Multiple hosts, multiple impacts: the role of 1 vertebrate host diversity in shaping mosquito 2 life history and pathogen transmission

The transmission of malaria parasites from mosquito to human is largely determined by the 15 dietary specialization of Anopheles mosquitoes to feed on humans. Few studies have explored


Introduction 34
The diet of female Anopheles mosquitoes, including those species capable of transmitting human 35 malaria parasites, is characterized by the ingestion of vertebrate blood during each ovarian cycle to sustain 36 vitellogenesis and egg production, while plant carbohydrates are mainly used for energy and maintenance 37 reserves (Clements 1992). With a rather short gonotrophic cycle, which can be as fast as 48h between two 38 egg-lays, mosquito females are recurrently looking for a blood meal during which they can transmit malaria 39 parasites. The successful transmission of the malaria parasite is highly dependent of the diet specialization 40 of the Anopheles vector and, in particular, its degree of anthropophagy (propensity to feed on human). 41 Female mosquitoes must bite a human host twice to potentially transmit malaria parasites. Therefore, the 42 higher the human feeding rate, the greater the transmission potential (Smith and McKenzie 2004) . 43 Furthermore, other key parameters of pathogen transmission such as mosquito longevity (Smith and 44 McKenzie 2004) can be influenced by blood meal source , Lyimo et al. 2013). 45 Consequently, dietary specialization on human could be associated with fitness benefits for mosquito 46 females which could increase parasite transmission rates. 47 Comparison of host use among 111 Anopheline mosquito populations drawn from 52 species showed 48 that 82% of the populations exhibits some level of dietary specialism ('≥ 50% bloodmeals taken from one 49 host type'; Lyimo and Ferguson 2009). Dietary specialization assumes a trade-off between exploitation of 50 different diets which results in fitness benefits when a specialist feeds on specialized resource and costs 51 when exploiting sub-optimal resources. On the contrary, generalism would be expected when the chances 52 of optimal host encounter are low and the costs of waiting are high (Lyimo and Ferguson 2009) and only 53 small differences between resources would be observed with no optimal use of one type of diet. For 54 example, Anopheles arabiensis is rather an opportunistic vector displaying either anthropophilic or 55 zoophilic preferences depending on the geographic area and the relative abundance of humans and cattle 56 (Costantini et al. 1999, Takken and Verhulst 2013). On the other hand, in Anopheles coluzzii, considered as 57 strongly anthropophagic, environmental changes such as the widespread usage of bed nets can induce 58 mosquitoes to feed on more accessible although less preferred host species ). Only a 59 handful of studies have investigated the fitness of anopheline mosquitoes fed on different vertebrate blood 60 (Lyimo et

61
While all studies observed some effects of host type on mosquito fitness traits, the exploitation of less 62 preferred hosts did not seem to strongly impact mosquito fitness so that the predicted relationship 63 between host-preference and fitness benefits was not always confirmed (Lyimo et al. 2013) or could be 64 offset by a second blood meal even on a different non -preferred host ). 65 Blood meal type is also likely to directly impact parasite fitness since parasite growth is fueled by host 66 resources (Shaw et al. 2022). For example, xanthurenic acid, a gametocytogenesis activation factor is 67 synthesized by the mosquito host (Billker et al. 1998), and essential amino acids such as valine, histidine 68 and methionine and leucine are incorporated by parasite oocysts (Beier 1998). Similarly, host lipids are 69 taken up by malaria parasites, probably to sustain its membrane biogenesis (Atella et al. 2009) while these 70 lipids are also central to mosquito immune defenses and reproduction (Briegel et  mosquitoes. Thus, the nutritive quality of the mosquito blood meals following malaria parasite invasion 79 might affect parasite fitness, competition for resources between the parasite and its mosquito host as well 80 as mosquito fitness and ability to cope with infection (Shaw et al. 2022). 81 To our knowledge, only two studies have investigated the effects of blood meal sources taken from 82 different vertebrate host species on mosquito competence for malaria parasites by providing a second 83 blood meal 4 or 8 days post-infectious blood meal, using laboratory colonies of mosquitoes, and cultured 84 clones of parasites (Emami et al. 2017, Pathak et al. 2022. Both studies revealed that the development of 85 the malaria parasite can be influenced by the source of blood consumed following the infection. 86 While Anopheles coluzzii is generally considered highly antropophilic, it can also feed on a wide range 87 of other vertebrate hosts (Lemasson et al. 1997 Emami et al. 2017). In nature, females can be exposed to a wide range of host species and seek 94 a blood meal every 2 to 4 days. Therefore we here used four vertebrate species, provided multiple blood 95 meals (3 to 4) and measured multiple fitness-related traits (feeding rate, blood meal size, competence to 96 parasites, survival, fecundity, F1 development time and wing length) to obtain a thorough picture of the 97 effect of blood-meal diversity on mosquito and parasite fitness. Mosquito females were first fed an 98 infectious or a non-infectious blood meal. They then received up to three subsequent blood meals from 99 either human, chicken, cow or sheep. We predicted that blood type would affect parasite development 100 and mosquito traits such as survival and fecundity and hence vectorial capacity and that effects would add 101 up with subsequent blood meals. Our results were combined into a theoretical model to predict the relative 102 contribution of different vertebrate hosts to overall malaria transmission. 103 Faso. Gametocyte carriers were selected by examining thick blood smears from children aged between 5 123

