Common predators and factors influencing their abundance in Anopheles funestus aquatic habitats in rural south-eastern Tanzania

Background: The role of larval predators in regulating the malaria vectors population remains relatively unknown. This study aimed to investigate the common predators that co-exist with Anopheles funestus group larvae and evaluate factors that influence their abundance in rural south-eastern Tanzania.

Methods: Mosquito larvae and predators were sampled concurrently using standard dipper (350 ml) or 10 L bucket in previously identified aquatic habitats in selected villages in southern Tanzania. Predators were identified using standard identification keys. All positive habitats were geo-located and their physical features characterized. Water physicochemical parameters such as dissolved oxygen (DO), pH, electrical conductivity (EC), total dissolved solids (TDS) and temperature were also recorded.

Results: A total of 85 previously identified An. funestus aquatic habitats in nine villages were sampled for larvae and potential predators. A total of 8,295 predators were sampled. Of these Coenagrionidae 57.7% (n=4785), Corixidae 12.8% (n=1,060), Notonectidae 9.9% (n=822), Aeshnidae 4.9% (n=405), Amphibian 4.5% (n=370), Dytiscidae 3.8% (n=313) were common.

A total of 5,260 mosquito larvae were sampled, whereby Anopheles funestus group were 60.3% (n= 3,170), Culex spp. 24.3% (n= 1,279), An. gambie s.l. 8.3% (n= 438) and other anophelines 7.1% (n= 373). Permanent and aquatic habitats larger than 100m2 were positively associated with An. funestus group larvae (P<0.05) and predator abundance (P<0.05). Habitats with submerged vegetation were negative associated with An. funestus group larvae (P<0.05). Only dissolved oxygen (DO) was positively and significantly affect the abundance of An. funestus group larvae (P<0.05). While predators’ abundance was not impacted by all physicochemical parameters.

Conclusion: Six potential predator families were common in aquatic habitats of An. funestus group larvae. Additional studies are needed to demonstrate the efficacy of different predators on larval density and adult fitness traits. Interventions leveraging the interaction between mosquitoes and predators can be established to disrupt the transmission potential and survival of the An. funestus mosquitoes.

Methods
Study area

A cross-sectional survey was conducted, between March and May 2022, in nine villages in south-eastern Tanzania, namely Chikuti (-8.6028°, 36.7288°), Mzelezi (-8.8934°, 36.7343°), Chirombola (-8.93041°, 36.75753°), Ebuyu (-8.9719°, 36.7608°), Mwaya (-8.91022°, 36.823139°) and Tulizamoyo (-8.35447°, 36.70546°) in Ulanga district and Ikwambi (-7.97927°, 36.81630°), Kisawasawa (-7.89657°, 36.88058°) and Mpofu (-8.17220°, 36.21651°) in Kilombero district (Fig 1). In this area An. funestus is responsible for more than 85% of overall malaria transmission [17]. The residents in these villages practise extensive rice farming, which creates suitable habitat for mosquito breeding. Common aquatic habitats for An. funestus in the villages are well known and have been previously characterized [24]. Eighty-five known habitats from the nine villages were sampled for both mosquito larvae and potential predators.

Fig1: Map of Kilombero and Ulanga districts showing the nine study villages

Sampling and morphological identification of mosquito larvae and aquatic predators

Mosquito larvae and predators were sampled using standard dippers (350 ml) or 10 L buckets, as previously described [8,15,24]. A minimum of 3 dips and a maximum of 20 dips were taken depending on the size and depth of the habitat. 

In a previous study mosquito larva from the same villages were taken to the laboratory in Ifakara, allowed to emerge and eventually identified to species by PCR. Of those identified 53% were An. funestus s.s. whilst 28% were An. rivulorum and 12% were An. leesoni [24]. All three species were found to occupy the same habitats. A similar approach was followed with samples of fourth instar larvae during the present study but identification to species level was not performed. Earlier stage larvae were identified based on their predominant characteristics as done in a previous study and separated into An. funestus group, An. gambiae s.l. or Anopheles sp. following the identification key by Gilles and Coetzee [25,26]. Culicines were identified to genera only. Predators were morphologically identified to family level using the keys of the Stroud Water Research Center [27] and Gerber and Gabriel [28]. Mosquito larvae and predators that were sampled by each dipper or bucket were counted and recorded. Additionally, geographical locations of the surveyed habitats were recorded at access points using a hand-held GPS device (Garmin eTrex 20x Handheld GPS Receiver).

