Introduction

The global emergence of epidemic arboviral diseases presents a public health threat that needs urgent management 1,2. Among these diseases, the geographical range and intensity of outbreaks of dengue fever have expanded most dramatically 3. Transmitted primarily by Aedes aegypti and to a lesser extent Aedes albopictus mosquitoes, the disease is estimated to affect 390 million people worldwide 4. This has resulted in considerable socio-economic burdens 5 which are expected to increase under climate change, growing human populations and increased urbanisation 6. As such, vector control measures need to be intensified to reduce dengue transmission 7.

The sterile insect technique (SIT) and incompatible insect technique (IIT) have gained traction in recent years for suppressing wild Aedes populations 8,9,10,11,12. The SIT relies on the release of males sterilised through irradiation. Meanwhile, the IIT involves the release of male mosquitoes infected with the endosymbiont bacteria Wolbachia and exploits the expression of cytoplasmic incompatibility between Wolbachia-infected males and females lacking Wolbachia or infected with a different Wolbachia strain. In both techniques, mating between released males and wild-type females in the field results in non-viable eggs. To date, trials in several countries have explored these strategies and successfully demonstrated suppression of wild Aedes populations 9,10,11,12.

In tropical and highly urbanised Singapore, Ae. aegypti is the primary vector of dengue while Ae. albopictus is a secondary vector 13. Singapore’s resident population has low herd immunity to the four circulating dengue serotypes, resulting in a cyclical pattern of dengue outbreaks in the country 14. Currently, environmental management, including premise inspections and source reduction efforts, forms the basis of Singapore’s vector and dengue control programme. While conducted year-round, these activities are enhanced ahead of the typical dengue season to reduce disease transmission 14. To complement this, the National Environment Agency (NEA) is assessing the use of Wolbachia-infected Ae. aegypti males to suppress wild Ae. aegypti in trial areas which include high-rise residential apartment blocks. To reduce the risk of introgression of the Wolbachia into the field Ae. aegypti population, released mosquitoes are additionally irradiated to render any residual females infertile 15. Results from these trials are promising, showing decreases in both the Ae. aegypti population and dengue incidence 12,16,17,18.

Despite the promising results, the potential effect of Ae. aegypti suppression on Ae. albopictus populations in the urban tropics is not well understood. Globally, both species can occur in sympatry and compete 19, with Ae. albopictus displacing Ae. aegypti in some instances where both species are invasive 20,21,22. It has been hypothesised that when Ae. aegypti populations in urban areas are suppressed, ecological niches originally occupied by Ae. aegypti may become available for Ae. albopictus 23, which may then emerge as a dominant disease vector. This is highly pertinent as the efficiency of Ae. albopictus in transmitting chikungunya is well documented, including in Singapore 24, and its potential to cause large dengue outbreaks has been demonstrated in Guangzhou 25.

There is little published information on this topic although a study in West Panama found no evidence for niche replacement by Ae. albopictus following Ae. aegypti suppression in a locality where both species were invasive 26. Singapore’s Wolbachia trial, with more than 90% suppression of wild type Ae. aegypti population in study sites 12, and the national Gravitrap surveillance system comprising more than 70,000 traps in residential premises, offer an opportunity to study the dynamics of the two species in a tropical city where Ae. albopictus is native.

This study aims to assess the potential impact of Ae. aegypti suppression on Ae. albopictus in high-rise public residential apartment blocks set in a vegetated, urbanised city. Using nationally representative and longitudinal surveillance data, we compared (1) the population trends of Ae. albopictus between sites with and without Ae. aegypti suppression and (2) the vertical distribution of Ae. albopictus in high-rise residential apartment blocks before and after Ae. aegypti suppression. Field Aedes mosquitoes were also collected and screened for dengue viruses to compare the relative dengue positive rates in these two species which may impact dengue transmission.

Results

Aedes albopictus population trends

Analysis of the population trends and vertical distribution (see below) is based on 36 datapoints (14 pre-intervention and 22 post-intervention) collected monthly from approximately 4,500 Gravitraps in four treatment sites and 30,000 Gravitraps in 11 control sites. The Gravitrap Ae. albopictus index (GIalbo) was then generated by normalising the number of adult female Ae. albopictus collected per month by the number of functional Gravitraps.

