Environmental drivers of SARS-CoV-2 lineage B.1.1.7 transmission intensity

Introduction

A year into the COVID-19 pandemic, quantifying the factors driving SARS-CoV-2 transmission is still key for the optimal implementation of control strategies. SARS-CoV-2 transmission is influenced by local environmental conditions (Poirier et al., 2020), and we have previously shown that colder temperatures and higher population densities explain spatial and temporal variation in transmission intensity (Smith et al., 2020). A critical insight from previous work has been that human behaviour drives SARS-CoV-2 transmission, and that environment only plays a marginal effect when effective non-pharmaceutical interventions (NPIs) are in place (Poirier et al., 2020, Smith et al., 2020). While several studies have investigated the impact of environmental factors on SARS-CoV-2 transmission during the early stages of the pandemic, numerous new variants have recently emerged (in, for example, the UK; Rambaut et al. 2020, Brazil; Voloch et al. 2020, and South Africa; Tegally et al. 2020) with greater transmissibility than strains present during the early stages of the pandemic (Faria et al., 2021, Volz et al., 2021). Any effect of environment must therefore be re-assessed in the context of a landscape of emerging viral lineages.

The UK SARS-CoV-2 ‘Variant of Concern’ (VOC; also known as lineage B.1.1.7; Rambaut et al., 2020) rapidly spread through the UK population in late 2020, showing greater transmission intensity (as measured by the reproduction number, R) than other circulating strains (Volz et al., 2021). The wealth of data collected for this strain makes the VOC a strong test-case for investigating environmental effects in the context of emerging lineages. To assess the impact of environment on VOC transmission, we extracted weekly R estimates calculated for fine-grained regions of England (Lower Tier Local Authorities— LTLAs; Mishra et al., 2020), and regressed these against VOC frequency, temperature, and population density for each LTLA. We assessed the impact of the environment by investigating the association between weekly R estimates and the predictors (i.e., spatial variation in R), for sequential weeks. To account for the VOC’s emergence in the South-East (which is among the warmest and most densely populated parts of England), and its subsequent spread into colder, less densely populated parts of the country, we incorporated interactions between VOC frequency and each of temperature and population density. Without including this interaction, any effect of temperature or population density on R might be compounded or masked by spatial autocorrelation with VOC frequency. Following Volz et al. (2021), we measured the effects of the weekly predictors on R in the subsequent week, to account for the generation time of SARS-CoV-2.

Results and Discussion

As observed by Volz et al. (2021), we find a strong positive effect of VOC frequency on transmission intensity, particularly during the initial emerging phase when VOC frequency was less consistent among regions (Table 1). While the UK’s national lockdown initially decreased SARS-CoV-2 transmission intensity, as the VOC spread, R increased across regions (Figure 1). We find that the impact of the environment on transmission is mediated by VOC frequency, and it is only at higher VOC frequencies that the effects of temperature and population density are pronounced (Table 2 and Figure 1). Furthermore, we find a greater environmental effect after the UK moved from full lockdown to a (less strict) tiered system, confirming our previous observations that the effect of temperature becomes pronounced when NPIs are relaxed (Smith et al., 2020). Thus, the effects of environmental drivers, and even those of the increased transmissibility of the VOC, are secondary to differences in human behaviour driven by differences in NPIs. As the VOC has spread to dominate the country’s viral population, we now expect environmental factors to become the dominant driver of spatial variation in R, with colder areas facing higher transmission intensities than warmer regions, unless NPIs or the accumulation of immunity (either naturally acquired or vaccine derived) sufficiently reduce transmission.

• Download figure
• Open in new tab

Figure 1: Transmission dynamics during emergence and spread of SARS-CoV-2 lineage B.1.1.7.

