The Genetic Basis for Salivary Gland Barriers to Arboviral Transmission

Simple Summary Mosquito salivary glands are infected with an arbovirus prior to entering saliva. The virus in the mosquito’s saliva can then infect a vertebrate host when the vector acquires its next bloodmeal. Salivary gland infection and escape barriers (SGIB and SGEB, respectively) have been identified that modulate arbovirus transmission. SGIB are manifest as the absence of virus in the salivary glands of mosquitoes having a disseminated infection, while SGEB are evident as an absence of virus in the saliva of mosquitoes even though salivary glands are infected. The interaction between Aedes aegypti and viruses is dynamic and related to the genetic background of each vector population and virus variant. There is little understanding of the genetic basis for SGIB and SGEB. Here, nested, half-sib families of Aedes aegypti were used to estimate genetic and environmental variances, in which daughters from an individual dam differ in their phenotype due to the sire. We found that SGIB has a strong genetic basis with dengue virus infections but not with Zika or chikungunya virus infections, in which all salivary glands became infected. SGEB has a moderate genetic basis for Zika or chikungunya infections but not with dengue virus infections. Abstract Arthropod-borne viruses (arboviruses) infect mosquito salivary glands and then escape to saliva prior to virus transmission. Arbovirus transmission from mosquitoes can be modulated by salivary gland infection barriers (SGIBs) and salivary gland escape barriers (SGEBs). We determined the influence of SGIBs and SGEBs by estimating the quantitative genetic contributions of Aedes aegypti half-sib families (Mapastepec, Mexico) infected with three dengue 2 (DENV2), two chikungunya (CHIKV), and two Zika (ZIKV) genotypes. We determined virus titer per salivary gland and saliva at seven days post-infection and virus prevalence in the half-sib population. CHIKV or ZIKV genotypes did not present SGIB, whereas DENV2 genotypes showed low rates of SGIB. However, virus titer and prevalence due to additive genetic factors in the half-sib family displayed a significant narrow-sense heritability (h2) for SGIB in two of the three DENV2 genotypes and one CHIKV and one ZIKV genotype. SGEBs were detected in all seven virus strains: 60–88% of DENV2 and 48–62% of CHIKV or ZIKV genotype infections. SGEB h2 was significant for all CHIKV or ZIKV genotypes but not for any of the DENV2 genotypes. SGIBs and SGEBs exhibited classical gene-by-gene interaction dynamics and are influenced by genetic factors in the mosquito and the virus.


Introduction
Vector competence (VC) for an arbovirus is affected by intrinsic (genetic) factors and mechanisms that control the ability of a vector mosquito to acquire, maintain, and transmit an arbovirus. There are several tissue barriers in the mosquito that the virus must overcome in order to establish a productive infection: the midgut infection barrier (MIB), midgut escape barrier (MEB), salivary gland infection barrier (SGIB), and salivary gland escape barrier (SGEB). In mosquito vectors of disease, the salivary gland is the organ leading to transmission of an arbovirus from an infected mosquito [1]. In all mosquitoes, the two salivary glands are located in the anterior portion of the thoracic hemocele. In females, each gland consists of two lateral and one medial cylindrical lobe ( Figure 1A). The distal end of each lobe is called the acinus, while the proximal end is called the neck ( Figure 1A). Each lobe consists of a basal lamina bounding a single layer of acinar cells ( Figure 1B). These cells are distributed around a central salivary duct containing an apical cavity for saliva storage. Apical plasma membranes have microvillar projections that are highly convoluted to increase the surface area for discharge of saliva into the apical cavity (diverticulum) ( Figure 1B). Following successful dissemination from the midgut, arboviruses typically infect hemocytes, fat body, neural tissue, and occasionally muscle tissue before infecting the salivary glands. Virus amplification in fat body and specifically in hemocytes is an important prerequisite before the virus is ready to infect the salivary glands [2]. Virus in the hemolymph penetrates the basal lamina of the salivary glands surrounding acinar cells to generate a salivary gland infection ( Figure 1C) [3]. Factors that prevent virus penetration of the salivary gland basal lamina constitute SGIB. Several studies suggest that tracheoles or nerve tissues may allow virus entry into acinar cells [4,5], where viral replication occurs. Barriers that prevent infection and virus replication in the salivary glands also constitute SGIB (SGI, Figure 1C). Viruses are eventually released in the apical cavities where saliva is stored prior to its release into the hypopharynx and deposition of virus into a vertebrate during probing and feeding. Barriers that block release of mature virus in the apical cavities are referred to as SGE1 and SGE2 ( Figure 1C). Virus replication in the salivary glands has been reported for DENV [6], CHIKV [3], and ZIKV [7].