Aquatic habitats characterization

Only positive aquatic habitats for An. funestus group larvae were sampled for mosquito larvae and predators. Their overall physical characteristics were recorded and physicochemical parameters of the water (pH, temperature, electrical conductivity (EC), total dissolved solids (TDS) were measured using a portable water quality meter (ZJ practical 4 in 1 Water Tester). A Trans Instruments Dissolved Oxygen Meter (HD3030) was used to measure dissolved oxygen (DO), using standard recording procedures. Habitats were classified as being either: swamp, stream, river, rice-field, stream-pool, ground-pool, ditch, spring-fed pool, puddle, hoof-print, man-made wells, brick or sand pit. Water colour was categorized as being clear (transparent and odourless) or coloured (cloudy, not transparent, turbid or with a film of oil).

The source of water was also classified as rainwater or others (non-rainwater). Algal quantities in the habitats were classified as none, moderate, or abundant. Algal type was classified as filamentous, green, blue-green or brown. Water was also classified as being stagnant, slow or fast moving. The land use surrounding the aquatic habitats was classified as scrub, cattle grazing or cultivated field. Shade over the habitats was classified as none, partial or heavy. Habitat size was measured using tape and classified as being less than 100 m2 or more than 100 m2. Vegetation quantity and vegetation type were also classified as (none, moderate or abundant) and (emergent, or submerged) respectively. Water bodies known to have existed for three months or more were considered to be permanent whilst other collections of water were considered to be ‘temporary’. Water depth was classified as being less than 50 cm or more than 50 cm deep.  The distance from aquatic habitats to the nearest houses were estimated visually and classified as being less than 100 m or more than 100 m.

 Statistical analysis

Analysis was done using open source software R version 4.2.1. [29]. Generalised linear mixed effects models (GLMM)  using template model builder (TMB) with zero-inflated negative binomial implemented under the glmmTMB package [30] were used to (i) assess the associations between water physicochemical parameters and the abundance of aquatic predators ii) assess the associations between water physicochemical parameters and the abundance of  An. funestus group larvae (iii)  assess which habitat characteristics contributed to the abundance of predators and An. funestus group larvae and (iv) assess the impact of each predator family on the abundance of An. funestus group larvae. All variables (i-iv) were assessed individually and later combined in the final model.

Due to a large number of dips with zero larvae the negative binomial with zero inflated models were used. In all models, the study villages in which the aquatic habitats were identified and habitat ID were used as random terms to capture unexplained variations between villages and habitats. The best fitting models were selected using Akaike Information Criterion (AIC) and results presented as risk ratios (RR) at 95% CI and statistical significance was considered when the P-value < 0.05.

Ethical considerations

Research proposal was presented to the Nelson Mandela Institute of Science and Technology and approval for this study was obtained from the institutional review board of Ifakara Health Institute (Ref: IHI/IRB/No: 13-2022) and from the Medical Research Coordinating Committee (MRCC) at the National Institutes of Medical Research (NIMR) (Ref: NIMR/HQ/R.8a/Vol. IX/3353).  The consent of publication this manuscript was obtained from the National Institute for Medical Research (NIMR) (Ref. No: NIMR/HQ/P.12 VOL XXXV/61). Written permission to conduct study was obtained from local leaders in each village whereby the purpose, procedure and benefits of the study were clearly explained. Verbal and written informed consents were obtained from community members who assisted to sample aquatic predators.