There was a significant increase (p < 0.001) in the GIalbo in all study sites (both treatment and control) where the average monthly GIalbo had approximately tripled in 2021–2022 (GIalbo ≈ 0.10) as compared to 2019–2020 (GIalbo ≈ 0.03) (Table 1a, Fig. 1, Supplementary Material 1– 3). The before-after-control-impact (BACI) analysis also indicated that the GIalbo of treatment sites were, on average, significantly higher by 39.3% (p < 0.001, 95% CI 23.1% – 55.4%) compared to the control sites after the start of releases (Table 1a). However, analyses at the site-level indicate inconsistency in how the abundance of Ae. albopictus changed across the treatment sites. GIalbo in Choa Chu Kang and Yishun were significantly higher by 0.03 (p < 0.001, 95% CI 0.02 – 0.05) and 0.04 (p < 0.001, 95% CI 0.03 – 0.06), respectively relative to the controls. Meanwhile, a small and non-significant increase in GIalbo of 0.01 (p = 0.09, 95% CI 0 – 0.03) was observed in Bukit Batok. In contrast, there was a non-significant decrease in GIalbo in Tampines of 0.01 compared to the control sites (p = 0.14, 95% CI -0.03 – 0.0) (Fig. 2). This was despite a large reduction in the Ae. aegypti population in all treatment sites (Fig. 1, Supplementary Material 1, 4).

Table 1 Output of the linear mixed models used to compare the Aedes albopictus (a) population abundance and (b) vertical distribution across study locations and periods.
Full size table
Fig. 1
figure 1

Monthly Gravitrap Aedes albopictus index for each treatment site and the aggregated trend for all control sites (bold lines) in relation to the aggregated Gravitrap Ae. aegypti index for the four treatment sites (dash-dotted line). The vertical dashed line indicates the start of male Wolbachia-infected Ae. aegypti releases in selected sites in Yishun and Tampines in March 2020, while the vertical dotted line indicates the start of releases at selected sites in Bukit Batok and Choa Chu Kang in June 2020.

Full size image
Fig. 2
figure 2

BACI estimates of the average difference in the monthly Gravitrap Aedes albopictus index (GIalbo) between each treatment site to the control sites after the start of male Wolbachia-infected Ae. aegypti releases. Horizontal lines indicate 95% confidence intervals. Intervals above zero indicate significantly higher site-level GIalbo than the control site after the start of releases. Treatment sites: BB – Bukit Batok; CCK – Choa Chu Kang; TP – Tampines; YS – Yishun. Control sites: AMK – Ang Mo Kio; BM – Bukit Merah, BP – Bukit Panjang; HG – Hougang; JW – Jurong West; KL – Kallang; PR – Pasir Ris; QT – Queenstown; SK – Sengkang; TPY – Toa Payoh; WL – Woodlands.

Full size image

Vertical distribution of Ae. albopictus

In contrast to the population trends, BACI analysis indicated no significant change in the proportion of Ae. albopictus collected at low floors between treatment and control sites after the start of releases (p = 0.83) (Table 1b). Across all study sites and both periods, most individuals were consistently collected at low floors (median proportion > 0.80) (Fig. 3, Supplementary Material 5). However, two treatment sites (Choa Chu Kang and Yishun) recorded slightly higher proportions of Ae. albopictus collected at lower floors in the period after releases (median proportion increased from 0.81 to 0.87 and from 0.87 to 0.93 in Choa Chu Kang and Yishun, respectively). Overall, the distribution of Ae. albopictus across floor categories has remained largely unchanged.

Fig. 3
figure 3

Boxplots comparing the monthly proportion of Aedes albopictus collected in the low floor category in high-rise residential apartment blocks for both before and after periods across the study sites. The site labels are provided in Fig. 2.

Full size image

Dengue prevalence in wild Aedes mosquitoes

From January 2010 to December 2022, a total of 17,850 Ae. albopictus and 47,294 Ae. aegypti females were screened for dengue (Table 2). The number of individuals screened in each year was associated with the intensity of dengue transmission (see Ho et al. 12 for details on yearly transmission). The yearly prevalence of dengue viruses in both Aedes species was low, ranging from 0% (Ae. albopictus only) to 5.73% (for Ae. aegypti in 2010). Overall, dengue prevalence in field Ae. albopictus (n = 12, yearly average = 0.09% ± 0.04 SE) was approximately 18 times lower than that of field Ae. aegypti (n = 467, yearly average = 1.57% ± 0.42 SE). In addition, no dengue-positive Ae. albopictus were detected in six out of the 13 years of screening, while dengue was detected yearly in Ae. aegypti.