The existence of SGEB has been definitively demonstrated for Japanese encephalitis virus (JEV) in Culex tritaeniorhynchus [8], Snowshoe hare virus in Aedes triseriatus [9], La Crosse virus (LACV) in Aedes hendersoni [10], and Sindbis virus (SINV) in Culex theileri [11]. An exceptionally strong SGEB was detected in Ae. hendersoni for LACV, wherein 65% of females had salivary gland infections by plaque assay, but only 5% actually transmitted infectious virus [12]. These results were confirmed by immunofluorescence assays (IFA), which showed high concentrations of LACV antigens in the salivary glands. More recently, SGEB have been reported for Rift Valley fever virus (RVFV) in Culex nigripalpus, Anopheles crucians, and Aedes infirmatus [13,14]. A recent transmission electron microscopy study [3] nicely illustrated CHIKV replication and storage in Ae. albopictus salivary glands and showed that CHIKV nucleocapsids bud from two different sites in acinar cells to become mature particles. The first site is at vesicular membranes located in the apical cavity of acinar cells (SGE1, Figure 1C), while the second occurs at cellular plasma membranes (SGE2, Figure 1C). Mature viral particles clearly accumulated in acinar cell junctions ( Figure 1D) [3]. Furthermore, it has been shown that different viruses preferentially replicate in different lobes of the salivary glands [15]. Arbovirus infection of salivary glands typically begins in the distal lateral lobes [15][16][17][18][19]. Arboviruses such as DENV2 and CHIKV infect the proximal lateral and medial [20] lobes of Ae. aegypti, whereas SINV does not seem to infect the median lobes of Aedes albopictus or Ae. aegypti [15,16,20,21]. Electron micrographic studies confirm that clumps of crystalline flavi-and alphaviruses [3,22] occur in the salivary glands and saliva. It has been reported that the lateral and medial salivary gland lobes interact with monoclonal antibodies differently, suggesting there is a biochemical difference between these two lobe surfaces [23]. Others have observed that specific proteins are differentially expressed by different salivary gland lobes [24], which could account for the observed differences between lobes or between mosquitoes in susceptibility to virus infection [2,25]. cation occurs. Barriers that prevent infection and virus replication in the salivary glands also constitute SGIB (SGI, Figure 1C). Viruses are eventually released in the apical cavities where saliva is stored prior to its release into the hypopharynx and deposition of virus into a vertebrate during probing and feeding. Barriers that block release of mature virus in the apical cavities are referred to as SGE1 and SGE2 ( Figure 1C). Virus replication in the salivary glands has been reported for DENV [6], CHIKV [3], and ZIKV [7].  Salivary gland escape (SGE)1 and SGE2 refer to potential barriers (entry into the apical cavity and acinar cell junctions, respectively) for virus escape from salivary glands (D) Immunofluorescence of salivary gland lobe infected with DENV-2 Merida (Cosmopolitan) at 7 dpi shows that mature viral particles clearly accumulated in acinar cell junctions (red arrow). (E) Uninfected salivary glands showed no immunofluorescence signal. N: nuclei. SD: salivary duct. AC: apical cavity for saliva storage. BL: basal lamina. Cyt: Cytoplasm. CM: Cell membrane. DLL: distal lateral lobe, ML: medial lobe. * Represents a virus cluster (1C).
To reach the salivary glands, arboviruses must overcome natural barriers, infect various tissues, exploit physical conduits, and counteract innate and cellular defenses [15,[26][27][28]. Meanwhile, vectorial capacity involves extrinsic factors, including temperature, availability of vertebrate hosts, feeding behavior, population density, longevity, and predation [29]. The interaction between Ae. aegypti and a virus is dynamic and is related to the genetic background of each particular vector population and arbovirus variant [30][31][32][33]. It is possible that the existence of mosquito genetic variants is related to these functions, which could affect the degree of susceptibility in mosquitoes [34]. There is little understanding of whether there is a genetic basis for SGIB and SGEB or whether they are random and subject to a few or many of the "environmental" factors listed in the previous paragraph. One of the principal applications of quantitative genetics is to determine the relative contributions of genetic and environmental factors to variation and covariation in one or more quantitative phenotypes [35]. Basically, the phenotypic variance (Vp) of a population is viewed as the sum of separate genetic (Vg) and environmental (Ve) variances. The genetic variance is in turn made up of additive (Va) and non-additive (dominance/epistatic) components. Va and the associated narrow-sense heritability (h 2 = Va/Vp) can vary from 0 to 1, and indicate the proportion of phenotypic variance explained by additive genetic effects [35].
