol of 8-N3-cAMP were bound per R-monomer; however, the major sites of covalent modification were altered as follows: R(Y371/W), Trp-371; R(E200/D), Tyr-371, and R(F247/Y), Tyr-371.     

Spontaneous mutations restore the viability of tick-borne encephalitis virus mutants with large deletions in protein C

The capsid protein, C, of tick-borne encephalitis virus has recently been found to tolerate deletions up to a length of 16 amino acid residues that partially removed the central hydrophobic domain, a sequence element conserved among flaviviruses which may be crucial for virion assembly. In this study, mutants with deletion lengths of 19, 21, 27, or 30 residues, removing more or all of this hydrophobic domain, were found to yield viable virus progeny, but this was without exception accompanied by the emergence of additional mutations within protein C. These point mutations or sequence duplications were located downstream of the engineered deletion and generally increased the hydrophobicity, suggesting that they may compensate for the loss of the central hydrophobic domain. Two of the second-site mutations, together with the corresponding deletion, were introduced into a wild-type genetic backbone, and the analysis of these "double mutants" provided direct evidence that the viability of the deletion mutant indeed depended on the presence of the second-site mutation. Our results corroborate the notion that hydrophobic interactions of protein C are essential for the assembly of infectious flavivirus particles but rule out the possibility that individual residues of the central hydrophobic domain are absolutely required for infectivity. Furthermore, the double mutants were found to be highly attenuated and capable of inducing a protective immune response in mice at even lower inoculation doses than the previously characterized 16-amino-acid-residue deletion mutant, suggesting that the combination of large deletions and second-site mutations may be a superior way to generate safe, attenuated flavivirus vaccine strains.     

A second-site suppressor strategy for chemical genetic analysis of diverse protein kinases

Chemical genetic analysis of protein kinases involves engineering kinases to be uniquely sensitive to inhibitors and ATP analogs that are not recognized by wild-type kinases. Despite the successful application of this approach to over two dozen kinases, several kinases do not tolerate the necessary modification to the ATP binding pocket, as they lose catalytic activity or cellular function upon mutation of the 'gatekeeper' residue that governs inhibitor and nucleotide substrate specificity. Here we describe the identification of second-site suppressor mutations to rescue the activity of 'intolerant' kinases. A bacterial genetic selection for second-site suppressors using an aminoglycoside kinase APH(3')-IIIa revealed several suppressor hotspots in the kinase domain. Informed by results from this selection, we focused on the beta sheet in the N-terminal subdomain and generated a structure-based sequence alignment of protein kinases in this region. From this alignment, we identified second-site suppressors for several divergent kinases including Cdc5, MEKK1, GRK2 and Pto. The ability to identify second-site suppressors to rescue the activity of intolerant kinases should facilitate chemical genetic analysis of the majority of protein kinases in the genome.     

Host suppressors in Arabidopsis thaliana of mutations in the movement protein gene of Cauliflower mosaic virus

A novel genetic screen was used to identify host factors in Arabidopsis thaliana that suppress mutations in the Cauliflower mosaic virus (CaMV) movement protein gene (gene I). A series of small mutations was made in gene I and the mutations were tested for their suitability in a suppressor screen. The first round of screening yielded only revertants or second-site mutations in gene I. A derivative of one of the second-site mutant viruses (N7) that was delayed in symptom production was used in a second round of screening for suppressor plants that accelerated symptom production. Two candidate suppressor plants were found that accelerated by 1 to 4 days the first appearance of symptoms caused by the mutant viruses. One of the suppressors (5-2), called asc1 (acceleration of symptoms by CaMV N7), was mapped to chromosome 1. Two additional loci that differentially affect N7 virus susceptibility in the parental Columbia and Ler ecotypes were mapped to chromosomes 3 and 4 by quantitative trait locus (QTL) analysis.