Table 2 Number of field Aedes albopictus and Ae. aegypti individuals screened for dengue using NS1 antigen rapid test kits or PCR, as well as the percentage and number of dengue positive individuals, from January 2010 to December 2022.
Full size table

Discussion

We found that, across the study locations, the population trends of Ae. albopictus varied but had not consistently increased compared to sites without Ae. aegypti suppression (Fig. 1, 2, Supplementary Material 2). We also observed that the proportion of Ae. albopictus collected across floor categories did not differ significantly between periods prior to and after Ae. aegypti suppression (Fig. 3). Thus, the suppression of Ae. aegypti using a hybrid SIT-IIT approach appears to have limited impact on the abundance and vertical distribution of Ae. albopictus in high-rise residential neighbourhoods. These findings corroborate with those described by Bansal et al. and Gorman et al. 12,26.

The habitat preferences of the two species likely explain these findings. The native Ae. albopictus is typically associated with outdoor settings, and thus lower floors of apartment blocks which are closer to managed vegetations on the ground floor. In contrast, Ae. aegypti tends to prefer urbanised areas where it breeds in domestic settings 27,28. Additionally, Ae. albopictus is suggested to be less anthropophilic than Ae. aegypti and hence less likely to be in close proximity to humans 29,30. Though the suppression of Ae. aegypti in high-rise residential apartment blocks is expected to increase the “ecological space” (such as breeding sites) available for Ae. albopictus, these environments may not be favoured by Ae. albopictus. Overall, these results corroborate findings of Ae. albopictus habitat use from previous work 30,31 and provide little evidence of ecological niche shifts or expansions by the non-targeted species in the local context 32.

The variable outcome of Ae. albopictus abundance in the different treatment sites following successful suppression of Ae. aegypti suggests that any site-specific increase in the Ae. albopictus population is not directly due to a lowered Ae. aegypti population. The increase in Ae. albopictus abundance in Choa Chu Kang and Yishun may thus be the result of local factors that warrant further investigation. First, differences in vegetation cover can affect the abundance of Ae. albopictus in urban settings 31. However, a related study found no such substantial difference across the four sites 12, and this remains an unlikely contributing factor in this instance. Meanwhile, mosquito control in Singapore is very much driven by the need to control transmission of arboviruses, such as dengue and Zika 13. The successful suppression of Ae. aegypti significantly decreased the risk of arboviral transmission in treatment sites, and we note that the two sites with increased Ae. albopictus abundance also recorded greater reduction in dengue incidence rates 17. Therefore, it may be that the reduced dengue transmission contributed to a decline in vigilance among stakeholders with potential benefits for breeding by Ae. albopictus 33, which was a risk we identified prior to the start of the field trials 34.

The very low prevalence of dengue virus in both Aedes species concurs with past observations from Singapore 35 and other countries such as Indonesia 36. Moreover, the dengue prevalence rate in field Ae. albopictus is substantially lower (approximately 18 times) than that for Ae. aegypti, a trend which mirrors that reported by Chung & Pang in 2002 35. This is consistent with other findings indicating that Ae. albopictus, compared to Ae. aegypti, is a less competent vector for the virus as well as being less anthropophagic, and that it plays a relatively minor role in dengue transmission in most urban residential environments 29,37,38,39.

The analysis also revealed a significant increase in Ae. albopictus abundance islandwide. Several factors could have contributed to this pattern. While warmer weather and ageing infrastructure may be contributing factors 31, the period of increase (mid-2020) also coincided with a period of intensive greening (e.g. promotion of skyrise greenery) and promotion of urban gardening 40. As Ae. albopictus is associated with greenery and public spaces 30,31, this may have increased the availability of habitats and breeding sites favourable for Ae. albopictus.

Though Ae. albopictus is not a major dengue vector, any increase in its abundance is a cause for concern. First, a larger vector population translates directly to a larger vectorial capacity 41. While Ong et al. did not find an increased odds of dengue transmission with increasing Ae. albopictus population in Singapore 39, the species is capable of driving transmission as seen in countries where the species predominates 25. Next, changes in the Ae. albopictus population structure 42 or adaptation of the dengue virus to Ae. albopictus 37 may render the species a more competent vector for dengue that then increases disease transmission. Finally, beyond dengue, Ae. albopictus is the primary vector for chikungunya 24 and a competent vector for zika in Singapore 43,44. As such, a high Ae. albopictus population can pose considerable public health risks given its opportunistic feeding behaviour 45. These potential risks highlight the importance of routine surveillance to detect population changes, and of adopting an integrated vector management approach to holistically manage vector population. Ongoing public engagement remains critical to ensure sustained community vigilance 13,33.

