Colonisation Patterns of Nosema ceranae in the Azores Archipelago

Simple Summary Nosema ceranae is an emergent honey bee pathogen that has now invaded most of the world. However, geographically-isolated places that are free of this pathogen may still exist, and the Azores may be one of them. Here, we used molecular tools to see whether N. ceranae has entered the Azores and how far it has colonised the archipelago. In 2014/2015 we sampled 474 colonies from eight islands, and in 2020 we re-sampled 91 colonies from four islands. Our results showed that N. ceranae was present on all islands but Santa Maria and Flores. In the 2014/2015 sampling, Pico, the island of Varroa destructor entry in the Azores, showed the greatest prevalence. Resampling in 2020 revealed that N. ceranae built-up on Terceira and São Jorge. Our findings suggest that N. ceranae colonised the archipelago recently, and it spread across the other islands. Santa Maria is also free of V. destructor, making it one of the remaining areas in the world where bees are naive to both stressors. This study will help the veterinary authority establish biosecurity rules for the movement of bees and hive products among islands to maintain the N. ceranae-free status of Santa Maria and Flores. Abstract Nosema ceranae is a highly prevalent pathogen of Apis mellifera, which is distributed worldwide. However, there may still exist isolated areas that remain free of N. ceranae. Herein, we used molecular tools to survey the Azores to detect N. ceranae and unravel its colonisation patterns. To that end, we sampled 474 colonies from eight islands in 2014/2015 and 91 from four islands in 2020. The findings revealed that N. ceranae was not only present but also the dominant species in the Azores. In 2014/2015, N. apis was rare and N. ceranae prevalence varied between 2.7% in São Jorge and 50.7% in Pico. In 2020, N. ceranae prevalence increased significantly (p < 0.001) in Terceira and São Jorge also showing higher infection levels. The spatiotemporal patterns suggest that N. ceranae colonised the archipelago recently, and it rapidly spread across other islands, where at least two independent introductions might have occurred. Flores and Santa Maria have escaped the N. ceranae invasion, and it is remarkable that Santa Maria is also free of Varroa destructor, which makes it one of the last places in Europe where the honey bee remains naive to these two major biotic stressors.


Introduction
Western honey bees (Apis mellifera Linnaeus; Hymenoptera: Apidae) play a crucial role in the maintenance of ecosystems' biodiversity through their pollination services. Yet, Since 2008, Nosema spp. has been annually surveyed in the Azores by the regional veterinary authority. According to the annual reports, the only islands that are negative to the microsporidia are the islands of Santa Maria and Flores, together with the most recently colonised island of Corvo, while the island of São Jorge became positive in 2018. [48]. However, since spore detection was based on morphological methods under light microscopy, the question remains as to which Nosema species underlies the detection. This is because the spores of N. apis and N. ceranae are very similar, and there is no data on field symptoms that could help discriminate between the two species. It was supposed that nosemosis was caused by N. apis, the species associated with A. mellifera before the recent worldwide spread of N. ceranae [16,18,19,49,50], as the geographical isolation and the early biosecurity measures, imposed by the veterinary authority after arrival of the V. destructor to Pico (Dr. Paula Vieira, Direção Regional da Agricultura e Desenvolvimento Rural dos Açores, pers. comm.), could be able to deter the new honey bee pathogen from entering the Azores. However, if N. ceranae succeeded in reaching the Azores, its transmission among honey bee populations that are isolated on each island would be difficult, allowing unique conditions for studying an emergent pathogen colonising a pristine territory. While V. destructor and N. ceranae are serious pests for a colony [32,51,52], when they are simultaneously present, it is possible that their impact is further aggravated by a summative or synergistic interaction. Whether this interaction exists is unclear, although some studies have found a positive correlation between the two parasites [53][54][55][56][57]. In the event that N. ceranae is detected, then the Azores provide a unique stage for addressing this issue because there are islands where the mite is present and islands that are mite free. Hereby, the objectives of this study were to (i) determine whether N. ceranae is present in the Azores; (ii) infer its prevalence and levels of infection throughout the archipelago; (iii) search for any association between N. ceranae and V. destructor; finally, (iii) explore possible colonisation scenarios of this emergent pathogen.

