Alpha

The analyses revealed that during the Alpha wave period, there were at least 99 independent introduction events (95% highest posterior density [HPD]: 83–113) from other countries to Spain (Fig. 2a, b). By far the most introductions to Spain during the Alpha period came from France (83, 95% HPD: 70–97). However, most introductions produced fewer than ten sampled descendants (84, 95% HPD: 68–96). Of those, around half were singletons, meaning there was only one sampled descendant (48, 95% HPD: 39–56), and thus most introductions did not produce substantial local transmission. On the other hand, 15 introductions (95% HPD: 11–19) resulted in a sizeable number of descendants (i.e. >10 sampled tip nodes) and these introductions originated almost exclusively in France.

Circular migration flow plots based on the Markov jumps between different countries and Spain (a: Alpha; c: Delta; e: Omicron-BA.1). Mean and number of transitions from those countries to Spain, computed from all the trees from the posterior distribution. The error bars represent the 95% highest posterior density (HPD). The width of the violin plots represents the density of observations, derived from 638 M, 794 M, and 1388 M generations for Alpha, Delta, and Omicron-BA.1, respectively (b: Alpha; d: Delta; f: Omicron-BA.1). Codes ISO 3166-2 AD, IN, UK, and US correspond with Andorra, India, the United Kingdom and the United States of America, respectively.

Delta

In comparison with Alpha, during the Delta period there was an increase in the number of introductions and the number of originating countries (Fig. 2c, d). We detected 202 (95% HPD: 186-221) introductions, with France again being the most important source of viral introductions (117, 95% HPD: 99–133). The number of introductions from the UK increased considerably (38, 95% HPD: 31–44), followed by Germany (14, 95% HPD: 7–20) and the Netherlands (11, 95% HPD: 5–15). There were also introductions from the USA, Italy, Portugal, and Belgium. Again, most introductions did not lead to substantial transmission clusters (187 tip nodes, 95% HPD: 170–205), with most being singletons (127, 95% HPD: 112–142). However, we did not detect introductions from India to Spain (or vice versa), despite it being the geographical origin of the Delta variant (Fig. 2d).

Omicron-BA.1

In the case of the first wave of Omicron (BA.1 and its sublineages) (Fig. 2e, f), there was a decrease in the number of identified introduction events (114, 95% HPD: 100–128) compared to the Delta period, although in both periods many introductions were small (104, 95% HPD: 90–118), primarily singletons (76, 95% HPD: 62–88). France was again the primary source of introductions, with around 27 (95% HPD: 18–35) estimated introductions, but the number of detected introductions from the UK and Germany was quite similar, with 24 (95% HPD: 17–31) and 22 (95% HPD: 14–28) respectively. Italy and the USA were also important sources of Omicron-BA.1 introductions to Spain. In this case, we also detected at least five introductions (95% HPD: 2–9) of SARS-CoV-2 from Portugal to Spain.

Discussion

The introductions of SARS-CoV-2 to Spain for the Alpha, Delta, and Omicron-BA.1 VOCs (epidemic waves three to six) likely reflects the changing landscape of COVID-19 restrictions and containment measures (Fig. 1c), as well as the mobility peaks linked to vacation periods, notably around August (Supplementary Fig. 1).

Contrary to the previous SARS-CoV-2 waves, most of the Alpha, Delta, and Omicron-BA.1 introductions came from France. For example, for the first wave, most of the detected introductions came from Italy, the Netherlands, England, and Austria⁹. This large contribution from France could have been shaped by the combination of a high number of cases and a high connectivity to Spain, making it the country with the largest number of representative sequences in this study, after Spain. However, we also have to take into account that, together with Portugal, which is much smaller and is connected primarily by terrestrial ways to Spain, France was the primary way of entrance to Spain from the rest of Europe, especially in a context where air connections were much more restricted. Peninsular Spain also shares borders with Andorra and the UK (via Gibraltar), but their much smaller population and territory renders their importance at a national level negligible.

