**4. Discussion**

The current tunnel test uses one sample of ITN with 100 mosquitoes as the unit of replication and based on the current work, it is proposed that a larger number of nets or two samples per ITN can be tested using 50 mosquitoes per replicate to improve laboratory throughput. Biological durability monitoring requires large sample sizes, as nets are exposed to highly variable use patterns [44–48] and environmental conditions [49,50] that result in a high degree of heterogeneity between individual nets. The goal of biological durability monitoring is the precise estimation of the biological efficacy of a population of ITNs. As the ITN is the unit of replication, greater precision is obtained by evaluating larger numbers of ITNs.

The current experiment confirmed that using the Hemotek® membrane feeding system as a blood source, together with a worn sock emitting human odor with a replicate size of 50 mosquitoes, a similar mortality and feeding inhibition as the standard WHO tunnel bioassay with rabbit and a replicate size of 100 mosquitoes for both pyrethroid and mixture pyrethroid and chlorfenapyr ITNs is estimated. Our results sugges<sup>t</sup> that a membrane bioassay can evaluate the difference between ITNs because the membrane assay estimates the superiority of Interceptor® G2 over Interceptor® on the M72 outcome using metabolic-resistant mosquitoes, which was also measured by the gold standard rabbit-100 assay and has been consistently seen in other studies in the WHO tunnel, I-ACT, and experimental hut [23]. It was also able to predict the superior blood feeding inhibition of Interceptor®, which has a higher concentration of the pyrethroid alpha-cypermethrin (200 mg/m<sup>2</sup> alpha-cypermethrin in Interceptor® and 100 mg/m<sup>2</sup> alpha-cypermethrin in Interceptor® G2). Being able to test differences between products is the goal of durability monitoring bioassays that track the bio-efficacy of ITNs over time (biological durability) and compare them to unwashed positive controls [41].

Having a reliable bioassay that can be conducted routinely without animal welfare concerns will be extremely useful. The data generated by the current work are promising and further work is planned to improve mosquito feeding success on the membrane as it was seen that differences in blood feeding success do impact on the mortality estimates. While this did not impact on the predictions of superiority, and therefore mortality can still be compared to an unwashed positive control net, if thresholds are used, i.e., the proportion of nets that meet WHO bio-efficacy criteria, then this might affect the interpretation of the bioassay results. It is recommended that the results are replicated in additional laboratories, since having an assay that can accurately predict the differences between net samples in multiple laboratories with several pyrethroid-resistant mosquito strains and that can predict the results of experimental hut studies is optimal. Data from the experiments

demonstrated that several factors influenced the mean mortality and feeding inhibition estimated in WHO tunnel tests [22,51,52].

#### *4.1. Impact of the Bait*

The use of different baits had an enormous influence on the bioassays. By using a human arm as bait, feeding inhibition was substantially lower compared to membrane or rabbit baits [53]. This has also been seen in early versions of the tunnel test [51]. This preference for the human arm is unsurprising, since the colony used in the experiments is anthropophilic. Therefore, although it is more representative of end-user conditions, the use of a human is not recommended for ITN evaluation, because the results were not comparable to those of the rabbit bioassay that was shown to predict the results of experimental hut trials in this setting [23] and elsewhere [9,22,52]. Study findings using *An. arabiensis* mosquitoes were consistent with the existing literature on vector host preference [25,54], confirming that mosquitoes are most attracted to humans as bait, followed by rabbits, and were least attracted to the membrane. Lower attraction in assays using the Hemotek® membrane system and rabbits reduces the number of mosquitoes passing the ITN tested, resulting in higher feeding inhibition compared to when the human arm was used as bait. Several other studies have shown that host-seeking *An. arabiensis* are more attracted to humans than to live animals [24,25]. The lower attraction and consequent higher feeding inhibition when using a membrane is likely due to the absence of carbon dioxide (CO2) that increases mosquito responses to kairomones [55] and the small size of the membrane feeder's surface, which reduces the amount of heat and moisture available, which are both important short-range attractants to mosquitoes [56–58]. The validity of the experiment relies on the negative control feeding success of (>50%) for rabbits. In this assay, with the membrane, augmentation with socks that contained human kairomones improved the attraction of the membrane to mosquitoes [59]. However, it was not possible to use the same threshold value for feeding success with the less attractive membrane. For this reason, further work is needed to optimize the attraction of the membranes for use in the WHO tunnel test. Further improvements to the attractiveness of the membrane could be achieved by making a larger surface area available [60,61] and the addition of 2-butanone [62] or CO2 [63] to augmen<sup>t</sup> mosquito response to kairomones until 50% feeding success in the negative control is consistently achieved.

#### *4.2. Impact of Exposure Time*

Exposure time was important with 12 h exposure, increasing both mortality and feeding success, indicating that the mosquitoes make repeated contact with the ITN sample overnight. Consistently, prolonged exposure (12 h) increased the efficacy of insecticide and host-seeking activities compared to 1 h exposure, resulting in increased mortality because of a higher dose of insecticide picked up by the mosquitoes when resting, bouncing, and passing the ITNs repeatedly. This is also likely in experimental huts and in the community where ITNs are in use. Therefore, the use of a 12 h overnight exposure is recommended. For insecticides that require the mosquitoes to be metabolically active, such as chlorfenapyr, prolonging exposure to 12 h allows the conversion of parent molecules into active forms, because of mosquitoes' metabolic activity when flying in the tunnel. Interestingly, results show that with either the pyrethroid only Interceptor® or the pyrethroid-chlorfenapyr Interceptor® G2 ITNs higher mortality was observed among unfed mosquitoes. Therefore, the results of this study underline the WHO recommendation that feeding success should always be reported when conducting WHO tunnel tests, as low feeding rates will affect the interpretation of results.

#### *4.3. Effects of Mosquito Density*

It was observed that the use of 50 or 100 mosquitoes per testing sample with the rabbit bait did not significantly alter the mortality and blood feeding success measured with either resistant *An. arabiensis* or susceptible *An. gambiae* for the pyrethroid only net or the

mixture ITNs. These results sugges<sup>t</sup> that fewer mosquitoes can be used in WHO tunnel bioassays and still correctly measure the efficacy of ITNs. As would be expected, with 50 mosquitoes there is a slight increase in blood feeding success and a consequent slight decrease in mortality compared to assays using 100 mosquitoes. Higher feeding success at a lower density is likely due to less competition between mosquitoes on the membrane during host-seeking [35], which may also reduce the host defensiveness of the rabbit [64,65]. Increasing the number of mosquitoes in the tunnel may lead to density-dependent mortality effects of crowding as mosquitoes can disturb each other when at a high density [66]. Our results sugges<sup>t</sup> that regardless of the insecticides on the ITNs tested, mortality was higher among unfed mosquitoes, revealing an impact of blood feeding on increased mosquito resilience to insecticides after a blood meal. A similar study on the effects of bites through permethrin nets shows that successfully fed mosquitoes survive longer than unfed ones [67]. This has been reported for chlorfenapyr, where observed mortality was lower among blood-fed mosquitoes compared to those who were unfed [10]. Blood feeding elevates detoxifying enzymes (glutathione, monooxygenase), which then assist in the detoxification of insecticides [68], although this did not translate into substantially lower bio-efficacy with Interceptor® G2 as upregulation of metabolization converts the parent molecule into the potent n-dealkylated form that elicits increased mosquito mortality [21]. It is also important to report control blood feeding success because unfit colony mosquitoes are less likely to fly and feed, which reduces the likelihood that the mosquitoes contact treated nets [67], nullifying the bioassay.