Methods
and 11 from two villages in southwestern Burkina Faso (Dande and Soumousso, located 60km north and 124 40km southeast of Bobo-Dioulasso, respectively, Figure 1). Malaria positive individuals were treated 125 according to national recommendations. Venous blood from gametocyte carriers was collected in 126 heparinized tubes. As a negative control (uninfected mosquitoes), females were fed on the same blood in 127 which gametocytes were heat-inactivated. This heat-inactivation inhibits the infection and does not affect 128 the blood nutritive quality (Sangare et al. 2013 In addition to the first infectious/uninfectious feed, mosquitoes received two to three additional blood 154 meals every three days through membranes on venous blood drawn from one of four different vertebrate 155 species ("blood type" hereafter): human, cow, sheep or chicken (Fig 2) ). After each blood meal, unfed 156 females were discarded and lost to follow-up. As a result, two different experimental designs were 157 employed: one utilizing cups (Fig 2A), which enabled the tracking of small groups of females and the 158 measurement of several life history traits (see below), and a second design using cages (Fig 2B), which 159 allowed for the monitoring of a larger number of females but without measuring all life-history traits. This 160 second design was solely used for measuring mosquito competence and was analyzed separately (see 161 details below). Mosquitoes were fed on the same vertebrate species for either three successive blood 162 meals resulting in a total of four blood meals (replicates 1 to 3) or two successive blood meals resulting in 163 a total of three blood meals (replicates 4 & 5, Fig 2) (Foster 2022). Therefore, 173 we did not provide a sugar solution to the mosquitoes during the whole experiment as it could hide or 174 compensate for the fitness effects of the different blood types. 175 The absence of malaria parasite in human blood donors at feeding episodes 2, 3 and 4 was confirmed 176 by a blood smear prior to blood collection. This study was carried out in strict accordance with the 177 recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of 178 Health. Animals were cared for by trained personnel and veterinarians. 179

Traits measurements 180
The effect of blood type on a series of mosquito life-history traits, namely (i) competence for P. 181 falciparum (oocyst / sporozoite prevalence and intensity), (ii) feeding rate, (iii) blood-meal size, (iv) survival, 182 (v) fecundity, (vi) progeny developmental time, and (vii) progeny body size, was assessed. 183 For replicates 1 and 2, following the first human (infectious or non-infectious) blood meal, a total of 184 1818 fully-fed An. coluzzii females were randomly distributed in 121 paper cups (10 cm height X 7.5 cm 185 down diameter X 9.5 cm up diameter) by group of 9 to 22 (median 12) per cup. Cups were divided in four 186 groups, of which the first three groups of cups contained females fed on infectious blood (thus containing 187 both exposed-infected and exposed-uninfected females) and the fourth group females fed on non-188 infectious blood. The first (32 cups) and second group (32 cups) were used to investigate the effect of blood 189 type on vector competence, respectively at the parasite oocyst stage 8 days post-bloodmeal (dpbm: days 190 post-bloodmeal) and the sporozoite stage 14 dpbm. The third (30 cups) and fourth group (27 cups) were 191 used to investigate the effect of blood type and P. falciparum exposure on mosquito survival. All other life-192 history traits were measured on mosquitoes from all the cups (n = 121 cups). F1 traits were measured on 193 offspring from replicate 1. Mosquitoes from replicates 3 to 5, maintained in cages, were used to measure 194 competence traits only (Fig 2). 195 Competence -Oocyst prevalence (i.e. proportion of females harboring at least one oocyst on their 196 midgut) and intensity (i.e. number of P. falciparum oocysts in the midgut of infected females) were 197 determined at 8 dpbm. At this time, females had received two additional bloods meal post-infection.