This predominance of French introductions was especially noticeable during the Alpha period (third and fourth waves), when, despite the movement restrictions put in place by Spain and France, almost all of the introductions into Spain (>80%) came from France. At the same time, we detected markedly fewer introductions from other countries with high connectivity with Spain. Towards the Alpha period, Spain and France were under stringent containment measures, including curfews and perimeter closures that restricted mobility around their territories. Apart from this, movements across the border were restricted to essential ones, such as cross-border workers and freight transports, or required the presentation of a negative PCR test (or a vaccination certificate, when these became more widely available). Despite these aspects and the recommendations to reduce traveling, the Spanish-French border was not outright closed, so there could still be some mobility, allowing a steady exchange of travelers and, therefore, of viruses. The influx of viruses from France was especially prominent towards the beginning of the Alpha period and declined as time passed, without peaks that could be linked to holiday periods such as the Holy Week. On the other hand, the border with Portugal was closed from the end of January to the beginning of May, except for exceptional reasons, which could help explain why we did not detect almost any introductions from Portugal to Spain. We must also consider that Portugal is much less populated than Spain and France, which also relates to fewer infected people, and its geographical location, not bordering any other country. All these factors combined could help to explain Portugal’s limited role as a source of introductions, at least towards Spain.

Additionally, in the case of the UK, there were several extra restrictions from both sides to travel between Spain and the UK due to the latter being the origin of the Alpha VOC^13,36. While these measures aimed to limit SARS-CoV-2 transmission and the number of introductions, they differed for both countries. While Spain banned UK travelers from entering the country unless they were Spanish nationals or residents, the UK imposed a self-quarantine period upon return, in both cases combined with other restrictions in the national territory. According to our analyses, these measures could have helped limit direct introductions from the UK to Spain.

The peak of Alpha introductions took place towards the beginning of the study period, while during the last weeks, i.e., in summer, as mobility increased (Supplementary Fig. 1A) and restrictions eased, it was already being replaced by Delta. Meanwhile, the sharp increase in introductions to Spain during the Delta wave could be linked to the fact that it started during the summer. This period was characterized by increased international tourism (Supplementary Fig. 1B) and higher national mobility (Supplementary Fig. 2), while the level of restrictions was the lowest until that point. Lifting the outdoor mask mandates and appropriate social distancing, combined with the relaxation derived from the vaccination rollout, may have helped to propel this peak in introductions as more people decided to travel. Also, at the end of September, towards the middle of the Delta period, with around 80% of the Spanish population already vaccinated (Supplementary Fig. 3), restrictions were further relaxed, and several regions entered the so-called “new normality” without specific restrictions.

The increase in tourism and lowered restrictions is likely reflected in the fact that, although France still ranked first in the number of introductions, it only represented around 58% of all introductions, a sharp decrease compared to the Alpha period. On the other hand, lifting the travel restrictions from and to the UK could explain its second position as a virus source. The sources of introductions during this period were much more diverse than during the Alpha wave, which could also be related to the lower level of restrictions across Europe. The increased transmissibility of Delta^37,38, as well as the higher risk of infection by this variant after vaccination³⁹, could also help explain the rapid variant replacement. Despite this, the peak in cases was slightly lower for Delta, the number of deaths was dramatically reduced, and an even higher percentage of the introductions were non-expansive, which means that pre-existing immunity in the population (by previous infections or vaccination), as well as some of the measures in place, may have played a role in restricting Delta spread.

Lastly, the Omicron-BA.1 period or sixth wave, which was associated with lower national mobility than for the Delta period, albeit higher than for Alpha (Supplementary Fig. 2) and not linked either with a significant peak in international tourism (Supplementary Fig. 1C), showed fewer introductions than the Delta wave. Nonetheless, with the lowest stringency of measures in place, despite the mandatory use of the mask outdoors again, the number of detected introductions was still higher than for Alpha. The fact that most of the introductions were non-expansive appears to contrast with the number of cases of this wave, which was the highest of the whole pandemic in Spain. This higher number of cases, on the other hand, might be explained by the increased transmissibility of Omicron-BA.1 compared to the previous VOCs⁴⁰, to a reduced sense of risk from the population, when everyone who wanted had already been vaccinated, and to the so-called “pandemic fatigue”^41,42.

The reasons for the limited number of Omicron-BA.1 introductions in Spain could be multiple. One possible explanation is that the Omicron period, taking only BA.1 into account, is much shorter than the Alpha and Delta period (35, 26, and 16 weeks for Alpha, Delta, and Omicron-BA.1, respectively). Additionally, the proportion of sequenced cases respective to the total cases was the lowest of the three waves (Supplementary Fig. 4), as the sequencing reached peaks similar to Delta while the number of cases skyrocketed. This low proportion of sequenced cases could potentially limit the detection of introductions in the country, as many transmission chains could be undetected or underrepresented.