Herein, we assume that if the variances in SGIB and SGEB among siblings or between half-sib families are due to one or all of the random causes listed above, including variation in the amount of saliva produced by a female, then Ve will be large, Va will be small, and h 2 will approach zero. Alternatively, if the amount of virus collected in saliva is due primarily to genetic, rather than environmental, factors, then the variation in SGIB and SGEB will have a large h 2 , approaching one. Herein, we perform a half-sib analysis to estimate the quantitative genetic and environmental basis of SGIB and SGEB for three DENV2 genotypes (American, Cosmopolitan, and Asian genotype), two CHIKV genotypes (Asian and ECSA-IOL genotype) and two ZIKV genotypes (Asian and African genotype). Basically, we are testing if, given all the environmental effects listed above, a significant genetic effect can be detected. Hereafter, viruses are designated in the text as DENV2 American, DENV2 Cosmopolitan, DENV2 Asian, CHIKV Asian, CHIKV ECSA-IOL, ZIKV Asian, and ZIKV African, respectively. DENV2 isolates were endemic strains from Mexico. The virus strains used in this study have different levels of vector competence: low for DENV2 American, moderate for DENV2 Asian, CHIKV Asian, and ZIKV Asian, and high for DENV2 Cosmopolitan and ZIKV African. The American, Cosmopolitan, and Asian DENV2 isolates used here were available in our lab and afforded us the ability to determine heritability in the context of three DENV2 genotypes from Mexico. DENV2 were propagated in C6/36 cells at a multiplicity of infection (MOI) of 0.01. ZIKV and CHIKV were propagated similarly in Vero cells. Supernatant was harvested (3-5 days post-infection (dpi) for CHIKV and ZIKV and 10-12 dpi for DENV), clarified by centrifugation at 4 • C, and aliquoted into single-use vials before freezing at −80 • C. Aedes aegypti strain Mapastepec (15 • 26 44.05 N, 92 • 54 9.74 W), a domestic mosquito from an urban region in from the state of Chiapas, Mexico, was originally collected in 2016. Adult mosquitoes were fed on raisins and water, and the females were allowed to feed on uninfected blood meals to stimulate oogenesis. Mosquitoes were maintained at 28 • C, 80% relative humidity with 16-8 h light-dark periods. Half-sib families for this study were created using generations F4, F5, and F6.

Intrathoracic Inoculation
Intrathoracic injections of DENV, ZIKV, and CHIKV were conducted to assess virus replication in salivary glands and other mosquito tissues outside the midgut. Intrathoracic inoculation of mosquitoes with arboviruses usually leads to more rapid, and higher proportions of, disseminated infections, compared to oral challenge, and mosquitoes become infected with a constant dose of virus [14,15,36,37]. We acknowledge that intrathoracic injections bypass the midgut, arguably the most important barrier to mosquito infection. However, past experience [15] has shown that only a few orally-infected mosquitoes in a family actually go on to develop infected salivary glands. These small numbers preclude an accurate quantitative genetic analysis of SGIB and SGEB. Notably, the peak concentration of virus detected in the salivary glands is often similar, regardless of which method is initially used [15]. The method for intrathoracic inoculation of mosquitoes is described in [38] with some modifications. Adult females (7 days post-eclosion) were infected by injecting 500 plaque forming units (PFU) in a volume of 69 nL of the virus stock using a Nanojet II (Drummond Scientific Company, Broomall, PA, USA). All injections were performed under a dissecting microscope using glass needles that were prepared using a vertical pipette puller (P-30, Sutter Instrument Co., Novato, CA, USA). Mosquitoes were left for 7 days at 28 • C and 80% relative humidity before samples were collected.

Saliva Collection and Salivary Gland Dissections
At 7 dpi, saliva from each of 10-15 daughters/dam were collected [39]. Briefly, females were chilled, and their wings and legs removed. Their proboscises were inserted into a 1 µL micropipette (microcaps ® , Drummond Scientific Company, PA, USA) filled with immersion oil type B, and the mosquitoes were allowed to salivate into the oil at room temperature. After 5 min (probing) and 45-60 min, each individual capillary was observed under the stereoscope Olympus SZH10 at 2X magnification for the presence of saliva. Mosquitoes that did not visibly salivate were eliminated. To determine the relationship between saliva volume and saliva titer, categories were assigned 1 (≈4.7 nL), 2 (≈7 nL), 3 (≈10.8 nL), or 4 (≈13.5 nL), based on the total observed volume compared with a reference capillary ( Figure 2) [40]. injections bypass the midgut, arguably the most important barrier to mosquito infection. However, past experience [15] has shown that only a few orally-infected mosquitoes in a family actually go on to develop infected salivary glands. These small numbers preclude an accurate quantitative genetic analysis of SGIB and SGEB. Notably, the peak concentration of virus detected in the salivary glands is often similar, regardless of which method is initially used [15]. The method for intrathoracic inoculation of mosquitoes is described in [38] with some modifications. Adult females (7 days post-eclosion) were infected by injecting 500 plaque forming units (PFU) in a volume of 69 nL of the virus stock using a Nanojet II (Drummond Scientific Company, Broomall, PA, USA). All injections were performed under a dissecting microscope using glass needles that were prepared using a vertical pipette puller (P-30, Sutter Instrument Co., Novato, CA, USA). Mosquitoes were left for 7 days at 28 °C and 80% relative humidity before samples were collected.