Survey and Sampling Collection
The number of colonies to be sampled was calculated in relation to the number of apiaries registered in 2013 with an expected prevalence of Nosema species of 15%, a precision rate of 10% and a confidence level of 95%. The number of samples was subsequently distributed in proportion to the number of apiaries on each island in which colonies were selected randomly (Table S1). In this cross-sectional study, 474 colonies (representing a total of 156 apiaries) were sampled from eight Azorean islands in July and August of 2014 and 2015 ( Figure 1 and Table 1). While most apiaries were represented by three colonies, there were a few apiaries that were represented by either two or four colonies (Table S2). Over 150 workers were collected from the outside frame of brood nest of the 474 colonies to appropriate card boxes and then shipped alive to Centro de Investigação de Montanha (CIMO; latitude 41 • 47 53.19 N, longitude 6 • 45 56.89 W).
Given the rise of Nosema spp. on São Jorge in 2018, as detected by the veterinary authority, in 2020 we decided to re-sample this island and the following three additional islands for comparison purposes: Santa Maria (island negative to Nosema spp., according to the morphological reports), Faial (island positive to V. destructor) and Terceira (island negative to V. destructor but positive to Nosema spp., according to the morphological reports). A total of 91 colonies, representing 34 apiaries, were sampled between July and August of 2020 using the same protocol as in 2014/2015 ( Figure 1 and Table 1). The 565 samples were stored at −80 • C until molecular analyses.
Vet. Sci. 2022, 9, 320 4 of 16 to the morphological reports), Faial (island positive to V. destructor) an negative to V. destructor but positive to Nosema spp., according to th reports). A total of 91 colonies, representing 34 apiaries, were sampled August of 2020 using the same protocol as in 2014/2015 (Figure 1 and samples were stored at −80 °C until molecular analyses.

Nosema spp. Extraction and Detection for the Prevalence Study
Total DNA was extracted from a pool of 120 workers for each o (collected in 2014/2015), following the protocol described in Martín-Her with minor modifications. Briefly, 120 workers were added to a do (BA6040, Seward, Worthing, UK) with 18 mL of 50% AL Buffer ( Germany) and crushed in Stomacher 80 (Stomacher 80-Microbiom Worthing, UK) for two minutes at low speed, followed by adding 9 mL and one more cycle of homogenization (60 s). Subsequently, those centrifuged at 3000 rpm for 10 min, and the pellet was resuspended i water. To improve DNA yield, the mechanical lysis of the spores was a 400 µL of resuspended pellet to a 96-well plate (Qiagen ® , Hilden, DE) beads (2 mm diameter, Sigma), which was shaken in a TissueLyser m Hilden, DE) for three minutes. A negative control with no honey bee s

Nosema spp. Extraction and Detection for the Prevalence Study
Total DNA was extracted from a pool of 120 workers for each of the 474 samples (collected in 2014/2015), following the protocol described in Martín-Hernández et al. [58] with minor modifications. Briefly, 120 workers were added to a double bag strainer (BA6040, Seward, Worthing, UK) with 18 mL of 50% AL Buffer (Qiagen ® , Hilden, Germany) and crushed in Stomacher 80 (Stomacher 80-Microbiomaster ® , Seward, Worthing, UK) for two minutes at low speed, followed by adding 9 mL of 50% AL buffer and one more cycle of homogenization (60 s). Subsequently, those macerates were centrifuged at 3000 rpm for 10 min, and the pellet was resuspended in 3 mL of MilliQ water. To improve DNA yield, the mechanical lysis of the spores was applied by adding 400 µL of resuspended pellet to a 96-well plate (Qiagen ® , Hilden, DE) containing glass beads (2 mm diameter, Sigma), which was shaken in a TissueLyser machine (Qiagen ® , Hilden, DE) for three minutes. A negative control with no honey bee sample was set for every 20 samples and processed in parallel. Finally, 150 µL of macerate was used for DNA extraction following the BS96 DNA Tissue extraction in a BioSprint workstation (Qiagen ® , Hilden, DE). The DNA extracts were frozen at −20 • C until the next analysis.
N. ceranae, N. apis and an internal control of A. mellifera DNA (COI gene) were simultaneously screened by using a multiplex-PCR developed by Martín-Hernández et al. [58]. PCR primers are shown in Table S3 whereas the thermal profile and reactions set up are detailed in Martín-Hernández et al. [58]. Nontemplate (NTC) and positive controls were included in each run. PCR products were analysed by capillary electrophoresis in the QIAxcel advanced apparatus using a QIAxcel DNA High Resolution Kit (Qiagen ® , Hilden, DE).

