Time series data collection

To analyse the emergence of SBD and its potential link to climate, we analysed complete time-series data of disease occurrence in France and Switzerland during the last three decades, from 1990 to 2021 and modelled this occurrence as a function of different climatic variables. The French disease records during these three decades were obtained from the database of the French Forest Health Department (DSF, French acronym). This database contains annual records of forest health problems observed in France by a network of foresters trained for the diagnosis of abiotic, entomological or pathological damages. The Swiss data were obtained from the forest protection reports generated by the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL) (Queloz et al. 2020). The records in this database are based on the specific symptoms of the disease and diagnostic in the respective laboratories when symptoms are not conclusive. To account for potential sampling bias in our database (i.e. different monitoring intensities across time and regions), we standardised our data following a procedure commonly-used in medical epidemiology (Lawson 2001), that has also been applied in forest pathology (Fabre et al. 2012). Briefly, we computed a record rate RR_ij = NSBD_ij/NRef_ij where NSBD_ij and NRef_ij are, respectively, the number of SBD cases and the number of other reported health problems concerning sycamore maple other than SBD, for year i and country j. The NSBD_ij is used here as a proxy for both the observation pressure and the density of the host which cannot be separated in the dataset. This report rate (RR_ij) was then standardised by dividing by the report rate over the entire data set, i.e. including all years (Eq. 1).

$SR{R}_{ij}=\frac{NSB{D}_{ij}}{N{Ref}_{ij}}·\frac{NRef}{NSBD}$ Eq. 1

where NSBD and NRef are, respectively, the total number of SBD and reference cases for the entire data set. Thus, a value of X for SRR_ij means that the report rate is X times the average report rate. Therefore, we assumed that a SRR_ij higher than 1 was an outbreak of the disease, as the reported number of cases exceeds the the basal level of the disease (considered to be the global average of our database, i.e. NSBD/NRef). The distribution of the SBD records in France and Switzerland are shown in Suppl. material 1.

The climatic data were obtained from Météo-France (SAFRAN database) computed on a daily basis on an 8-km resolution grid throughout France and Switzerland (except for the Tessin region, where these data were not available) (Suppl. material 1). We selected all points of the climate grid close to each SBD record made in the 1989–2021 period. We computed eight variables, related to high summer temperature and drought (Ogris et al. 2021) and to limiting winter conditions, to be used as predictors of the SBD occurrence: the average daily maximal temperature in summer (July-August, T_Xsummer), in spring (April-June, T_Xspring) and in the vegetative season (April-August, T_Xveg), the water balance calculated as the sum of the daily difference between rainfall (P) and Penman-Monteith evapotranspiration (ETP) in summer (July-August, P-ETP_summer), in spring (April-June, P-ETP_spring), in the vegetative season (April-August, P-ETP_veg), the number of days in the year where the temperature exceeds 25 °C (n25) and the average daily minimum temperature in winter (January-March, T_Nwinter). We chose the threshold of 25 °C, as it has been reported as the optimal growth temperature of Cryptostroma corticale (Ogris et al. 2021). The eight variables were computed for one, two and three years preceding each disease record, which potentially contributed to or impeded disease development. We did not include the year when the disease was recorded, because disease records occurred throughout the year and not only after summer.

Aerobiological study: experimental design, sample collection and data sources

The samples used as starting material in our aerobiological study consisted of microscope slides with a ca. 48-mm portion of Melinex tape (corresponding to 24 h ± 2 h, depending on the sampling time) from Hirst-type volumetric air samplers used to monitor airborne pollen grains and fungal spores by the aerobiology networks of the involved European countries. The Hirst-type air samplers (Hirst 1952) are active vacuum-pumped suction traps with a rotating drum containing the Melinex tape covered by an adhesive solution which captures the particles present in the air (Fig. 1). Further details can be found in Lacey and West (2006) and in European Norm EN 16868:2019. The samplers of the network are placed on rooftops, at least 10–15 m high. We performed DNA extractions and qPCR targeting C. corticale (see following sections) to assess the detectability of C. corticale in aerobiological samples and to quantify the spores captured during a 24-h period.