198
Midguts were dissected in 1% Mercurochrome ® stain and the presence and number of oocysts were 199 determined under a microscope (at 20× objective). A total of 119 individuals were used to determine oocyst 200 prevalence derived from 13 cups (4 cups from human and sheep blood, 3 cups from cow blood, 2 cups from 201 chicken blood -between 1 to 5 females per cup) in replicate 1 and 16 cups (4 cups per type of blood -202 between 2 to 11 females per cup) in replicate 2. A total of 757 individuals were used to determine oocyst 203 prevalence in replicates 3 to 5 (202 females in replicate 3, 215 females in replicate 4 and 340 females in 204 replicate 5). Sporozoite prevalence was determined at 14 dpbm. At this time, females had received three 205 (replicates 1 to 3) or two (replicates 4 and 5) additional blood meals post-infection. For each individual, the 206 abdomen was removed and the head and thorax were stored at -20°C. Prevalence was determined using 207 PCR assays on the crushed head and thorax (Morassin et al. 2002). Sporozoite prevalence was determined 208 on 34 individuals derived from 6 cups (2 cups from human, cow and sheep) in replicate 1 and 10 cups in 209 replicate 2 (3 cups from human and cow and 4 cups from sheep). Sporozoite prevalence was assessed on 210 235 individuals in replicates 3 and 5 (175 individuals from replicate 3 and 60 individuals from replicate 5). 211 Feeding rate. The feeding rate of females during meals 2 to 4 was calculated as the number of fully-fed 212 females over the total number of females. This trait was assessed on 121 cups for blood meal 2, on 119 213 cups for blood meal 3 and on 71 cups for blood meal 4. Feeding rate was not measured for the first meal 214 on infectious or non-infectious human blood. 215 Mosquito blood meal size was estimated following each meal (1 to 4) by measuring the amount of 216 haematin (a by-product of the decomposition of haemoglobin) excreted in each paper cup and averaging 217 it by the number of females in the cup (Briegel 1990). One ml of 1% lithium carbonate solution was 218 distributed in each cup to elute faeces and the absorbance of the resulting solution was read at 387nm, 219 using LiCO3 solution as a blank, and compared with a standard curve made with porcine serum haematin 220 (Sigma-Aldrich). Blood meal size was assessed on 116 cups for the first blood meal, on 119 cups for the 221 second blood meal, on 74 cups for the third blood meal and on 60 cups for the fourth blood meal. 222 Mosquito survival was recorded daily at 8:00 by counting and removing the number of dead individuals 223 in the cups. Survival was derived from a total of 855 individuals followed from 57 cups. 224 Fecundity -At each gonotrophic cycle (n=4), petri dishes containing humid cotton covered with a piece 225 of Whatmann ® paper were placed in the cups two days post-bloodmeal. Eggs laid on the Whatmann ® paper 226 were recovered the following morning, pictured and placed in a plastic weighing pan with 25 ml of water. 227 Two days later, pictures of the 1 st instar larvae were taken. The number of eggs were counted using the between egglay 1 and 4. 240 F1 development time was assessed by introducing 1 st instar larvae (median number of larvae = 9, range 241 = 1-13) randomly selected from each weighing pan in a plastic cup with 50ml of water. Mosquito larvae 242 were provided with Tetramin ® Baby Fish Food ad libitum once a day, and excess food was removed to avoid 243 water pollution. Development time was calculated as the duration from egg-lay to emergence and was 244 measured for a total of 1,035 individuals from the four egg-lays of the first replicate (464 individuals from 245 56 cups in egg-lay 1, 303 individuals from 36 cups in egg-lay 2, 142 individuals from 18 cups in egg-lay 3, 246 126 individuals from 15 cups in egg-lay 4).