N. ceranae Extraction and Load Determination
Since the previous DNA extracts of the 2014/2015 sampling period were accidentally lost, DNA was isolated de novo from a pool of 30 remaining workers and from all the samples collected in 2020, for real-time qPCR purposes. Fifteen per cent of the colonies were excluded for further analysis due to the insufficient number of workers to pool, reducing the sample size of the 2014/2015 collection to 403 (Table S2). These samples were used to determine the N. ceranae load and to compare the prevalence between both periods (2014/2015 and 2020). Briefly, the 30 workers were placed in a double bag strainer (BA6040, Seward, Worthing, UK) and added 6 mL of cool DEPC water (E476, VWR, Pennsylvania, US). Subsequently, the homogenization step was carried out using the MixWell Lab Blender (Alliance Bio Expertise ® , Guipry-Messac, FR) for the following two cycles: one of 60 s followed by another 30 s, with 30 s pause between each cycle. DNA was extracted from the tissue homogenates by using the standard protocol for animal tissue from NucleoSpin Tissue commercial kit (Macherey-Nagel TM , Düren, Germany), with minor modifications. To that end, 50 µL of the homogenates and 180 µL of T1 buffer were added to 2 mL tubes containing two zirconia beads (3 mm, Specanalítica Lda., Carcavelos, PT) to proceed with the mechanical tissue disruption using Precellys (Bertin Instruments, Montigny-le-Bretonneux, FR) with the following protocol: 6200 rpm; 5 s; 3 times. Subsequently, the NucleoSpin Tissue protocol extraction was followed without any modification. The yield and DNA quality were verified through spectrophotometry (Lvis chip, SpectroStar Nano, BMG Labtech, Ortenberg, Germany) and DNA extracts were normalized to 10 ng/µL. Our approach to quantifying N. ceranae load per colony combined the primers described by Martín-Hernández et al. [58] and the SYBR ® Green chemistry (primers shown in Table S3). Real-time qPCR assays were performed in the QuantStudio 5 apparatus (Applied Biosystems ® , Masschusetts, USA). qPCR reactions were carried out in 10 µL total volume (two replicates per sample), containing 10 ng of DNA, 5 µL of 2× iTaq Universal SYBR ® Green Supermix (Biorad ® , California, USA), 300 nM of each primer (Table S3) and 2 µL of DEPC water. The thermal conditions were set according to the SYBR ® Green manufacturer's instructions. Positive controls used to establish the standard curves were made by concentrating known positive samples for N. ceranae through PCR. PCR products were quantified using the LVis Chip (SpectroStar Nano, BMG Labtech, Ortenberg, Germany) apparatus. Subsequently, PCR products were used to prepare ten-fold serial dilutions with concentrations (7 points in duplicate) ranging from 1.75 × 10 −2 to 1.75 × 10 −8 ng/µL ( Figure S1). Melting curve analysis was conducted to detect primer dimers and/or unspecific amplicons. All samples amplifying before the last point of the standard curve, with an amplification plot showing an exponential increase and a melting profile matching the melting temperature of the positive controls, were classified as N. ceranae positive. In addition, the RPL8 gene of A. mellifera was analysed in those samples to check whether Vet. Sci. 2022, 9, 320 6 of 16 DNA extraction was successful ( [59]; see primers in Table S3). N. ceranae loads were used to establish the following three infection classes: low for loads < 10 −8 ng/µL, medium for loads between 10 −5 ng/µL and 10 −8 ng/µL and, high for loads ≥ 10 −5 ng/µL.

Statistical Analysis
To evaluate N. ceranae prevalence differences between sampling periods and its association with V. destructor, the Chi-square test (χ 2 ) was applied. Whenever the sample size of one of the variables was smaller than 5 counts, one of the Chi-square test assumptions was violated, and the Fisher's exact test was applied. The strength of the association between N. ceranae prevalence and V. destructor (presence/absence at the island level) was assessed with Cramer's V measurement (ϕ c ).
Since the qPCR data did not meet the assumptions of parametric tests, differences in N. ceranae loads among islands were assessed using the Kruskal-Wallis or Mann-Whitney tests. All statistical analyses and graphical representations were carried out in R studio software (version 4.0.2) [60] and the level of significance was set at 95% (α = 0.05).