Figure 1.

7-day volumetric air sampler (Burkard Manufacturing Co Ltd, Hertfordshire, UK) in Brno (Czechia) installed on the roof of the University hospital, 15 m above ground to ensure landscape-scale monitoring. Photo credit for Aneta Lukačevičová.

We undertook two studies, at a regional and a continental scale, to evaluate the use of permanent aerobiological networks to assess C. corticale epidemiological surveillance. The regional study focused on French samplers, while the continental study covered locations in six European countries over a wide latitudinal and longitudinal range: Czechia, France, Italy, Portugal, Sweden and Switzerland.

Regional study

We selected samplers to cover the SBD outbreak in north-eastern France in 2017 and 2018, following a two-year-spanned drought episode (from 2017 to 2018). We selected four samplers located in Mulhouse (the main focus of the outbreak), and in three locations at different distances from the main focus (with less records of the disease): Bart, Besançon and Strasbourg (Table 1, Fig. 2). We selected four additional locations in the south of France, which were available for the year 2018, and had lower historical records of the disease (Angoulême, Aurillac, Avignon and Gap; Table 1, Fig. 2). To determine the optimal sampling period, we analysed the detectability of C. corticale in a temporal series of aerobiological samples from Mulhouse every three days from the 1^st of May to the 30^th of September of 2018. The highest frequency of spores was detected in May and June. Accordingly, the sampling period and intensity for the regional study were fixed in May-June with a three-day frequency, i.e. 10 samples per location and year.

Figure 2.

Selected air samplers, SBD records for years 2017 and 2018 considered for aerobiological sampling in 2017, and years 2018 and 2019 considered for 2018 sampling; and total sycamore maple basal area (m²) in a 16×16 km grid.

Table 1.

Download as

CSV

XLSX

Selected French air samplers for the regional study with different SBD incidence.

City	Code	GPS Coordinates	Year of the first record at < 50 km	Year of the first record at < 100 km	Year of the first record at < 180 km
Mulhouse	MUL	47.7524, 7.3591	2010	2010	1992
Bart	BAR	47.4856, 6.7694	no records	2010	1992
Besançon	BES	47.2324, 6.0231	no records	2006	1992
Strasbourg	STR	48.5833, 7.7500	no records	2010	2010
Angoûleme	ANG	45.6494, 0.1645	no records	2016	1991
Aurillac	AUR	44.9258, 2.4341	no records	no records	2014
Avignon	AVI	43.9203, 4.8021	no records	no records	2002
Gap	GAP	44.5575, 6.0761	no records	no records	2002

In order to align the aerobiological data with the presence of the disease, we used the disease records from the DSF database (as described above). From 1989 to 2021, 1708 health reports were done on maples, of which 1351 were on A. pseudoplatanus and 172 corresponded to the SBD. We modelled the number of spores as a function of two variables: the distance to the disease and the maple basal area. We computed the distance to the closest disease record for each aerobiological sample (i.e. each captor) and year. We consider all the disease records taking place in both the year of the sampling and the following one, to capture the dispersion of the spores once the disease has been detected. Finally, we obtained host density data from the French National Forest Inventory (IFN, French acronym). We assigned to each sampler the sum of the total sycamore maple basal area in IFN plots in a radius of 50 km from each sampler, which is the reference area of influence of an aerobiological sampler (i.e. average distance at which the pollen is dispersed, Oteros et al. 2017), as a proxy for the host density. To test whether the radius at which we computed the maple basal area had an impact on its link to the number of spores, we tested a gradient of radius, from 40 to 130 km (by 10 km). Even though C. corticale can infect Acer spp. other than A. pseudoplatanus, such as A. platanoides and A. campestre, we only considered the latter in our study. Based on the French database, from all the SBD records from 1989 to 2021, A. pseudoplatanus is the main host (97.6% of cases).