247
F1 wing length was used as a surrogate of body size and was measured from the alula to the wing tip, 248 excluding scales (Van Handel and Day 1989). One wing per F1 individual was dissected on the day following 249 emergence on a subset of individuals. The wing was pictured with a stereomicroscope and measured with 250 ImageJ software (Wayne Rasband, rsb.info.nih.gov/ij/). Wing length was measured on 656 individuals of 251 the first replicate (300 individuals from 55 cups in egg-lay 1, 176 individuals from 34 cups in egg-lay 2, 100 252 individuals from 18 cups in egg-lay 3, 80 individuals from 15 cups in egg-lay 4). 253

Statistical analyses 254
Competence -Parasite prevalence (oocyst or sporozoite stages) and intensity (oocyst stage only) were 255 analysed using Generalized Linear Mixed Models (GLMMs) with a binomial and a zero-truncated negative 256 binomial error structure respectively. The replicates in cups and the replicates in cages were analyzed 257 separately. In these GLMMs, blood type (four levels: cow, sheep, chicken or human blood), gametocytemia 258 and their interaction (only for replicates 3-5) were coded as fixed factors, and cup and parasite isolate 259 nested in replicate (for replicates 3-5) as random factors. 260 Feeding rate was analysed using a GLMM with a binomial error structure. In this model, blood type, P. 261 falciparum exposure (two levels: mosquito previously fed an infectious blood meal vs fed the same heat-262 inactivated blood), blood feeding episode (three levels: 2 to 4) and their interactions as well as parasite 263 isolate were coded as fixed factors and cup as a random factor. 264 For the following traits, data from the first gonotrophic cycle (resulting from infectious vs. non-265 infectious human blood) were analysed separately from data from gonotrophic cycles 2-4 for which 266 mosquitoes were fed on four different types of blood (human, cow, sheep, chicken). Data analyses and 267 results from the first gonotrophic cycle are presented in the supplementary material. 268 Mosquito blood meal size -Data from the blood meals 2 to 4 were log-transformed before being 269 analyzed with a GLMM with a Gaussian distribution. In this model, blood type, mosquito exposure, blood-270 feeding episode and their interactions as well as parasite isolate were coded as fixed factors and cup as a 271 random factor. 272 Survival data were analysed using Cox proportional hazard mixed models (coxme package) with 273 exposure to infectious blood, blood type, parasite isolate and their interactions coded as fixed factors and 274 mosquito cup as a random factor. Since unfed females from blood meals 2 to 4 were removed, they were 275 given a censoring status of 0 indicating that the individual was alive when last seen. 276 Fecundity -Egg-laying rate, hatching rate, average number of eggs (log-transformed), and average 277 number of 1 st instar larvae (log-transformed) over gonotrophic cycles 2-4 were analysed using GLMMs with 278 binomial or Gaussian error structures. Blood type, exposure, isolate, blood meal size and gonotrophic cycle 279 were coded as fixed factors and cup as a random factor. In addition, GLMs with quasipoisson structure (to 280 correct for overdispersion) were used to analyze the effect of blood type, exposure, isolate and their 281 interactions on the lifetime fecundity and lifetime production of larvae corresponding to the sum of the 282 average number of eggs and 1 st instar larvae over gonotrophic cycles 2-4. 283 The development time of larvae from gonotrophic cycle 2-4 was analyzed using a a Cox proportional 284 hazard mixed effect model with maternal exposure, maternal blood type, gonotrophic cycle, larval density 285 and mosquito sex coded as fixed factors, and rearing cup as a random factor. The effect of blood type on 286 the sex ratio of the progeny was analyzed using a binomial GLMM with blood type coded as a fixed factor 287 and rearing cup as a random factor. 288 F1 wing length-A Gaussian GLMM was used to explore the effects of maternal blood type, maternal 289 exposure, egg-lay episode, larval density and mosquito sex on log-transformed wing length of the progeny 290 from gonotrophic cycles 2 to 4. 291 For model selection, we used the stepwise removal of terms, followed by likelihood ratio tests (LRT). 292 Term removals that significantly reduced explanatory power (P<0.05) were retained in the minimal 293 adequate model (Crawley 2007). All analyses were performed in R v. 3.0.3 (R Core Team 2020). Results are 294 presented as mean ± standard error (se) and proportion ± confidence interval (CI). 295 Theoretical modelling 296 We explored the relative contribution of the blood type on mosquito mean individual vectorial capacity 297 (Saul et al. 1990). Individual vectorial capacity (IC) is the mean number of infectious bites given by an 298 infected vector (i.e. the number of bites it gives after the Plasmodium extrinsic incubation period is 299 completed). Therefore, IC expresses the efficiency with which individual mosquitoes transmit malaria. To 300 estimate IC, we developed a model that simulates the daily life history of individual mosquito vectors after 301 taking an infectious blood meal on a human under various scenarios (Fig 3). The environment (= scenario) 302 was characterized by the presence of humans and an alternative host (either chicken, cow or sheep) with 303 varying availability (0 to 3 consecutive possible feeding attempts during Plasmodium incubation period). 304 There was therefore 12 scenarios tested (3 alt. host x 4 availability levels). 250 000 individuals (representing 305 500 populations of 500 individuals) were simulated per scenario. The model allowed to track daily 306 physiological states (either Host-Seeking, HS; Blood-Fed, BF; or Resting, R) of individuals. Daily transitions 307 from one state to another depended on survival probability (related to the origin of the previous blood 308 meal) and blood-feeding success probability (related to the host that the mosquito is attempting to bite: 309 human, chicken, cow or sheep; only for transition from HS to BF). A binomial GLMM of feeding success and 310 a COXPH model of survival were fitted to the data presented in the manuscript and used to calculate host-311 specific probabilities of feeding success and daily survival. For each individual simulation, the number of 312 days spent in state BF (= number of successful feeding attempts) following the duration of Plasmodium 313 extrinsic incubation period (n = 11 days) was counted and the mean (= IC) was calculated for each 314 population.