The sources of Omicron-BA.1 introductions were considerably more diverse than for previous VOCs. Although still in first place, France was only accountable for 24% of the introductions, and a very similar number of introductions was detected from the UK and Germany, in line with a lower level of preventive measures and travel limitations. We should note that even though the connectivity from France to Spain was much higher than that of the rest of the countries, the same number of Omicron-BA.1 sequences from France, Germany, Italy, and the UK were present in the study, as they reached the limit of 500 sequences per country of origin (Supplementary Table 1). The much larger number of initially selected sequences for this variant is related to the peak of cases that Omicron caused in the connected countries, as well as in Spain, and limiting those numbers was necessary to maintain the analysis computationally feasible. Even so, we expect the number of sequences from each country in the study to have a limited influence over the number of introductions as, for example, for Delta we included more UK sequences than French ones. Still, we detected three times as many introductions from France as from the UK. We should then consider that the proportion of German and British travelers that caused an introduction of SARS-CoV-2 in Spain had to be substantially higher and could be related to the different anti-COVID-19 measures put in place by each region, as the regions that receive most of those travelers are not the same. In the case of France, the mobility was higher towards Catalonia (Supplementary Fig. 5A), in the northeast of Spain and bordering France, which imposed strict control measures during the Christmas period and January, including nocturnal curfews, the closure of nightlife and the mandatory presentation of the COVID-19 certificate. On the other hand, the highest mobility from Germany and the UK occurred towards the Canary Islands (Supplementary Fig. 5B, C), which only required presenting the COVID-19 certificate. These differences in mobility trends, coupled with the asymmetric containment measures within Spain, could explain the lower proportion of French introductions compared to the previous waves and possibly highlight the importance and impact of those regional measures.

One trend common to the three VOCs studied was that most introductions produced only a handful of descendants. This could be due to the various prevention and containment measures, which greatly limited the interaction between people without masks for long periods, possibly reducing the odds of a massive transmission event. Apart from this, quarantines after testing positive were mandatory at the beginning and then at least recommended, as well as maintaining social distancing and mask-wearing, which could also help deter local transmission. This situation is also typical of SARS-CoV-2 transmission, as there is significant variability in the number of descendants produced, which means most of the introductions into a territory tend to make a low number of descendants, while some others are responsible for most of the cases^43,44,45. It is also important to highlight that most of the detected introductions happened in the first stages of the studied periods or even a bit before them, so these variants may have already been circulating in the country before they were detected.

The number of introductions into Spain per variant also highlights the different impacts of external introductions during the pandemic, depending on the country and the specific VOC. In the case of Delta, having detected 202 individual introductions after using 1000 Spanish sequences, it nearly doubles the capacity of Alpha and Omicron-BA.1 of establishing local transmission chains and is noticeably higher than the one observed in the UK for the same variant (1458 detected introductions for 52,992 sampled sequences⁴⁶). Conversely, the contribution of the Omicron-BA.1 introductions was only slightly higher relative to the UK (6455 detected introductions for 81,039 sampled sequences⁴⁴) but lower than in Mexico (160 detected introductions for 641 sequences⁴⁷).

Lastly, we acknowledge that the mobility data we used present limitations as they depend on the number of users sharing their location with the application and, therefore, may only be representative of some of the population. Nonetheless, this type of data has already been used to study SARS-CoV-2 transmission linked to human mobility^6,11,48, and can still be considered a good approximation that should reduce the sampling bias in our analysis. Indeed, other types of sampling bias still exist, such as the difference in sequencing effort and strategy between countries and between Spanish regions, which can be substantial (Supplementary Fig. 6). Despite this, in most cases, we believe the number of sequences available relative to the number of sequences used was large enough to ensure an adequate distribution of sampling dates and locations, mitigating this source of bias.

Another caveat is that it is not easy to back up some of the conclusions about the impact of the control measures on the number of introductions and their local spread using more quantitative analyses. This is because control policies were not always applied consistently or in the same period across Spanish territories and cities. Besides, the time series of the introductions is very sparse, diminishing statistical power. All in all, our analysis of the SARS-CoV-2 introductions in Spain suggests that the control measures might have reduced further importations of the virus into the country. Moreover, given that the number of cases does not correlate with the number of introductions and the relatively small number of descendants detected for most introductions, it seems that foreign importations had a limited impact on the progression of the pandemic in Spain.