Quality Control
All samples analysed with the multiplex PCR method amplified the internal control of A. mellifera DNA (COI gene), confirming DNA integrity. Furthermore, none of the DNA extraction controls or non-template controls (NTCs) were amplified, suggesting no cross-contamination during sample processing and analysis. Regarding real-time qPCR, all samples validated DNA integrity by amplifying the A. mellifera RPL8 gene, and all NTCs showed no amplification, indicating that there was no cross-contamination throughout the study. The amplification efficiency of N. ceranae was 97.4% ( Figure S1). Altogether, these quality control results indicate that the data analysed below is of good quality.

Detection and Prevalence of Nosema spp. across the Azores
The prevalence of Nosema infection in the Azores was inferred from the multiplex-PCR output for the samples from the 2014/2015 period, as allowed by the sampling design. A total of 56 (11.8%), 9 (1.9%), and 6 (1.3%), out of 474 colonies, tested positive for N. ceranae, N. apis, and co-infection, respectively, using the multiplex PCR approach (Table 2 and Figure 2). N. ceranae showed the highest prevalence on Pico (50.7%), followed by Graciosa and Terceira, with 14.3% and 10.3%, respectively. On São Miguel, Faial and São Jorge, N. ceranae prevalence was ≤4.0% and Santa Maria and Flores did not have any positive colony. N. apis was detected in five islands (Flores, Graciosa, São Jorge, São Miguel, and Terceira), in general, with a lower prevalence than N. ceranae, ranging from 1.3 to 5.1% ( Table 2). The co-infection was even rarer, with one positive colony in São Jorge (2.7%), one positive colony in Terceira (1.3%), and four positive colonies in São Miguel (4.0%).
To assess whether the infection changed between 2014/2015 and 2020, samples were processed for both periods using the same qPCR conditions. The overall prevalence for the first sampling period showed that 34 (8.4%) of the 403 screened colonies were positive for N. ceranae (Table 2 and Figure 3). N. ceranae could be detected only on Pico (43.7%), Graciosa (5.6%), Terceira (1.4%), and São Miguel (1.1%). On Faial, the two colonies that tested positive for N. ceranae by the multiplex PCR approach were negative by qPCR. The single positive colony of São Jorge was not re-analysed because there were not enough workers for pooling. In the 2020 sampling, the highest number of N. ceranae positive colonies was recorded on Terceira (57.1%), closely followed by São Jorge (50%). No infection of N. ceranae was identified on Faial and Santa Maria. Notably, the months of July and August of 2020 were considerably dryer than those of 2014 and 2015 (Table 1), which may have influenced the build-up of the infection observed on these two islands.
single positive colony of São Jorge was not re-analysed because there were not workers for pooling. In the 2020 sampling, the highest number of N. ceranae colonies was recorded on Terceira (57.1%), closely followed by São Jorge (50 infection of N. ceranae was identified on Faial and Santa Maria. Notably, the mo July and August of 2020 were considerably dryer than those of 2014 and 2015 (T which may have influenced the build-up of the infection observed on these two is While the overall prevalence patterns are consistent between the qualitati (multiplex) and the quantitative approach (qPCR), a reanalysis of the 2014/2015 by qPCR, using a lower number of pooled workers (30 instead of 120), led to the d of a lower number of positives, indicating that false negatives may arise from sm sizes. This is expected when the proportion of infected individuals in the colony requiring a higher pool size for detecting infected colonies [61].
Statistical analysis of N. ceranae prevalence data, obtained from qPCR, rev highly significant difference (p < 0.001; Chi-square test) between sampling periods 3), which is explained by the dramatic rise in positive colonies observed in 2 Terceira and São Jorge (p = 0.001 for both islands; Fisher's exact test).  While the overall prevalence patterns are consistent between the qualitative PCR (multiplex) and the quantitative approach (qPCR), a reanalysis of the 2014/2015 dataset by qPCR, using a lower number of pooled workers (30 instead of 120), led to the detection of a lower number of positives, indicating that false negatives may arise from small pool sizes. This is expected when the proportion of infected individuals in the colony is low, requiring a higher pool size for detecting infected colonies [61].
When categorising the infection level of N. ceranae according to the qPCR loads in the 2014/2015 sampling period, most colonies exhibited a medium infection (23; 75%), one colony exhibited a severe infection (3.1%), and seven colonies exhibited a low infection (21.9%). Notably, the majority of the medium-infected colonies originated from Pico, and they were located on the north-western part of the island ( Figure 4C). In the 2020 sampling period, 14 (45.2%) positive colonies were categorised as having a high infection, 15 (48.4%) a medium infection, and 2 (6.4%) a low infection. The most highly infected colonies originated from São Jorge ( Figure 4B).