354
Design concept: Mosquito-to-human transmission may occur during each bite after the pathogen 355 incubation period is completed. Transmission is therefore dependent on both the longevity of the 356 mosquito and the duration of its gonotrophic cycle (i.e. the mean time between two consecutive bites). 357 Longevity vary according to the daily survival probability, duration of the gonotrophic cycle vary according 358 to feeding success probability (as feeding is delayed in case of failure).

359
Stochasticity occurs for daily transition between physiological states since feeding success and survival 360 are Bernoulli trials with host-specific probabilities of success calculated from (i) a binomial GLMM of 361 feeding success and (ii) a Cox proportional hazard model of survival (see sub-model section below), 362 respectively. 363 Observation: For each individual simulation, the number of days spent in state BF (= number of 364 successful feeding attempts) following the duration of Plasmodium extrinsic incubation period (n = 11 days) 365 is counted and population mean is calculated. Populations are made of 500 independent individuals. 366 Initialization: Initially, a female vector is in physiological state "Blood-Fed" with previous blood meal 367 taken on human (the infectious blood meal). 368 Sub-models: Probabilities used in the Bernoulli trial of feeding success were provided by a binomial 369 mixed effect model of feeding success fitted on data from the current study. Feeding success was modelled 370 according to blood meal source and with blood meal episodes and cup of origin (of the mosquito batch) as 371 crossed random effect, using the glmmTMB function. Feeding success probabilities according to blood 372 meal source were extracted using the emmeans function and are shown in Table 1. Probabilities used in 373 the Bernoulli trial of survival were derived from the result of a Cox Proportional Hazard mixed effect model 374 of survival fitted on data from the current study. Death events were modelled according to blood meal 375 source and replicates with cups of origin (of the mosquito batch) as random effect, using the coxme 376 function. Hazard ratios (relative to human blood source) were extracted using the emmeans function (Table  377 1). In the simulations, we set daily survival probability with a previous blood meal taken on human to be 378 0.8. Survival probabilities for mosquitoes having taken their previous blood meal on animals was therefore 379 0.8 exponentiated by the corresponding Hazard ratio (Table 1). 380

403
show the median values. Each colored point represents a P. falciparum-infected mosquito individual. Pies show the 404 infection prevalence (grey area). Numbers indicate the sample size (n = total number of mosquito females for parasite 405 prevalence or number of infected females for parasite counting) for each treatment and isolate.

418
Feeding rate -Blood type significantly affected mosquito feeding rate (X 2 3 = 14.4, P = 0.002) with 419 highest overall feeding on sheep blood (66.32±1.8%) followed by chicken blood (64.6±1.9%), human blood 420 (64.1±1.9%) and cow blood (58.8±1.9%). There was a significant interaction between blood type and 421 feeding episode (X 2 6 = 37.8, P< 0.0001; Fig 5A), with cow blood providing lowest feeding rate during the 422 second blood-meal and highest rate during blood-meals three and four. Mosquitoes that received an 423 infectious blood meal displayed lower feeding rate than mosquitoes previously fed an uninfectious blood-424 meal (61.4 ± 1.9% vs. 71.1 ± 1.8%; X 2 1 = 9.09, P = 0.003, Fig 5B), regardless of the blood type (exposure 425 :blood type: X 2 3 = 0.38, P=0.94, Fig 5B) and of the feeding episode (exposure:feeding episode: X 2 2 = 0.03, 426 P=0.99). Feeding rate consistently increased over the successive feeding episodes (53.1 ± 2, 81.2 ± 1.5, and 427 89.1 ± 1.2% at the 2 nd , 3 rd and 4 th episode, respectively; X 2 2 = 160.2, P< 0.0001, Fig 5A). No other effects 428 were found (Appendix 2- Table S2). 429 430 Figure 5. Effects of blood type on mosquito feeding rate and blood meal size. (A) Feeding rate (number of fed 432 females/number of alive females) ± 95% CI as a function of blood feeding episode and blood meal type. B) Feeding rate as 433 a function of blood meal type and infection group (females exposed vs. females unexposed to an infectious blood meal on 434 feeding episode 1). Bars show the average feeding rate across feeding episodes 2 to 4 + 95% CI. C) Average blood meal size 435 ± se as a function of blood feeding episodes and blood meal type. D) Average blood meal size + se as a function of blood 436 type and infection group (females exposed vs. females unexposed to an infectious blood meal). Bars show the average 437 meal size across feeding episodes 2 to 4.