Saliva Collection and Salivary Gland Dissections
At 7 dpi, saliva from each of 10-15 daughters/dam were collected [39]. Briefly, females were chilled, and their wings and legs removed. Their proboscises were inserted into a 1 µL micropipette (microcaps ® , Drummond Scientific Company, PA, USA) filled with immersion oil type B, and the mosquitoes were allowed to salivate into the oil at room temperature. After 5 min (probing) and 45-60 min, each individual capillary was observed under the stereoscope Olympus SZH10 at 2X magnification for the presence of saliva. Mosquitoes that did not visibly salivate were eliminated. To determine the relationship between saliva volume and saliva titer, categories were assigned 1 (≈4.7 nL), 2 (≈7 nL), 3 (≈10.8 nL), or 4 (≈13.5 nL), based on the total observed volume compared with a reference capillary ( Figure 2) [40]. Individual capillaries observed under the microscope for the presence of saliva and assigned saliva categories based on total volume. The volume of saliva was calculated using the cylinder volume formula (V = π (r2 × h) = 3.1416 × 0.01 × h = mm 3 or µL) and measuring the height of the saliva with a digital fractional caliper (precision: ±0.02 mm), as described previously.
The proboscis was removed from the capillary, and oil containing the saliva was expelled under pressure into 1.5 mL tubes containing 300 µL Dulbecco's Modified Eagle Medium (DMEM supplemented with 20% of heat inactivated FBS, 1% penicillin/ streptomycin, 1% glutamine, 1% non-essential amino acids) and frozen immediately on dry ice. Individual capillaries observed under the microscope for the presence of saliva and assigned saliva categories based on total volume. The volume of saliva was calculated using the cylinder volume formula (V = π (r 2 × h) = 3.1416 × 0.01 × h = mm 3 or µL) and measuring the height of the saliva with a digital fractional caliper (precision: ±0.02 mm), as described previously.
The proboscis was removed from the capillary, and oil containing the saliva was expelled under pressure into 1.5 mL tubes containing 300 µL Dulbecco's Modified Eagle Medium (DMEM supplemented with 20% of heat inactivated FBS, 1% penicillin/streptomycin, 1% glutamine, 1% non-essential amino acids) and frozen immediately on dry ice. After collecting saliva, salivary glands from mosquitoes were dissected and collected in 500 µL DMEM medium (7% heat inactivated FBS, 1% penicillin/streptomycin, 1% glutamine, 1% non-essential amino acids) and frozen immediately on dry ice. Virus in saliva and salivary glands were detected and quantified by plaque assay.

Quantitative Genetic Analyses
A nested half-sib design was used to provide estimates of genetic and environmental variances and covariances [35,43,44]. A single male was mated to several (usually 3-5) females. The females were allowed to bloodfeed and lay eggs. Eggs from each female were collected, and 5 days later, were reared separately, awaiting adult eclosion ( Figure 3). All 7-day-old daughters from an individual dam were intrathoracically inoculated with 500 PFU of virus, and after 7 days, were tested for the presence or absence of virus and the amount of virus in the salivary glands and saliva ( Figure 3). Half-sib families differ in their phenotype due to the sire. If the SGIB or SGEB have a genetic basis, then sharing the same sire is why half-sibs are similar for the SGIB or SGEB phenotype. Variance in phenotype between half-sib families is equal to the covariance in phenotype among half-sibs. A high variance between families should cause high covariance within families, because members of a family are alike and the heritability approaches one. Conversely, a low variance in phenotype between families will lead to low covariance within families, and the heritability approaches zero.
Restricted maximum likelihood (REML) was used to estimate quantitative genetic parameters [44,45]. REML does not require balanced designs, and missing cells have a minimal effect. Furthermore, REML allows a diversity of hypotheses to be tested. The computational limits that formerly made REML impractical for large datasets have mostly been overcome. For these reasons, REML is now the method of choice for estimating quantitative genetic parameters from most designs. We used PROC MIXED in SAS (SAS User's Guide, Cary, NC: Statistical Analysis System Institute, Inc., 1987) to obtain REML estimates and hypothesis tests for half-sib designs [44,45]. The SAS code for performing REML analyses appears in Figure S1. The total phenotypic variance (σ 2 P ) is a function of three observed sources of variance: the average variance among offspring of a single sire is σ 2 S , the average variance among full-sib families from dams mated to a common sire is σ 2 d , and the average variance within members of a full-sib family is σ 2 w . Additive genetic variance (Va) is calculated as: The variance among dams (Vd) bred to a common sire is calculated as: Insects 2021, 12, 73 The variance among full-sibs is (Vw): Summing Equations (1)- (3), Narrow-sense heritability (h 2 ) is: See SAS lines 5-10 for SGIB, and lines 24-30 for SGEB in Figure S1. The null hypothesis that σ 2 S = 0 was analyzed with a likelihood ratio test. The 2 restricted log likelihood for the full model (Equation (4)) was subtracted from 2 restricted log likelihood for the Va = 0 mode See SAS lines 12-23 for SGIB and lines 31-42 for SGEB in Supplement 1. This is a likelihood ratio test because the difference in log likelihoods equals the log of the ratio of the likelihoods. There is one degree of freedom (d. f.) because Equations (4) and (6)   Restricted maximum likelihood (REML) was used to estimate quantitative genetic parameters [44,45]. REML does not require balanced designs, and missing cells have a minimal effect. Furthermore, REML allows a diversity of hypotheses to be tested. The computational limits that formerly made REML impractical for large datasets have mostly been overcome. For these reasons, REML is now the method of choice for estimating quantitative genetic parameters from most designs. We used PROC MIXED in SAS (SAS User's Guide, Cary, NC: Statistical Analysis System Institute, Inc., 1987) to obtain REML estimates and hypothesis tests for half-sib designs [44,45]. The SAS code for performing REML analyses appears in Figure S1. The total phenotypic variance (σ 2 P) is a function of three observed sources of variance: the average variance among offspring of a single sire is σ 2 S, the average variance among full-sib families from dams mated to a common sire is σ 2 d, and the average variance within members of a full-sib family is σ 2 w. Additive genetic variance (Va) is calculated as:

Statistical Analyses
All statistical analyses were performed with GraphPad Prism (version 5.0, La Jolla, CA, USA). A two-tailed Fisher's exact test was used to compare infection rates for SGIB and SGEB and determine any correlation among virus titers in saliva and salivary glands (Pearson correlation). Analysis of variance (one-way and two-way ANOVA) was used to determine the statistical significance, and CORREL function (Excel 2016, Microsoft, Redmond, DC, USA) was used for determining the correlation coefficient. Significance was defined as p < 0.05.