Our study yields valuable insights, while also revealing areas which warrant further examination. First, our findings were obtained from data collected about two years post-initiation of male Wolbachia-infected Ae. aegypti releases. While this observation period surpasses that of a similar study by Gorman et al. 26, it remains possible that more time may be needed to detect potential niche replacement or other ecological changes. The present findings thus serve as an important baseline, and we are monitoring both Aedes species closely to elucidate long-term ecological dynamics. Next, Singapore’s context – urban landscape dominated by high-rise residential apartments – may not be representative of other regions where both Aedes species coexist. Thus, the study finding’s broad applicability may be limited, and efforts should be directed to assess the impact of similar population suppression programmes on non-target species in varying landscapes.

In summary, we found variable change in the abundance of Ae. albopictus in high-rise residential estates in Singapore following an SIT-IIT based Ae. aegypti suppression programme, with increases in some sites and minimal changes in others. In addition, we found limited change in the vertical distribution of Ae. albopictus within these estates, and very low prevalence of dengue virus in field Ae. albopictus. As such, the suppression of Ae. aegypti is unlikely to directly result in Ae. albopictus acting as a major dengue vector in Singapore. However, a decline in vigilance among stakeholders, environmental changes such as climate change and increases in managed vegetation as well as viral evolution may provide Ae. albopictus the opportunity to emerge as a dominant vector in Singapore. Therefore, continued surveillance and environmental management remain essential to suppress Ae. albopictus. Close monitoring of the vector population’s competence for dengue and other viruses is also critical, and novel approaches should be explored to manage Ae. albopictus.

Methods

Entomological data

As part of the island-wide Aedes surveillance programme, more than 70,000 Gravitraps have been deployed in high-rise public residential apartment blocks and landed homes in Singapore. These Gravitraps are sticky ovitraps that attract and capture gravid female Aedes mosquitoes 46. Within each apartment block, Gravitraps are placed along public corridors and distributed evenly vertically across floors to achieve a deployment ratio of approximately one trap per 20 households. The Gravitraps were monitored fortnightly and all mosquitoes collected were sexed and identified to species using morphological and molecular methods 46. Traps that were dry, missing, overturned, or had their sticky lining removed were considered non-functional and excluded from analysis.

Aedes aegypti and Ae. albopictus abundance was determined in the same manner as Ong et al. 47. In brief, residential apartment blocks were grouped into individual “sectors”, each comprising five to 30 blocks. Data from the Gravitraps in each sector were aggregated at the month level to reduce the noise of weeks with no or very few mosquitoes 31. The Ae. aegypti and Ae. albopictus abundances in each sector were then calculated using the monthly Gravitrap Ae. aegypti Index (GIaeg) and Gravitrap Ae. albopictus Index (GIalbo), defined as the mean number of adult female Ae. aegypti and Ae. albopictus collected per functional Gravitrap per month, respectively.

Site selection

The NEA currently conducts SIT-IIT suppression of Ae. aegypti in several high-rise public housing towns, of which four (Bukit Batok, Choa Chu Kang, Tampines and Yishun) have received mosquito releases for at least two years as of December 2022 (“treatment sites” henceforth) (Fig. 4) 48. Each site is comprised of 133 to 731 high-rise public residential apartment blocks, which are typically 12 floors high. In these sites, Wolbachia-infected Ae. aegypti males were distributed twice a week at a rate of up to six males per resident per week, and suppression of Ae. aegypti populations of around 60% and 90% was observed within three and 12 months of releases, respectively 12. In these four sites, male Wolbachia-infected Ae. aegypti releases were introduced in phases 12. To ensure data for each site sufficiently captures the Ae. albopictus abundance before and after the start of releases, only sectors with releases starting from March 2020 to February 2021 were selected for analyses (Fig. 4, Table 3). To estimate the Ae. albopictus abundance in each site, the site-level GIalbo was obtained by aggregating Gravitrap data across the selected sectors within each site. As a control, sectors in high-rise public housing towns without male Wolbachia-infected Ae. aegypti releases and with similar dengue risk profiles as treatment sites were identified using the method described in Ong et al. 49. In brief, residential areas were grouped into four risk groups using a Random Forest machine learning approach based on data comprising past dengue exposure, human population density, vector population and environmental indices. To ensure that data from control sites was representative, only towns with at least 20 selected sectors were included as control sites (number of sectors per town which meet the above criteria ranged from three to 42). Data from these sectors were then aggregated at the site-level to obtain the respective GIalbo for each control site (Fig. 4, Table 3).