439
Mosquito blood meal size -Blood type did not significantly affect meal size (X 2 3 = 4.2, P = 0.24, Fig 5C). 440 Meal size varied among feeding episodes (X 2 2 = 54.04, P < 0.0001) with biggest size observed for the fourth 441 bloodmeal. There also was a significant blood type by feeding episode interaction (X 2 6 = 23.7, P = 0.0006; 442 Fig. 5C) such that blood type providing highest or lowest meal size were not always the same across feeding 443 episodes. Meal size was not influenced by the previous exposure of mosquitoes to parasites (X 2 1 = 0.18, P 444 = 0.67) regardless of the blood type (exposure: blood type: X 2 3 = 4.7, P = 0.19; Fig 5D) or the feeding episode 445 (exposure: feeding episode: X 2 2 = 0.8, P = 0.67). No other effects were found (Appendix 2- Table S2). 446 Survival -Mosquito survival over the duration of the experiment was strongly influenced by the blood 447 type (X 2 3 = 68.26, P < 0.0001; Fig 6), with lowest survival observed when females were successively fed with 448 chicken blood (median survival time: chicken: 6 days, cow: 9 days, human: 10 days and sheep 11 days). 449 Females fed on isolate B during the first feeding episode survived significantly longer than females fed on 450 isolate A (10 and 6 days respectively; X 2 1 = 24.38, P < 0.0001, Appendix 3- Fig S1). Mosquito exposure to P. 451 falciparum gametocytes did not significantly influence mosquito survival (9 days for both unexposed and 452 exposed mosquitoes, X 2 1 = 0.03, P =0.87) regardless of the blood type (exposure: blood type: X 2 3 = 2.38, 453 P=0.50; Fig 6) or the isolate (exposure: isolate: X 2 1 =0.27, P =0.96). No other effects were found (Appendix 454 2- Table S2). 455 456 457 Figure 6. Effects of blood type (colored lines) and infection group (dashed lines: females exposed vs. solid lines: 458 females unexposed to an infectious blood meal on feeding episode 1) on mosquito survivorship.

Theoretical modelling 542
Our simulations showed an average IC (the average number of infecting bites transmitted by an 543 infected mosquito during its lifetime) of 0.16 infectious bites when females fed on human hosts only 544 (corresponding to the values with zero feeding attempts on the alternative host in Fig. 9). Compared to a 545 human blood meal, feeding on chicken blood drastically reduced the individual vectorial capacity. There 546 was 2.6 times fewer infectious bites after a single potential blood meal on chicken (with a 0.796 probability 547 of feeding success; mean IC=0.06) and even 26 times fewer after 3 potential blood meals on chicken (mean 548 IC=0.006; Fig. 9). Although less marked, a similar decrease was observed when females obtained blood 549 meals on cow with almost halved of the individual vectorial capacity after 3 potential blood meals on cow 550 (with a 0.717 probability of feeding success; mean IC=0.09; Fig.9). On the contrary, feeding on sheep during 551 Plasmodium development increased the vectorial capacity by 25% after 3 potential blood meals on sheep 552 (with a 0.774 probability of feeding success; mean IC=0.2; Fig. 9). 553 554 555