Infection Rates
Our breeding design was established with a minimum number of sires (15)(16)(17)(18)(19)(20), a minimum of 2-3 half-sib families (dam/sire) and a minimum of ten F1 females/half-sib families. Under this criterion, we obtained 24 half-sib families for DENV2 American genotype. We also established 21 half-sib families for DENV2 Asian and Cosmopolitan genotypes, ZIKV Asian and African genotypes, and 22 half-sib families for CHIKV genotypes. Table 1 lists the number of paired samples tested for each of the seven viral strains. The number of salivary glands that were not infected divided by the total number of salivary glands tested is the SGIB%. The number of infected salivary glands that failed to produce any virus in the saliva are also listed. This number divided by the number of infected salivary glands is the SGEB%. The percentage transmission is 100 − SGEB% (Table 1). These calculations assume that viral dissemination occurred in each intrathoracicallyinoculated mosquito based on data that shows 100% infection of whole mosquitoes (n = 20-30) by plaque assay for each viral strain at 7 dpi ( Figure 4).
When comparing the spread of sire means between DENV2 genotypes, the greatest variation in log 10 (PFU/mL) means/sire occurred with DENV2 American (Figure 6A) infection, and the smallest spread of sire means occurred with DENV2 Cosmopolitan infection ( Figure 6C). Variation among sire families with CHIKV ECSA-IOL ( Figure 6D) was slightly greater than that among CHIKV Asian families ( Figure 6E). Variation among sire families with ZIKV Asian and ZIKV African were also small ( Figure 6F,G). Narrow-sense heritabilities (h 2 ) for SGIB were calculated to characterize each viral infection (Table 2) once the observational components σ 2 S , σ 2 d , and σ 2 w and the causal components of variance (Vp, Vsire, Vdam, Ve) for log 10 (PFU/mL)/salivary gland were obtained (Table S1).
There were large differences between the CHIKV and ZIKV genotypes. SGIB exhibited strong genetic bases with CHIKV ECSA-IOL and ZIKV African, but not CHIKV or ZIKV Asian genotypes. For CHIKV ECSA-IOL and ZIKV African, the variance among sire means and the h 2 were significantly different (for CHIKV ECSA-IOL: σ 2 S = 0.0337, p = 0.005; h 2 = 0.5174, p = 6.2 × 10 −6 ; for ZIKV African: σ 2 S = 0.0762, p = 0.0133; h 2 = 0.3106, p = 0.0009). For CHIKV Asian genotype, the variance among sire means and the h 2 were not significantly different (σ 2 S = 0.0012, p = 0.4537; h 2 = 0.0385, p = 0.5). For ZIKV Asian genotype, the variance among sire means was 0.008 and not significant (p = 0.0918), and h 2 was 0.1526 and marginally significant (p = 0.0416). There was a positive correlation between the variance among sire means and the narrow-sense heritability (h 2 ). As predicted by Equation (5) (h 2 = Va/Vp), the greater the variance in sire means, the larger the heritability. In other words, when the amount of virus was low (e.g., Figure 6A,B, DENV2 American and Asian), estimates of heritability and variance among sire means tended to be large and significant. However, once viral titers reach a high enough level in the salivary gland, (DENV2 Cosmopolitan, CHIKV, and ZIKV genotypes in Figure 6C-G), genes in the mosquito may no longer have an effect, and heritability estimates were small.