**研究背景**：幼虫捕食者对疟疾媒介种群的调控作用目前仍相对不明。本研究旨在探明与富氏按蚊（Anopheles funestus）组幼虫共存的常见捕食类群，并评估坦桑尼亚东南部农村地区影响其丰度的相关因素。

**研究方法**：在坦桑尼亚南部选定村庄的已确认水生滋生地中，同时采用标准采样勺（350ml）或10L水桶采集蚊幼虫与捕食者。捕食类群通过标准分类检索表进行鉴定。对所有阳性滋生地进行地理定位，并记录其物理特征。同时测定并记录水体的理化参数，包括溶解氧（dissolved oxygen, DO）、pH值、电导率（electrical conductivity, EC）、总溶解固体（total dissolved solids, TDS）以及水温。

**研究结果**：本研究共对9个村庄的85处已确认的富氏按蚊滋生地开展幼虫与潜在捕食者采样，累计采集到8295头捕食类个体。其中常见类群包括：蟌科（Coenagrionidae）57.7%（n=4785）、划蝽科（Corixidae）12.8%（n=1060）、仰泳蝽科（Notonectidae）9.9%（n=822）、蜓科（Aeshnidae）4.9%（n=405）、两栖类（Amphibian）4.5%（n=370）以及龙虱科（Dytiscidae）3.8%（n=313）。
本次采样共采集到5260头蚊幼虫，其中富氏按蚊组占60.3%（n=3170）、库蚊属（Culex spp.）占24.3%（n=1279）、冈比亚按蚊复合组（An. gambiae s.l.）占8.3%（n=438），其余按蚊类群占7.1%（n=373）。永久性且面积大于100㎡的水生滋生地与富氏按蚊组幼虫丰度（P<0.05）及捕食者丰度（P<0.05）均呈显著正相关；带有沉水植被的滋生地则与富氏按蚊组幼虫丰度呈显著负相关（P<0.05）。仅溶解氧（DO）水平对富氏按蚊组幼虫丰度存在显著正向影响（P<0.05），而捕食者丰度未受所有测定的水体理化参数的影响。

**结论**：富氏按蚊组幼虫的水生滋生地中存在6类常见潜在捕食类群。后续仍需开展更多研究，以明确不同捕食类群对蚊幼虫密度及成蚊适合度性状的调控效果。可基于蚊与捕食者的互作关系开发干预策略，以阻断富氏按蚊的传播潜能并降低其种群存活率。

### 研究区域
本研究于2022年3月至5月间，在坦桑尼亚东南部的9个村庄开展横断面调查，其中乌隆加区（Ulanga district）的村庄包括奇库提（Chikuti，-8.6028°, 36.7288°）、姆泽莱齐（Mzelezi，-8.8934°, 36.7343°）、奇隆博拉（Chirombola，-8.93041°, 36.75753°）、埃布尤（Ebuyu，-8.9719°, 36.7608°）、姆瓦亚（Mwaya，-8.91022°, 36.823139°）以及图利扎莫约（Tulizamoyo，-8.35447°, 36.70546°）；基洛姆贝罗区（Kilombero district）的村庄包括伊克万比（Ikwambi，-7.97927°, 36.81630°）、基萨瓦萨瓦（Kisawasawa，-7.89657°, 36.88058°）以及姆波富（Mpofu，-8.17220°, 36.21651°）（图1）。该区域内富氏按蚊介导了超过85%的疟疾传播[17]。当地村民广泛开展水稻种植，为蚊类滋生提供了适宜环境。本研究涉及村庄内富氏按蚊的常见水生滋生地已为学界熟知且此前已完成表征[24]，本次共对9个村庄的85处已知滋生地开展蚊幼虫与潜在捕食者采样。
**图1**：基洛姆贝罗与乌隆加地区地图，标注本次研究涉及的9个村庄