563
The relationships between blood meal type, and mosquito and parasite fitness were explored using a 564 total of 2810 An. coluzzii females initially fed an infectious or a non-infectious blood meal and then followed 565 by up to three subsequent blood meals from either human, chicken, cow or sheep. We found no significant 566 effect of blood type on malaria parasite development both at the non-transmissible (oocyst) and the 567 transmissible (sporozoite) stages for either parasite prevalence or intensity. No effects at the oocyst stage 568 or negative effects at the sporozoite stage were also found in An. gambiae s.s. females consuming a second 569 blood meal on cow compared to human blood or unfed controls, whereas higher oocyst and sporozoite 570 prevalences were observed in An. arabiensis females having a second blood meal on cow compared to 571 human blood or unfed controls (Emami et al. 2017). In another study, a second blood meal shortened 572 parasite development in P. falciparum with no group specific differences, while a marginal increase was 573 observed with human blood compared to unfed controls for P. berghei development (Pathak et al. 2022). 574 Thus, the effects of blood type on Plasmodium sp. development seem to be variable both between and 575 within Plasmodium species and our results with sympatric field strains do not seem to confirm those 576 previous findings. Findings obtained in the laboratory using unnatural host-parasite associations or long-577 time derived strains do not always reflect natural interactions and an increasing number of studies 578 highlights the importance of confirming laboratory observations with more natural systems for studying 579 disease ecology and evolution (Aguilar et al. 2005, Tripet et al. 2008. 580 Exposure to malaria parasite with wild isolates resulted in 78% (isolate A) and 52% (isolate B) infected 581 females and had, overall, no effect on female fitness traits. Parasite exposure did not affect female's 582 survival regardless of the host type they fed on nor did it affect female fecundity nor their progeny 583 development time and wing size. The only indirect cost we observed was a lower feeding rate in the 584 following blood meals of exposed females compared to unexposed females. The existence of fitness costs 585 of malaria parasite infection in the mosquito host has long been debated and seem to depend on the 586 environmental conditions under which the fitness traits are measured. First, mortality is more commonly 587 reported in unnatural parasite-vector combinations . Second, fitness costs are 588 also more commonly observed in stressful environmental conditions (Lalubin et al. 2014, Sangare et al. 589 2014, Roux et al. 2015) and can depend on the genetic background (Alout et al. 2016). In our experiment, 590 we used a natural parasite-host combination and did not provide any sugar-meals to not alleviate potential 591 fitness costs. Indeed, sugar feeding can affect mosquito competence, survival and fecundity ((Ferguson and 592 Read 2002, Lambrechts et al. 2006, Foster 2022) and could compensate for the fitness costs of the different 593 blood types. One explanation could be that the cost of infection are minimal in our system and that 594 infection might only be costly for exposed-infected females. Following the infectious blood meal, our setup 595 did not separate exposed-infected and exposed-uninfected females, thus we were not able to measure the 596 cost of infection in exposed-infected females only. Another possibility is that those costs are quickly offset 597 by the following blood meals the females received, although a study on Plasmodium relictum and Culex 598 pipiens observed fecundity costs following the infection which lasted for three consecutive gonotrophic 599 cycles (Pigeault and Villa 2018). 600 Blood type strongly affected mosquito survival. In particular, chicken blood reduced mosquito 601 survivorship by 40% (Fig 6). The larger and nucleated red blood cells of chicken compared to human might 602 be more difficult to digest for An. coluzzii (Wintrobe 1933). In addition, although anopheline mosquitoes 603 can reduce their body temperature while blood feeding (Lahondère and Lazzari 2012), chicken host 604 temperature might be too high for evaporative cooling as we saw increased mortality after each blood 605 meal. Interestingly, females fed on sheep blood had the highest survival followed by human which 606 translated in an increased vectorial capacity in our transmission model (Fig 9). Indeed, the type of host the 607 female feeds on strongly increase or decrease the average number of infectious bites in the population (Fig  608  9). Modeling of malaria transmission showed that mosquito survival is the factor with the biggest impact 609 on transmission (MacDonald 1957). Indeed, parasite development is relatively long (10-14 days) compared 610 to mosquito survival (2-3 weeks). Consequently, females will be infectious for a limited period of time only 611 and any small changes in mosquito longevity will dramatically affect malaria transmission (Smith and 612 McKenzie 2004).