N. ceranae and V. destructor Associations
The association between N. ceranae prevalence and V. destructor was tested using the multiplex PCR and the qPCR datasets separately. A statistically significant positive association was found in the 2014/2015 sampling period (φc = 0.25; p < 0.001, Chi-square test) ( Figure 5A). This finding was corroborated by the qPCR qualitative data (φc = 0.32; p < 0.001, Fisher's exact test), despite the lower number of tested samples ( Figure 5B). In contrast, the prevalence of N. apis (φc = 0.08; p = 0.163, Fisher's exact test) or co-infection (φc = 0.09; p = 0.087, Fisher's exact test) did not reveal any relationship with V. destructor presence. Since only one (Faial) of the sampled islands from 2020 has V. destructor, the association analysis could not be performed.

N. ceranae and V. destructor Associations
The association between N. ceranae prevalence and V. destructor was tested using the multiplex PCR and the qPCR datasets separately. A statistically significant positive association was found in the 2014/2015 sampling period (ϕ c = 0.25; p < 0.001, Chi-square test) ( Figure 5A). This finding was corroborated by the qPCR qualitative data (ϕ c = 0.32; p < 0.001, Fisher's exact test), despite the lower number of tested samples ( Figure 5B). In contrast, the prevalence of N. apis (ϕ c = 0.08; p = 0.163, Fisher's exact test) or co-infection (ϕ c = 0.09; p = 0.087, Fisher's exact test) did not reveal any relationship with V. destructor presence. Since only one (Faial) of the sampled islands from 2020 has V. destructor, the association analysis could not be performed.

Discussion
This study documents for the first time the presence of N. ceranae in the Azore ther extending its distributional range in the world (reviewed by Klee et al. [16] and and Quandt [49]) and specifically in Macaronesia, where it had only been reported Canaries [43]. Nosema infection has occurred in the Azores at least since 2008, acc to the first spore identification by microscopical analysis produced by the local vete authority [46]. Whether the early cases were due to N. apis or N. ceranae is unknown. over, unknown is the year and exact location of the first arrival of N. ceranae to the pelago, contrasting with the known putative N. ceranae-free status of the following its islands: Santa Maria and Flores. Remarkably, despite the widespread distributio ceranae in the Azores, samples from these two islands were all negative in bo 2014/2015 and 2020 sampling periods. These molecular results, together with the m logical results reported by the Azorean veterinary authority for several years (from to 2021; Direção Regional da Agricultura e Desenvolvimento Rural dos Açores [46,48]), st suggest that Santa Maria and Flores are still free of this harmful pathogen.
In the 2014/2015 sampling, N. ceranae prevalence was below 10.3% for six of th surveyed islands, with Santa Maria and Flores exhibiting values of 0%. However, 2020 sampling, while N. ceranae went undetected on Santa Maria and Faial, the ep ological situation aggravated considerably in São Jorge and Terceira, as suggested b 50% of samples testing positive in qPCR on both islands. Due to the geographically l sampling effort in 2020, these results should be interpreted with caution, particula Faial where only two apiaries were examined. Nonetheless, the increasing trend ob on São Jorge and Terceira, which is supported by the morphological data [48], can ignored. These findings call for a thorough molecular survey across the entire archi so that a more rigorous appraisal of the epidemiological status of the Azorean hon populations, and a further confirmation of the N. ceranae-free status of Flores and Maria, can be carried out. This is particularly important for Santa Maria as this is also free of V. destructor and harbours the purest Iberian honey bee (A. m. iberiensis ulation in the Azores [45].

