key: cord-0991405-7qlk8m69
authors: Alizon, S.; Haim-Boukobza, S.; Foulongne, V.; Verdurme, L.; Trombert-Paolantoni, S.; Lecorche, E.; Roquebert, B.; Sofonea, M. T.
title: Rapid spread of the SARS-CoV-2 δ variant in the area of Paris (France) in June 2021
date: 2021-06-20
journal: nan
DOI: 10.1101/2021.06.16.21259052
sha: 9aef58f5d7e1b24c97ba3749ed76852932c01868
doc_id: 991405
cord_uid: 7qlk8m69

Analysing 5,061 variant-specific tests performed on SARS-CoV-2 positive samples collected in France between 31 May and 8 June 2021 reveals a rapid growth of the {delta} variant in the Ile-de-France region. The next weeks will prove decisive but the magnitude of the estimated transmission advantage (with a 95% confidence interval between 67 and 120%) could represent a major challenge for public health authorities.

The evolution of SARS-CoV-2 variants has caused epidemic rebounds in many countries, such as the one in May 2021 in the United Kingdom (UK) caused by the δ first detected in India [1] . According to a report from Public Health England [2] , this variant has a transmission advantage of approximately 66% (95% confidence interval (CI): [28,113]%) over the α variant based on within-household transmission and is slightly more prone to immune evasion [3] . As shown in 2021 with the emergence of the α variant, there is a two-month shift between the epidemic situation in France and the UK [4] . Therefore, the time is ripe to investigate the potential early spread of the δ variant to devise appropriate public health responses.

We analysed 

Using a multinomial regression model as in [5] , we found that samples bearing the L452R mutation and neither the E484K nor the E484Q mutation, i.e. consistent with the variant δ genotype, tended to increase in the Île-de-France region compared to samples without any of the three mutations, i.e. consistent with the variant α genotype (Table 1) .

Furthermore, samples collected outside hospital settings are less likely to orginate from β, γ, or η variants than from α variant (set as reference). Finally, in the Hauts-de-France region, we found an increasing trend of β, γ, or η variants compared to the α variant (set as reference). In most regions, we found a temporal increase of 'other' variants compared to the α variant (set as reference). This is more difficult to analyse since a part of these latest tests are uninterpretable (16,4% of the total). Finally, we find a temporal increase of the δ variant compared to the α variant in the regions that were less densely sampled. 

To further quantify the temporal trends, we performed a statistical analysis using a logistic growth model following previous studies [6, 4] . We found that in Île-de-France the transmission advantage of the positive samples bearing the L452R mutation and neither the E484K nor the E484Q mutations (i.e. potential δ variants) compared to all the other samples is 92% (95% CI: [67,120]%) ( Figure 1A) . In Hauts-de-France, we estimated the transmission advantage of the positive samples bearing the E484K mutation and not the E484Q mutations (i.e. potential β, γ, or η variants) compared to all other samples was 56% (95% CI: [42,71.9]%).

The analysis of the first tests designed to screen SARS-CoV-2 mutations bore by the δ variant collected between 31 May and 8 June 2021 indicate that this variant might already be spreading rapidly in the area of Paris, and potentially in other regions of France. According to our estimates, by mid-June 2021 its frequency in the Île-de-France region could already be above 10%.

There are several potential biases to this analysis, which we attempted to correct. First, since our dataset gathers results from local laboratories, there can be delays in data collection. To address this, we collected the data available on 16 June 2021, which included tests results as recent as June 14, and only analysed the data up to June 8. Note that analysing all the samples yielded qualitatively similar results except for an increase in the δ variant in the Hauts-de-France region according to the multinomial model.

Another potential issue has to do with the specificity of the test used. However, the frequency of δ variant we estimate, which was in the order of 4% of the positive tests on June 8, is consistent with the outcome of national sequencing studies (0.2% on 11 May 2021 [7] ) although with a short doubling time (approximately 8 days). Furthermore, our estimate is conservative since we treat all the samples that are uninterpretable or where the signal is weak as non-δ variant samples.