A. Median (bold) and interquartile range (box) of the transmission intensity estimates (R) for weeks 43-50 of 2020, with point estimates for each LTLA overlaid. Points are coloured by the frequency of the VOC in positive COVID-19 tests (%), boxplots are coloured by the median VOC frequency across all LTLAs. The first date of each week is given within parentheses. As the VOC emerged, the UK entered a national lockdown, leading to an initial decrease in R across regions. However, as the VOC spread through space to become the dominant lineage, R increased across regions despite the national lockdown. When the UK re-entered a tiered NPI system on the 2nd of December 2020, average R further increased across regions. Panels B and C show the observed (points) and expected (background) effects of temperature and VOC frequency on R during early (week 46) and late (week 50) phases of the variant sweep, respectively (legend shared across both plots).

To investigate the potential role of immunity in our results, we perform a sensitivity analysis and re-run our regression analyses on R estimates corrected for the relative attack rate (AR) estimates. This accounts for the potential slowdown in R due to the accumulation of natural immunity across the population through time. We obtain qualitatively identical results in the AR-corrected analyses, but generally lower correlation (additive decrease in of 16% in weeks 45 and 46) as the VOC emerged, versus in the later stages (average decrease in of only 3%, Supplementary Table S1). Note that the time-frame of our analyses was prior to mass vaccination of the UK public and even in the latest time-point (week 50), median attack rate across regions was less than 10% (Supplementary Figure S1).

We additionally investigated whether the VOC may respond differently to the environment as compared to non-VOC strains. Volz et al. (2021) estimated the transmission intensity of VOC vs non-VOC strains and here we regressed the ratios of VOC R to non-VOC R against temperature, to understand under which environmental conditions the VOC may have an advantage. This was performed at a larger spatial scale than our LTLA analyses (English NHS sustainability and transformation partnerships – STP regions). We found a small, positive correlation between temperature and the R ratio (Supplementary Table S2), i.e., at warmer temperatures the VOC was even more transmissible than non-VOC strains, than in colder conditions. This may be expected, as differences in fitness between strains are likely to be magnified in harsher environments where fitness is reduced overall (Soberon and Peterson, 2005). This effect may have played into the dynamics of the spread of the VOC in the South East of England by enhancing its spread even more in regions where previous strains’ transmission intensities were lower.

Whereas, previously, a central question has been whether winter conditions would enhance SARS-CoV-2 transmission, in the context of these new strains it is important to ask whether warmer summer conditions could slow transmission. Here we have shown that temperature can drive variation in transmission intensity between regions and between variants, but this effect is likely secondary to the overall increased transmission intensity of the new variant. Our results show that variant B.1.1.7 is more transmissible in colder areas. Using the model fitted to the week 50 data, our results suggest that when the variant accounts for 100% of SARS-CoV-2 cases in the population, R increases by approximately 0.12 per °C of temperature decrease. However, we also find that in warmer conditions, the difference in transmissibility between VOC and non-VOC strains is greater. Our results suggest that for each °C of temperature increase, the ratio of VOC to non-VOC R may increase by approximately 0.12-0.22 (Supplementary Table S2).

While we caution against extrapolating from winter to summer conditions, our results highlight that there is no reason to suppose that summer weather alone will slow down the invasion dynamics of B.1.1.7 and significantly reduce the transmission intensity of SARS-CoV-2. Thus, it is imperative to quantify and continue to monitor the impact of interventions (e.g., vaccines and NPIs) to inform policy. In that regard, the speed with which the VOC has spread through the UK, while concerning from a public health perspective, provides the perfect opportunity to parameterise models of its responses. Importantly in the context of new strains, transmission of respiratory viruses in tropical climates (such as Brazil, where a new variant has arisen; Voloch et al., 2020) are often more sensitive to rainfall than temperature (Pica and Bouvier, 2012). Better characterisation of past, and new potential relationships between environment and transmission, as new variants emerge and spread internationally, is a research priority.

Materials and Methods

Data and code availability

All code to replicate these analyses and associated data are available from the following github repository: https://github.com/smithtp/covid19-variant-N501Y