Discussion
This study documents for the first time the presence of N. ceranae in the Azores, further extending its distributional range in the world (reviewed by Klee et al. [16] and Grupe and Quandt [49]) and specifically in Macaronesia, where it had only been reported for the Canaries [43]. Nosema infection has occurred in the Azores at least since 2008, according to the first spore identification by microscopical analysis produced by the local veterinary authority [46]. Whether the early cases were due to N. apis or N. ceranae is unknown. Moreover, unknown is the year and exact location of the first arrival of N. ceranae to the archipelago, contrasting with the known putative N. ceranae-free status of the following two of its islands: Santa Maria and Flores. Remarkably, despite the widespread distribution of N. ceranae in the Azores, samples from these two islands were all negative in both the 2014/2015 and 2020 sampling periods. These molecular results, together with the morphological results reported by the Azorean veterinary authority for several years (from 2008 to 2021; Direção Regional da Agricultura e Desenvolvimento Rural dos Açores [46,48]), strongly suggest that Santa Maria and Flores are still free of this harmful pathogen.
In the 2014/2015 sampling, N. ceranae prevalence was below 10.3% for six of the eight surveyed islands, with Santa Maria and Flores exhibiting values of 0%. However, in the 2020 sampling, while N. ceranae went undetected on Santa Maria and Faial, the epidemiological situation aggravated considerably in São Jorge and Terceira, as suggested by over 50% of samples testing positive in qPCR on both islands. Due to the geographically limited sampling effort in 2020, these results should be interpreted with caution, particularly on Faial where only two apiaries were examined. Nonetheless, the increasing trend observed on São Jorge and Terceira, which is supported by the morphological data [48], cannot be ignored. These findings call for a thorough molecular survey across the entire archipelago so that a more rigorous appraisal of the epidemiological status of the Azorean honey bee populations, and a further confirmation of the N. ceranae-free status of Flores and Santa Maria, can be carried out. This is particularly important for Santa Maria as this island is also free of V. destructor and harbours the purest Iberian honey bee (A. m. iberiensis) population in the Azores [45].
Since the first discovery in colonies sampled in 2005 in Spain [14] and in Taiwan [15], N. ceranae has been detected with high prevalence rates in all continents where A. mellifera is present, indicating a very high invasive potential [16,49,62]. In the European continent, N. ceranae reached 95-98% prevalence in Hungary [63], 84% in the Balkan countries [20], 77.2% in the northern part of Bulgaria [64], 59.7% in Belgium [65], over 63.2% in Italy [23,66], or 58.5% in Spain [67]. In mainland Portugal, the possible origin of the spores introduced in the Azores, N. ceranae prevalence was over 60% in a survey conducted in 2012 (report published by Federação Nacional dos Apicultores de Portugal [68]). High prevalence rates have also been reported for islands. For instance, N. ceranae reached 97.1% prevalence in Cuba [35], 63% in the Dominica Islands [36], or 75% in the Macaronesia archipelago of the Canaries [43].
All those regions were in stark contrast with the epidemiological situation of the Azores, where two islands seemed to remain free of N. ceranae and five islands exhibited prevalence rates lower than 14.3% in the 2014/2015 sampling. Only Pico, the island where V. destructor was first introduced, registered a large proportion of infected colonies, regardless of the molecular approach used (43.7-50.7%). This result suggests that, as for V. destructor, Pico might have acted as the first entry point for N. ceranae in the Azores. If this was the case, it is possible that N. ceranae spores hitchhiked along with V. destructor in the queens parcels illegally imported in 2000, as this pathogen has been shown to be infecting A. mellifera since at least 1995 in the USA [17] and 1998 in Europe [69]. Alternatively, Pico did not act as the entry point in the Azores and N. ceranae could have been introduced later to this island through contaminated hive products [70][71][72], originating from the varroa-free islands, mainland or elsewhere, and V. destructor aided infection development by lowering the immune defences of the honey bees [73][74][75]. If the first hypothesis is true, then the stringent restrictions on the circulation of honey bees and hive products from the varroainvaded islands onto the varroa-free islands, imposed early by the veterinary authority (Dr. Paula Vieira, pers. comm.), imply that at least one additional independent introduction of N. ceranae occurred on São Miguel, São Jorge, Terceira, or Graciosa. On Flores, the other illegal queen import was seemingly free of N. ceranae, as suggested by the negative results obtained from either molecular or morphological methods.
The introduction of N. ceranae on the varroa-free islands likely occurred via commercial hive products, as the importation of queens or packaged honey bees has been restricted for over 22 years. Honey, pollen, royal jelly, and wax foundation can all contain viable spores and thus operate as transmission vehicles of N. ceranae through trading [70][71][72]. Among these, the wax foundation is the more probable original source of introduction in the varroa-free islands because there is a high demand for this hive product in the Azores and sterilisation of wax imports has only been compulsory since 2010 (Dr. Paula Vieira, pers. comm.). Regardless of the places and means of N. ceranae's entrance, the spatial and temporal prevalence patterns suggest that N. ceranae arrived recently in the Azores, and it is invading the archipelago at a fast pace. Since its entry, it is behaving as an emerging pathogen, as evidenced by the rising prevalence observed in some islands.
In addition to prevalence, N. ceranae infection loads were determined for all positive samples identified by real-time qPCR. In 2014/2015, the three single positive samples, originating from varroa-free islands, had low (Graciosa and São Miguel) and medium (Terceira) infection intensities. On Pico, there were samples with low as well as high infection intensities, but the great majority (75%) exhibited medium intensity, and these were mostly located in the northwest. This part of the island has a higher concentration of apiaries, which might have facilitated the transmission of N. ceranae as the spores are considered bioaerosols that can be carried in the air and deposited in natural environments, including flowers [76]. Other ways to spread the infection could be through sharing food or water sources, or the effects of drifting or robbing infected colonies. In the 2020 sampling, the two islands (Terceira and São Jorge) that showed high prevalence also showed high loads, particularly São Jorge. Strikingly, in the previous sampling, only one sample was identified as positive on this island, and now, in addition to the high prevalence (50.0%), most samples displayed infection levels classified as high. The infection loads, together with the temporal prevalence patterns, suggest a recent establishment of N. ceranae on these two varroa-free islands, and now the pathogen is rapidly spreading, showing the typical behaviour of an emergent pathogen, with an increase in the prevalence and in the level of infection. Therefore, despite the positive association between V. destructor and N. ceranae prevalence detected in an earlier stage of the invasion (2014/2015), and also reported in other studies [53][54][55][56][57], these findings suggest that the mite is not a mandatory condition for a successful invasion of the pathogen.
Once N. ceranae is introduced into a pristine territory, several beekeeper-independent factors, such as pesticide exposure, quality of pollen forage, and climate, can influence the establishment of the pathogen and the build-up of the infection [77][78][79][80][81][82]. Yet, new introductions into geographically isolated places such as the Azores are dependent on humans for transportation of propagules. In the case of N. ceranae, long-distance transmission is facilitated by the trade of infected honey bees or contaminated hive products [70,71]. There are only two islands, Flores and Santa Maria (and perhaps Corvo), that remain free of N. ceranae (according to the veterinary report for 2021 [48]), and this status can only be perpetuated if biosecurity restrictions are capable of continuing to prohibit any importation attempt of honey bees and decontamination of imported hive products used in beekeeping (e.g., wax) is assured.