The selected primers and probe used in this study were ccITS2F (AGGTTGTGCTGTCCGGTG), reported in the study by Kelnarová et al. (2017), and the new reverse primer and probe developed here: SBD3R (AGCTCCTACCAACTACAGGGT) and SBD5P (FAM-ACCCTGTAGGAGGAGCTACCCTGTA-BHQ1), respectively. The LOD was fixed at 0.01 pg/μl (Cq 35.9 ± 0.2) in DNA extracts from mycelium samples (see Suppl. material 3). The detection of spores with our test ranged from 2 to 1144 spores/μl.

The climatic variable best explaining the standardised SBD report rate was the water balance (P-ETP) in the vegetative season (April-August) of the year preceding the disease report (Table 3). Other models that yielded low deviance were the water balance in the summer of the year preceding disease, the mean number of days with temperature exceeding 25 °C of the two previous years of disease record, and the water balance in the spring of the year preceding disease (Table 3). We found that SRR was predicted to exceed 1 when at least 33 days per year (95% CI 29, 36) had a temperature higher than 25 °C during the two years preceding disease. The distribution of the residuals of the best model (water balance in the vegetative season) can be found in Suppl. material 4.

The number of SBD cases increased exponentially with more negative water balance (Fig. 3a). On average, the model predicted a standardised SBD report rate higher than 1 (i.e. SBD occurrence higher than average) for total water balance in spring and summer lower than -132 mm (95% CI -170, -93, Fig. 3a), which qualifies a mild drought (extreme drought events taking place around -300 and -400 mm of P-ETP, Candel-Pérez et al. 2012). The number of SBD cases in France was on average higher than in Switzerland, corresponding to more negative accumulated water deficit during the vegetative season (Fig. 3b). SBD peaks in France paralleled those in Switzerland, with a marked increase from 2018 to 2021. Even though drought peaks in both countries did not appear to increase in magnitude across the years, they did tend to be more frequent after 2005 (Fig. 3b). No annual point earlier than 2005 exceeded the average report rate (SRR above 1, Fig. 3b). According to our model, the probability of disease absence not being linked to the water balance of the vegetative season (i.e. Bernoulli process, Eq. 3) was 0.05 (95% CI 0, 0.19). This zero-inflation, not explained by the water deficit, was mostly observed during the first decade of the time-series (Fig. 3).

Figure 3.

Model prediction of standardised SBD report rate as a function of water balance (measured as P-ETP) of the vegetative season (April-August) of the year previous to the disease report (a). Evolution of standardised SBD reports from 1990 to 2021 in France and Switzerland (b), and model predictions (Eq. 2). A dotted grey line indicates a standardised record rate that equals 1, above which the number of cases of the SBD is higher than average, and hence considered an outbreak of the disease.

The number of spores detected per week was more abundant in samplers closer to disease reports (Fig. 4a). However, the magnitude of the increase was not large, and the coefficient estimate for the variable distance was not significant (i.e. the 95% CI contained the 0; Table 4). Our model predicted a detection of 27 spores per week (two sampling days per week) (95% CI 16, 40) at 10 km from the closest disease report, while 10 spores (95% CI 2, 25) were detected at a distance of 300 km (Fig. 4a). An increasing number of spores were detected in areas with a higher sycamore maple density (Fig. 4b). Even in areas with no sycamore maple in a radius of 50 km, the aerobiological samples presented C. corticale DNA, up to 20 spores. The coefficient estimate for the total sycamore maple area was positive on average, but not significant (Table 4). The distribution of the residuals of the best models (distance to the closest disease report and total maple basal area in a 50-km radius from the sampler) can be found in Suppl. materials 4, 5, respectively. The other radius tested for the computation of the maple basal area showed the same deviance, hence no other radius value led to an estimation of maple basal area that better explained the presence of spores in the air (Suppl. material 7).