Regarding the sampling scheme, the French authorities have announced a strict monitoring of the spread of the δ variant, which means that contact tracing could be more intense. This could, therefore, artificially enrich the data in δ variant positive test. We do not have access to the details regarding the context of the test. However, we do know that some tests were performed in a hospital setting, mainly for hospitalised patients. If there was a strong bias generated by better contact tracing, we would expect samples from hospitals to differ because it would likely be less important here (aside from nosocomial transmission, the contacts of a hospitalised person are likely not to be hospitalised). We find no effect of the hospitalisation status in the multinomial model and running the analyses on the hospitalised samples only yielded a similar trend (although with a large 95% CI: 81% [19,166]%).

Finally, from a methodlogical point of view, our estimation of the transmission advantage assumes that the transmission fitnesses of the different lineages to not vary over time. This is unlikely to be the case with the ongoing vaccination campaign and the variations in natural immunity. However, the short time period considered for this analysis justifies this assumption.

The estimated transmission advantage in the Île-de-France appears to be higher than that estimated in the UK [2] . One possibility could be due to specific contact-tracing. Another could be due to the vaccination coverage in France compared to the UK [8] . Indeed, early evidence suggests that the δ variant could be evading immunity more than the α variant [3] .

These results have major public health implications because, given the current estimates for the French epidemic reproduction number (R t ≈ 0.8), they imply that variant δ could be spreading rapidly (with R δ t ≥ 1.2). In absence of specific interventions and based on the current vaccine rollout, a model tailored to the French epidemic [9] suggests that this could cause another epidemic wave starting in August 2021. The magnitude and exact timing of this wave would depend on the exact transmission advantage of the variant. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted June 20, 2021. ; https://doi.org/10.1101/2021.06.16.21259052 doi: medRxiv preprint

World Health Organisation. Tracking SARS-CoV-2 variants

SARS-CoV-2 variants of concern and variants under investigation -Technical briefing 15; 2021. 15

Effectiveness of COVID-19 vaccines against the B.1.617.2 variant. medRxiv

Detection of Rapid SARS-CoV-2 Variant Spread

The SARS-CoV-2 B.1.351 lineage (VOC β) is outgrowing the B.1.1.7 lineage (VOC α) in some French regions in

On measuring selection in experimental evolution

Quelle est l'évolution moléculaire des virus SARS-CoV-2 circulant sur le territoire ? Résultats de l'enquête Flash#9

Coronavirus pandemic (COVID-19)

Memory is key in capturing COVID-19 epidemiological dynamics

We thank Gilles Roullin, Eric Hedbaut and Virginie Dubois from CERBA for their technical assistance, and the ETE team from CNRS, IRD, and University of Montpellier for discussion.

This study has been approved by the Institutional Review Board of the CHU of Montpellier and is registered at ClinicalTrials.gov with identifier NCT04738331.

We analyse the qualitative outcome of the SARS-CoV-2 VirSNiP E484Q/K and L452R assay from Tib Mol Biol performed on samples that tested positive for SARS-CoV-2 in partner laboratories, mainly using partly the PerkinElmer SARS-CoV-2 Real-time RT-PCR Assay (Perkin Elmer).The VirSNiP test targets three mutations specifically (E484K, E484Q, and L462R). Based on the genomics of the different lineages and their prevalence in France, the test results are interpreted as follows:• E484K-/E484Q-/L462R-: potential infection by the α variant • E484K-/E484Q-/L462R+: potential infection by the δ variant • E484K+/E484Q-/L462R-: potential infection by the β, γ, or η variant All the other combinations were rare and grouped in the "other" category (see below for the details).The data was extracted on June 16, 2021 and the most recent test results were from June 14, 2021. However, to correct for potential delays in the recording of the results from local partners, we did not include any tests performed after June 8, 2021. We also only included from individuals from 5 to 80 years old, and included at most one sample per individual.

The characteristics of the data are shown in Table S1 .Among the 'other' test results, the majority corresponded to uninterpretable test results (751/1394, i.e. 54%) and a large fraction to tests without the L452R mutant and an ambiguity for the E484 mutation (524/1394, i.e. 38%), which could be indicative of an infected by a β, γ, or η variant with a low virus load. Note that this assumption is conservative to avoid overestimating the spread of the δ variant. In fact, some of the tests in the 'other' category (74/1394, i.e. 5%) find a clear signal for the L452R mutation with an ambiguity regarding the E484 mutation and could, therefore, be associated with this variant.