613
Vectorial capacity is also very sensitive to feeding rates (Brady et al. 2016). Our model highlights how 614 even minute differences in survival and feeding rates, such as those observed between females fed with 615 human and sheep blood (Fig 5B & 6), can cause large variation in vectorial capacity (Fig.9). Mosquito 616 feeding rate was highest on sheep blood followed by chicken, human and cow blood. Although membrane 617 feeders were maintained at a temperature corresponding to each vertebrate body temperature, the 618 difference is likely linked to the blood characteristics for the females as sheep blood was maintained at 619 39°C which is close to cow blood temperature, 38.5°C. Comparison of feeding rates between several 620 vertebrates with blood maintained at the same temperature showed large variation depending on the 621 mosquito species (Phasomkusolsil et

627
We measured several individual fecundity traits (egg laying rate, average number of eggs, average 628 number of 1 st instar larvae, eggs and larvae prevalence, hatching rate) and found no effect of blood type. 629 However, the lifetime fecundity and production of larvae corresponding to the sum of the average number 630 of eggs or larvae of gonotrophic cycles 2, 3 and 4, were affected by the blood type which was lower for 631 chicken compared to human (Fig 7C, D). Thus, our proxy of the lifetime fecundity showed that the fitness 632 of females fed on chicken or cow blood was lower than the fitness of females fed on human or sheep blood. 633 Indeed, the small differences observed at each gonotrophic cycle were not significantly different, but the 634 accumulation of all those small differences gave overall a difference when looking over the lifetime 635 production of eggs and larvae. 636 We observed a donor effect on fecundity with females fed on blood from donor A on their first blood 637 meal laying more eggs and having a higher hatching rate and more larvae than females fed on donor B. The 638 difference between the two donors in the average number of eggs and larvae in blood meals 2, 3 and 4 639 tended to blur over the successive gonotrophic cycles (Fig S4A & S4D), such that the successive blood meals 640 slowly offset the difference in donor blood quality. 641 The survival rate was negatively associated with the average number of eggs laid ( Fig S3B)

649
Although blood type did not inluence progeny development time, progeny from females fed on chicken 650 blood was smaller than the progeny from females fed on all other vertebrates blood (Fig. 8), which 651 highlights the fitness cost of feeding on chicken blood. Larger females generally take larger blood meals, 652 lay more eggs, live longer and are more competent (Briegel 1990 limitations. First, we measured averaged fecundity traits per cup since multiple blood-feeding over the 660 lifetime of each female individually would have been technically challenging (i.e. using a single membrane 661 feeder for each individual female), and at the minimum would have strongly reduced our sample size. 662 Second, the laboratory colony used is replenished with wild mosquitoes, however processes such as 663 genetic drift or selection by artificial feeding in the laboratory can happen on a very short time frame 664 especially in small population sizes and those could have eroded the specialization on human blood. Third, 665 specialization on humans could be linked to other ecological or behavioural factors which might exert 666 stronger selection pressures than blood characteristics. While we investigated the effects of the blood type, 667 this was disconnected from the effects of the host type as a whole since other characteristics were not 668 taken into account such as e.g. defensive behavior ( ). In addition, even though we observed a lower 671 feeding rate of exposed females compared to unexposed females in the following blood meals, all 672 successive blood meals were carried out on membrane feeders and results could be different on whole-673 body hosts. Fourth, our laboratory setting did not take into account the natural blood foraging rhythm of 674 the vector nor the circadian rhythm of the parasite which have both been shown to influence mosquito 675 and parasite fitness (Schneider et  Here, successive meals were taken from one of the four vertebrate species used. Under natural conditions, 682 mosquitoes can shift from one host species to another between their gonotrophic cycles. It would be of 683 particular interest to examine the effect of successive meals taken from different host species on the traits 684 measured here. The type and frequency of blood meals not only has consequences on reproduction, 685 survival and epidemiology but also on many other physiological and ecological aspects such as e.g. the 686 maintenance of insecticide resistance phenotype for longer period (Oliver and Brooke 2014, Oliver et al. 687 2022) or an increased ivermectin susceptibility in previously bloodfed females (Seaman et al. 2015). Our 688 findings emphasize that considering the diversity of vertebrate blood-meal sources is important to better 689 understand the ecology of mosquitoes as well as their capacity to transmit malaria parasites. 690 Overall, blood type had a significant impact on mosquito survival and feeding rates, leading to 691 considerable variation in vectorial capacity and differences in progeny sizes. These findings imply that the 692 diversity of vertebrate hosts (including both the number of species and their relative abundance) within 693 villages could influence the transmission of malaria parasites. Specifically, transmission may decrease when 694 chickens or cows make up the majority of available blood sources, while it may increase when a relatively 695 large number of sheep are present. However, the host selection patterns of mosquitoes are not solely 696 driven by vertebrate abundance, but are also influenced by mosquito innate preferences and host 697 defensive behavior (Lefèvre et al. 2009, Lyimo andFerguson 2009