Conclusions
This is the first molecular survey of Nosema spp. carried out in the Azores and consequently the first detection of N. ceranae in this Macaronesia region. Nosema spp. was identified on some of the islands in the first morphological report released by the local veterinary authority in 2008. Whether these early cases were due to N. apis or N. ceranae is unknown. However, there is a chance that N. ceranae was already present in 2008, at least on Pico, as suggested by the high prevalence rate and infection loads found on this island in the 2014/2015 sampling. The spatial and temporal patterns are compatible with a recent colonisation hypothesis, after N. ceranae has been introduced in the Azores, likely multiple times. Flores and Santa Maria have so far seemed to avoid N. ceranae invasion, a situation that can be sustained if beekeepers comply with biosecurity regulations. While Flores has V. destructor, Santa Maria is free of this parasite. Therefore, Santa Maria is one of the last places in Europe, and perhaps in the world, where honey bees remain naive to two of the major honey bee biotic stressors, making this island unique for beekeeping activity.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/vetsci9070320/s1; Figure S1: Real-time qPCR standard curves for N. ceranae; Table S1: Distribution of registered apiaries and colonies per island in 2013; Table S2: Metadata of sampled apiaries; Table S3: Primers used for molecular detection of N. ceranae; N. apis and internal controls for Apis mellifera.