Fig. 4
figure 4

Map of Singapore with the four sites that had received male Wolbachia-infected Ae. aegypti releases for at least two years demarcated in red. Within each treatment site, sectors selected for analysis are indicated as polygons outlined in black. Sectors without releases that were selected as controls are indicated as blue polygons outlined in black, with varying shades of blue representing different control sites. The map was generated using QGIS (version 3.16) 50 with basemaps obtained from OneMap Singapore 51. The site label is provided in Fig. 2.

Full size image
Table 3 Number of sectors per treatment and control site, and the delineation of “before” and “after” periods for each site for the before-after-control-impact analysis.
Full size table

Statistical analysis

To investigate the impact of Ae. aegypti suppression on Ae. albopictus abundance, a before-after-control-impact (BACI) statistical design was used 52. The Ae. aegypti SIT-IIT suppression programme was introduced in phases from March 2020 to February 2021. To align the period across all sectors, the “before” period was designated as months when surveillance first started (January/February 2019) until February 2020, while months from March 2021 until December 2022 were assigned to the “after” period (Table 3).

Using a linear mixed model, the Ae. albopictus abundance was modelled as a function of the treatment group (control or treatment), period (before or after) and the interaction between group and period. In addition, site and month of sampling were modelled as random effects to account for repeated measures of GIalbo per site over time. Using the model

$$GI_{albo} \sim Group + Period + Group:Period + \left( {1|Site} \right) + (1|Month)$$

a statistically significant interaction between group and period therefore points to a change in Ae. albopictus abundance following Ae. aegypti suppression.

Similarly, a BACI approach was used to investigate whether Ae. aegypti suppression affected the vertical distribution of Ae. albopictus in high-rise residential apartment blocks. Gravitrap data for treatment and control sites were stratified according to floor category (low: level 1 to 4; middle/high: level 5 and higher) and aggregated by month. The monthly proportion of Ae. albopictus collected per Gravitrap for each floor category was calculated and assigned to “before” and “after” periods as described in Table 3. The proportion of Ae. albopictus collected at low floors was modelled using a generalised linear mixed model with a binomial distribution:

$$Proportion_{low } \sim Group + Period + Group:Period + \left( {1|Site} \right) + (1|Month)$$

and a statistically significant interaction between group and period highlights a change in the proportion of Ae. albopictus collected at lower levels after Ae. aegypti suppression.

Notably, significant interactions in the models only point to significant differences between all treatment sites and all control sites. However, it is of interest to assess the changes for each treatment site. To additionally evaluate this at the site-level, data from the treatment sites were individually compared to the control sites using a generalised linear model:

$$GI_{albo} \sim Site + Period + Site:Period\tt, and\ \it Proportion_{low } \sim Site + Period + Site:Period\\ $$

and statistically significant interactions between site and period would indicate how the abundance and vertical distribution of Ae. albopictus in each treatment site compare in relation to the control sites.

Fitted models were checked for overdispersion to ensure robustness of the analysis. All statistical analyses were performed using R (version 4.3.3) 53, and statistical significance was assessed at the 5% level.

Prevalence of dengue viruses in Aedes mosquitoes

As Gravitraps are designed to attract gravid female Aedes mosquitoes 46, captured females which test dengue-positive are likely to be infected or infectious instead of freshly fed with an infected bloodmeal 54,55. As such, we used mosquitoes collected from Gravitraps to estimate and compare the rate of dengue prevalence in field Aedes mosquitoes. As part of routine monitoring in dengue clusters and ad hoc surveillance, Aedes mosquitoes collected from Gravitraps were screened individually or in pools of up to 10 mosquitoes for dengue viruses using dengue NS1 rapid antigen test kits or real-time polymerase chain reaction (PCR) 46. Data for these mosquitoes from January 2010 to December 2022 were then extracted from NEA’s database and summarised.