Figure 4.

Number of spores per day as a function of the distance to the closest disease report (a), and as a function of the total maple basal area (m²) in a radius of 50 km from the sampler (b).

The two models (distance to the disease, and maple basal area) estimated a similar probability of spore capture (0.73 and 0.72, respectively, Bernoulli process in Eq. 8, Table 4). Therefore, 27% (or 28%) of the lack of spore detection (zero inflation) was due to a process other than the distance to the disease report (or the host density), and hence not explained by our deterministic model.

The probability of disease occurrence increased with the number of detected spores at a given distance (Fig. 5, Suppl. material 8). A 95% probability of disease in a radius of 60 km and 120 km corresponded to 41 and 127 spores detected, respectively. However, the lack of spore detection did not correspond to the absence of disease reports. On the contrary, the disease could be detected in the field without being detected in aerobiological samples with two sampling days per week.

Figure 5.

Probability of disease report in an area of 60-km and 120-km radius from the sampler as a function of the number of detected spores per day.

The proportion of positive aerobiological samples based on the quantitative species-specific C. corticale PCR assay per year in Europe followed the reported presence of the disease in Europe (Fig. 6). Further, the regions that yielded more positive samples corresponded to those of the native range of sycamore maple, and hence with higher potential to maintain a sustained C. corticale population (samplers BES, PAY, MÜN and BRN, Fig. 6). Sweden and Portugal, the two countries where the disease had not been reported yet, did not present any positive samples (Figs 6, 7c, d). The highest proportions of positive samples were detected in France, Switzerland and Czechia (Figs 6, 7). Spores tended to be detected earlier in western than eastern locations. French samplers that produced positive samples, Bordeaux and Besançon, peaked spore detection in June and July, respectively (Fig. 7a). Payerne, in Switzerland, also presented a peak in July (Fig. 7b). Swiss location Münsterlingen and the Italian Bologna, presented higher C. corticale detection in August (Fig. 7b, e). Finally, the easternmost location in Czechia, Brno, peaked in September (Fig. 7f).

Figure 6.

Proportion of aerobiological samples that tested positive from May to September of 2018 (n = 10, except for Gap where n = 5) in the different European samplers following the natural distribution of the host Acer pseudoplatanus.

Figure 7.

Phenology of spore emission of Cryptostroma corticale in the studied European countries showing the proportion of positive aerobiological samples per year. Jun: June; Jul: July; Aug: August; Sep: September.

Disease peaks increased exponentially in magnitude with time. C. corticale is an invasive pathogen, reported for the first time in continental Europe in 1952, in France. In Switzerland, the first known report was in 1991. The highest peaks of the standardized number of disease records in 2020 in both countries suggest that the pathogen, after several outbreaks of the disease, might have colonized more forest plots, where the disease was eventually able to develop after conducive weather conditions. Two processes may have then taken place that explain the exponential increase in the last decades in France and Switzerland: (1) the dispersion and the establishment of the exotic pathogen, and (2) an increasing frequency of low water balance and high temperatures which are conducive conditions for the SBD disease. The relative contribution of the two processes in the SBD emergence cannot be fully disentangled from our model. However, they are likely to have occurred additively, as the main dispersion events of the pathogen are highly dependent on drought conditions. Hence, the increased frequency of drought events and heat peaks might have led to higher dispersion rates. The estimation of the zero-inflation at 5% in our climate model suggests that the disease is mainly climate-driven (i.e. only 5% of disease absence is not explained by high water deficit in the vegetative season). Before 2005, there was no SRR higher than one, which implies an outbreak of the SBD (higher recording of SBD compared to average). The lack of SBD reports in early years may be due to the absence of inoculum (early phases of the invasive process) and the less frequent conducive conditions to the SBD. However, the lack of awareness of the disease by surveillance agents and hence little attention to symptoms in the field might have also resulted in fewer reports. Although the disease is detectable some months before sporulating lesions develop on the trunk (sooty appearance), those early symptoms are not specific to the SBD: wilting of leaves, presence of stool shoots and branch dieback (Gregory and Waller 1951). This means the disease is usually reported at its later stage. Further, the apparent similarity of dark brown stroma of C. corticale to black stroma of the common saprophytic fungus Eutypa maura (Fr.) Sacc. (Saccardo 1882) may be another reason for misidentification. Finally, the SBD is commonly found in urban environments, which are recognized to be an entry pathway for exotic pathogens (Tubby and Webber 2010; Paap et al. 2017), and hence expected to be, in proportion, more frequently found in urban areas than natural ecosystems in the early phases of pathogen invasion. Urban environments are, in addition, known to be heat islands with warmer and drier conditions than forests. These two reasons could thus also explain why early cases of the disease, being more frequent in urban areas, were poorly reflected in our time-series data which mostly included forested areas. In any case, the SBD occurrence rate in this study focuses on its emergence in forests, giving a temporal frame of the development of an invasive disease linked to climate extremes.