Variation among sire families ( Figure 6) suggested that viral titer in the salivary gland is negatively correlated with the magnitude of heritability. We performed a correlation analysis between the average log 10 PFU/mL virus in the salivary glands and the heritability to evaluate this observation (Figure 7). Pearson's correlation coefficient was large and negative, as expected, when ZIKV African was excluded (R = −0.819, p = 0.0461). If ZIKV African was included, Pearson correlation coefficient was −0.24150 (p = 0.6019). titers. SGIB did affect the quantity of virus present in the salivary gland. It is very clear that different half-sib families can, and do, vary greatly in the amount of virus in salivary glands ( Figure 6).  analysis between the average log10 PFU/mL virus in the salivary glands and the heritability to evaluate this observation (Figure 7). Pearson's correlation coefficient was large and negative, as expected, when ZIKV African was excluded (R = −0.819, p = 0.0461). If ZIKV African was included, Pearson correlation coefficient was -0.24150 (p = 0.6019).
Narrow-sense heritabilities (h 2 ) for SGEB were also calculated to characterize each viral infection as SGIB. The observational components σ 2 S , σ 2 d , and σ 2 w and the causal components of variance (Vp, Vsire, Vdam, Ve) for log 10 (PFU/mL)/saliva were obtained (Table S2). In contrast to SGIB, the heritabilities for SGEB (log 10 (PFU/mL)/saliva) were not significant for any of the three DENV2 genotypes tested (American, h 2 = 0.319, p = 0.127; Asian, h 2 = 0.160, p = 0.185; Cosmopolitan, h 2 = 0.2888, p = 0.1103). Except for ZIKV African, the heritabilities for SGEB were likewise significant (for CHIKVECSA-IOL, h 2 = 0.461, p = 0.00041; for CHIKV Asian, h 2 = 0.437, p = 7.7932 × 10 −5 ; for ZIKV Asian, h 2 = 0.3286, p = 0.000226; and for ZIKV African, h 2 = 0.4050, p = 0.0415). Clearly, genetic factors in the mosquito do influence the amount of virus in the saliva for CHIKV and ZIKV Asian (Table 3). Table 3. Quantitative genetic analysis of SGEB in Aedes aegypti. Each analysis corresponds to one of the seven virus strains tested for log 10 (PFU)/saliva. We used PROC MIXED in SAS to obtain restricted maximum likelihood (REML) to estimate quantitative genetic parameters. The greater the variance in sire means, the larger the heritability. In examining Figure 8, note the positive trend (R = 0.86) between the variance among sire means and the h 2 . The average log 10 (PFU/mL) virus in the saliva was not correlated with heritability (R = 0.552, p = 0.2557). Once the virus has infected the salivary gland, SGEB appears to have a major effect on the presence or absence of virus in the saliva (Table 1). SGEB has a moderate genetic basis for Zika or chikungunya infections, but not with dengue virus infections.

Relationship between SGIB and SGEB
We tested for a statistical correlation between log 10 (PFU/mL)/salivary gland and log 10 (PFU/mL)/saliva to determine whether the amount of virus in the salivary glands affected the amount of virus in the saliva. We calculated Pearson's correlation coefficients between the log 10 (PFU/mL)/salivary gland and the average log 10 (PFU/mL)/saliva across all individual mosquitoes (Table 4) for all viruses tested. These correlations were only calculated among mosquitoes with virus in the salivary gland. Three of the seven genotypes have a significant correlation, but the magnitude of the correlations was small (DENV2 Asian, p = 0.0487; CHIKV ECSA-IOL1, p = 0.0026; ZIKV African, p < 0.0001).   An additional correlation analysis was completed between the average log 10 (PFU/mL)/ salivary gland and the average log 10 (PFU/mL)/saliva across all sire families for a genotype. When combined across all genotypes, a correlation of R = 0.6533 (p < 0.0001) was obtained.
However, significant correlation was only obtained for DENV2 American (p = 0.0095), DENV2 Asian (p = 0.0124) and ZIKV African (p = 0.0016). This suggests that the overall correlation is an artifact of combining results from different viral genotypes. In general, the titer of virus in the salivary glands was not strongly correlated with the titer of virus in the saliva.

Relationship between Saliva Volume, Virus Titer, and Prevalence
To evaluate whether we were examining the genetics of SGEB underlying the amount of virus in the saliva or the SGEB genetics underlying the amount of saliva produced by mosquito, an additional 100 mosquitoes were intrathoracially injected with 500 PFU/mL for each virus, and at the time of saliva collection. All samples were assigned categories (1+, 2+, 3+, etc.) based on observed volume (Figure 2), using the stereoscope Olympus SZH10 at 2X magnification. Mosquitoes that did not visibly salivate or samples that were too difficult to measure were eliminated. Techniques have been designed to quantify saliva [47,48], but they prevent subsequent work with collected virus, and these techniques require excessive standardization to correlate the amount of label protein with the volume of saliva. We are aware this method did not (or could not) accurately determine the amount of saliva collected from individual mosquitoes, but it gave a basis for correlating the volume of saliva to the virus titer.