### 蚊幼虫与水生捕食者的采样及形态鉴定
采用标准采样勺（350ml）或10L水桶开展蚊幼虫与捕食者采样，操作流程参照既往研究[8,15,24]。根据滋生地的大小与水深，每个采样点的采样次数为3~20次不等。
既往一项针对同一村庄的研究中，采集的蚊幼虫被运送至伊法卡拉实验室，待其羽化后通过PCR鉴定至物种水平，结果显示53%为纯系富氏按蚊（An. funestus s.s.）、28%为溪岸按蚊（An. rivulorum）、12%为利松按蚊（An. leesoni）[24]，且这3个物种共享同一滋生地。本研究对4龄幼虫样本采用了相似的处理流程，但未进行物种水平鉴定。早期龄期幼虫则参照既往研究的方法，基于其主要形态特征进行鉴定，依据Gilles与Coetzee的分类检索表[25,26]将其划分为富氏按蚊组、冈比亚按蚊复合组（An. gambiae s.l.）或按蚊属未知种（Anopheles sp.）；库蚊类仅鉴定至属水平。捕食类群则通过Stroud水研究中心[27]及Gerber与Gabriel[28]的分类检索表，完成科水平的形态鉴定。对每次采样勺/水桶采集到的蚊幼虫与捕食者个体进行计数并记录。此外，采用手持GPS设备（Garmin eTrex 20x手持GPS接收机）在采样点记录滋生地的地理位置。

### 水生滋生地表征
仅针对富氏按蚊组幼虫阳性的水生滋生地开展蚊幼虫与捕食者采样。记录其整体物理特征，并采用便携式水质检测仪（ZJ practical 4合1水质测试仪）测定水体pH值、水温、电导率（EC）及总溶解固体（TDS）；采用Trans Instruments溶解氧仪（HD3030）按照标准操作流程测定溶解氧（DO）。将滋生地划分为以下类型：沼泽、溪流、河流、稻田、溪塘、地面积水塘、沟渠、泉塘、水坑、蹄印坑、人工井、砖坑或沙坑。水体颜色分为清澈（透明无异味）与浑浊（不透明、混浊或带有油膜）两类。
水源分为雨水与非雨水两类。滋生地中的藻类数量分为无、中等或丰富三类；藻类类型分为丝状藻、绿藻、蓝藻或褐藻。水体流动性分为静止、缓流或急流三类。滋生地周边的土地利用类型分为灌丛、牧草地或耕地。滋生地的遮阴程度分为无遮阴、部分遮阴或重度遮阴三类。采用卷尺测量滋生地面积，并划分为小于100㎡与大于100㎡两类。植被数量与植被类型分别分为（无、中等或丰富）以及（挺水植物或沉水植物）两类。存在时长超过3个月的水体被定义为永久性滋生地，其余水体则为临时性滋生地。水深分为小于50cm与大于50cm两类。通过目视估算滋生地至最近住宅的距离，并划分为小于100m与大于100m两类。

### 统计分析
本研究采用开源软件R版本4.2.1[29]开展数据分析。基于glmmTMB包[30]实现的、采用模板模型构建器（template model builder, TMB）的零膨胀负二项分布广义线性混合模型（Generalised linear mixed effects models, GLMM），用于完成以下4项分析：(i) 评估水体理化参数与水生捕食者丰度之间的关联；(ii) 评估水体理化参数与富氏按蚊组幼虫丰度之间的关联；(iii) 明确影响捕食者与富氏按蚊组幼虫丰度的滋生地特征；(iv) 评估各捕食类群对富氏按蚊组幼虫丰度的影响。所有分析变量先单独评估，再整合至最终模型中。
由于多数采样点未采集到蚊幼虫，因此采用零膨胀负二项分布模型。所有模型中，将采样所在村庄及滋生地ID作为随机效应，以捕捉村庄与滋生地间未解释的变异。采用赤池信息准则（Akaike Information Criterion, AIC）筛选最优拟合模型，结果以95%置信区间（95% CI）的风险比（risk ratio, RR）呈现，以P<0.05作为统计学显著性阈值。

### 伦理考量
本研究方案已提交至纳尔逊·曼德拉科学技术学院，并获得伊法卡拉健康研究所伦理审查委员会（批准号：IHI/IRB/No:13-2022）及坦桑尼亚国家医学研究所（NIMR）医学研究协调委员会（批准号：NIMR/HQ/R.8a/Vol.IX/3353）的批准。本论文的发表许可已获得国家医学研究所（NIMR）批准（批准号：NIMR/HQ/P.12 VOL XXXV/61）。已获得各村庄当地领导人的书面研究许可，并向其清晰说明本研究的目的、流程与获益。协助开展水生捕食者采样的社区成员均已获得口头及书面知情同意。