If all mosquitoes produce the same concentration of virus, then we would expect the amount of virus to be proportional to the amount of saliva. The amount of live virus would also be a function of the volume of saliva released into the host, which is also a continuous variable. Alternatively, if the saliva contains clusters of viruses, then there should be no relationship between the amount of saliva and the amount of virus. It has been reported that saliva often contains clusters of viruses rather than a uniform distribution [3,22]. There is no significant difference between the amount of saliva and the virus titer (log 10 (PFU/mL)) for all virus strains tested in this study (DENV2 genotypes: p = 0.8530 for American, p = 0.7188 for Asian, and p = 0.7532 for Cosmopolitan; CHIKV genotypes: An additional correlation analysis was completed between the log 10 (PFU/mL)/saliva and saliva volume for all viruses (Table 5). These correlations were only calculated in saliva samples positive for virus. The only significant correlation that was obtained was for ZIKV African, but the magnitude of the correlation was small (R = 0.1506, p = 0.0018). In general, saliva titers were not correlated with saliva volumes.

Discussion
Herein, we report the results of a half-sib quantitative genetic analysis of SGIB and SGEBs in a freshly colonized strain of Ae. aegypti from Mapastepec, Mexico. Half-sib analysis allows for an assessment of whether there is a genetic basis for a phenotypic trait that varies in expression in nature or in the laboratory. The critical question addressed by this study is whether there is a genetic basis for SGIB and SGEB in Ae. aegypti. The answer is affirmative, but with the qualification that the heritability of these traits in the mosquito is dependent upon the viral genotype, or more specifically, on the growth rate of the virus in the salivary gland. In addition, many of the significant h 2 estimates are small, suggesting that most of the phenotypic variation is still unexplained, and that many uncontrolled factors in the environment contribute to the Vp exhibited in SGIB and SGEB. For each half-sib analysis, we assumed that the variance contributed by the virus should be minimal, because high-passaged virus in cell culture were used in this study. Usually there is abundant genetic variation present in Ae. aegypti field collections [25,49], and this genetic variation may affect the growth of the virus in the mosquito's salivary gland. Examining Figure 6A,B, it is very clear that different half-sib families vary greatly in the amount of virus in salivary glands. The high h 2 suggests that this is due to the segregation of genetic factors that affect viral reproduction among and within half-sib families. In contrast with this idea, a DENV2 genotype (Cosmopolitan) and a CHIKV genotype (Asian) ( Figure 6C,E) exhibited obviously less variation in the amount of virus and higher average viral titers ( Figure 5A), and the h 2 was close to zero. Table 2 indicates significant h 2 values for DENV2 American and Asian genotypes, for CHIKV ECSA-IOL genotype, and for ZIKV Asian genotype. Thus, it would appear that an SGIB following the introduction of virus into the hemocele has a minimal effect on salivary gland infection rates. However, SGIB did affect the quantity of virus present in the salivary gland.
A low percentage of mosquitoes injected with DENV2 genotypes remained uninfected, but all mosquitoes injected with CHIKV or ZIKV genotypes were infected ( Table 1). The extrinsic incubation period varies among virus genotype and mosquito strain, and it is greatly dependent on environmental factors, such as temperature and effective dose of virus [50][51][52]. However, a minimum of one week after initial infection is required for DENV to be released and detected in saliva, indicating an absence of SGEB and SGIB [51]. There is minimal information on relative growth rates of different arboviruses following intrathoracic inoculation in mosquitoes. Since we inoculated all mosquitoes with the same number of PFU, and since we collected all mosquitoes at 7 dpi, our observations suggest, for example, that DENV2 American genotype grows more slowly than the other genotypes and that ZIKV African genotype grows faster (Figure 4). To formally test this hypothesis, we processed salivary glands and saliva from DENV2 American infected mosquitoes at a later extrinsic incubation period (14 dpi). Average virus titers in salivary glands at 14 dpi (2.6 × 10 3 PFU/mL) were significantly higher (p = 0.0003) than at 7 dpi (1.7 × 10 4 PFU/mL). However, the percentage of salivary glands infected and the percentage of saliva infected was similar, at 7 dpi and 14 dpi.
One could also vary the amount of virus injected for ZIKV and CHIKV and then assess genetic contributions throughout the course of infection. We assumed that if the variances in SGIB and SGEB are due to one or all of the random causes listed in the Introduction (such as virus genotype or limit of detection of virus by plaque assay), then Ve will be large, Va will be small, and h 2 will approach zero. The distribution of phenotypes among the progeny of half-sib families will appear as in Figure 6C,E. Alternatively, if the amount of virus detected in salivary glands or from saliva is due, at least in part, to genetic factors, then the variation in SGIB and SGEB will have a significant h 2 . The distribution of phenotypes among the progeny of half-sib families appear as in Figure 6A,B,D,F,G. Once the virus has infected the salivary gland, SGEBs appear to have a major effect on the presence or absence of virus in the saliva (Table 1, Figure 8) and the concentration of virus in the saliva (Table 3). Saliva often contains clusters of virus rather than a uniform distribution [3,22]. Indeed, this explains the poor correlation between the amount of saliva and the amount of virus obtained for all virus strains tested in this study, as has been reported in Ae. albopictus infected with different CHIKV genotypes [40]. A similar argument could be made regarding the dissection of the salivary gland. It is likely that the dissected glands will be contaminated by adjacent or proximal tissues. However, if variation in the amount of virus in the dissected glands or saliva was attributable to random causes, then none of the h 2 estimates should have been significant. However, in fact, four of the seven half-sib analyses of virus titer/salivary gland had a significant h 2 , and in three of the seven half-sib analyses of virus titer/saliva, h 2 was significant. Thus, despite the random contribution of saliva volume and the potential presence of attached tissues during salivary gland dissections, significant heritabilities were still estimated. Further, it is unlikely that we may have underestimated h 2 for SGEB as a result of problems associated with determining the volume of saliva produced by individual females and other uncontrolled factors. It is very clear that the amount of virus can vary greatly in similar saliva volumes (Figure 9).
The heritabilities for DENV2 SGEB were not significant, nor were the variances among sires ( Figure 8, Table 3). In contrast, the heritabilities were significant, as were the variances among sires, when infected with either of the CHIKV genotypes or with ZIKV. Clearly, genetic factors in the mosquito do influence the abundance of CHIKV and ZIKV during salivation. The observed negative correlation between viral titers at the time of harvest and h 2 (Figure 7) is consistent with a threshold model wherein once a viral population has reached a critical density, then genetic factors in the mosquito that influence viral survival and titer no longer have an effect. As the viral population approaches the threshold, the amount of variation among means of sire and h 2 will approach zero ( Figure 7). This pattern suggests that when the amount of a virus is low, genetic factors in the mosquito may influence viral titer in the salivary glands. However, once viral titers reach a high enough level in the salivary glands, the mosquito's genetic contribution may no longer have any effect. Indeed, genetic factors associated with the innate immune response to virus infections are in play here, and the particular combination of genetic factors in a given mosquito determines how successful the virus is, so low or high titers reflect a genetic basis in the mosquito. Negatives values of Vd were obtained from DENV2 American genotype and CHIKV ECSA-IOL genotype for SGIB (Supplementary Table S1), and Vd values were negative for all six genotypes, except the DENV2 Cosmopolitan genotype, for SGEB (Supplementary Table S2). Mathematically, this occurs because the average variation virus titer (log 10 PFU/mL) among dams within sire is less than the average variation among all sires-families. The genetic explanation for this involves either (1) dominance wherein variance among heterozygotes is not intermediate in value between parental types or (2) common environmental variance arising when progeny of a single female is all raised in the same larval container. This leads to two types of environmental variance: variance arising from a common environment (all larvae in a family raised in the same container), and variance arising from half-sibs raised under different, albeit unapparent, conditions. Because Vp = Va + Vd + Ve and h 2 = Va/Vp, a negative dominance reduces the denominator and increases h 2 . Raising half-sib larvae in a common environment could reduce Ve and inflate the covariance among siblings, and this also could inflate h 2 . There is abundant evidence for gene by gene (G × G) interactions between arboviral genotypes and Ae. aegypti [53]. The genetics underlying this vector-pathogen association is still incompletely understood, but variation in vector competence appears to result, in part, from the interplay between two genomes. The findings reported here are consistent with G × G interactions. Mosquito innate immunity genes [54][55][56] impacting viral growth may only be able to act when the viral titer is low. Thus, viral genes that enable rapid growth may overcome innate immunity genes in the mosquito. In our study, variation in the mosquito arises from sampling directly from a field population, where potentially many innate immunity genes are segregating. It is clear that the different genotypes of the virus in any of the three species examined here generate a great deal of phenotypic variation in the virus propagation rates and overcome infection and transmission barriers in mosquitoes.
In terms of infection rates, it is clear that SGEB provides the greatest heritable barrier to transmission in mosquitoes with infected salivary glands.

Conclusions
Herein, we report the results of a half-sib quantitative genetic analysis of SGIB and SGEBs in a freshly colonized strain of Ae. aegypti from Mapastepec, Mexico. The critical question addressed by this study is whether there is a genetic basis for SGIB and SGEB in Ae. aegypti. The answer is affirmative, but with the qualification that the heritability of these traits in the mosquito is dependent upon the viral genotype, or more specifically, on the growth rate of the virus in the salivary gland. This pattern suggests that when the amount of a virus is low, genetic factors in the mosquito may have an influence on viral titer in the salivary glands. However, once viral titers reach a high enough level in the salivary glands, the mosquito's genetic contribution may no longer have any effect. Once the virus reaches a certain threshold, the effects of mosquito genes in determining the amount of virus declines, as does the variance in viral titers among sires. In addition, many of the significant h 2 estimates are small, suggesting that most of the phenotypic variation is still unexplained, and that many uncontrolled factors in the environment contribute to the Vp exhibited in SGIB and SGEB.
Supplementary Materials: The following are available online at https://www.mdpi.com/2075-4 450/12/1/73/s1, Figure S1: The SAS code for performing REML analyses, Table S1: Quantitative genetic analysis of SGIB in Aedes aegypti, Table S2. Quantitative genetic analysis of SGEB in Aedes aegypti.  Data Availability Statement: All data produced from this study are included in this published paper.