The presence of C. corticale in aerobiological samples paralleled the presence of the disease SBD in Europe. At the continental level, monitoring of aerobiological samples shows a great potential as a large-scale epidemio-surveillance method for the SBD in Europe. Especially, early aerial detection of C. corticale in disease-free countries, such as Portugal and Sweden, could help implement special measures for SBD detection and eradication in the field. The advantage of aerobiological monitoring is that the aerobiological networks are already established and samples can be potentially obtained periodically. This method has already been proved for other forest pathogens such as Hymenoscyphus fraxineus, Heterobasidion annosum s.l., Erysiphe alphitoides, and Melampsora larici-populina (Aguayo et al. 2020). The anamorphic C. corticale produces conidia emerging from conidiophores in the stroma formed between the inner and outer bark. Even though the release mechanism of conidia is still unknown, it has been suggested that, when the bark peels off, the spore mass of C. corticale is exposed and dispersed by wind (Gregory and Waller 1951; Bencheva 2014; Oliveira Longa et al. 2016). The number of spores has been reported to be approximately 30 to 179 million per square cm of black sooty layer (Abbey 1978). Our results suggest that the pathogen is effectively dispersed by wind. However, a few points should be considered if aerobiological surveillance is to be implemented for the SBD. First, our models estimated the detectability of the pathogen with a probability of 72%. This implies that in 28% of the cases, the pathogen’s spores may be present in the air but either not captured by the air samplers, or not detected by qPCR. This explains why our model of the probability of disease presence predicts SBD occurrence with a probability of 50% in a 60-km-radius area even when no spores are detected by the samplers. Second, spore capture efficiency may be improved by increasing the spatial resolution of the selected air samplers or by considering the movement of air masses in the selection of locations. Third, the sampling period can impact the probability of detection. In the regional study, we sampled every three days in May and June. This period could be either intensified or prolonged during additional months. The results of the European sampling show that the sporulating period may be larger than we expected based on the Mulhouse captor we had analysed before this study (not shown). A whole-year period would also be informative, in future research, to evaluate fluctuations in spore release. Finally, our aerobiological data consists of two years in the regional study and one year in the continental study, in both cases with conducive climatic conditions. A longer time-series aerobiological data would allow for assessing whether inoculum production is detectable outside the period of outbreaks.

Distance tended to decrease the number of detected spores, but the magnitude of the effect was low and it was not significant (the confidence interval of parameter estimates contained 0). We detected C. corticale spores as far as 310 km from the closest disease report. This result suggests that the fungus can disperse long distances by wind. However, we cannot rule out the possibility of underreporting with unobserved SBD occurring closer to the samplers in urban settings. It is reasonable to assume that the main SBD foci in forests were registered in our database from 2017–2018 onwards, as the disease was well known at that time by the surveillance agents. But, outbreaks in parks or along roads may not have been as comprehensively included in our database. We did not sample locations at distances farther than 310 km. Therefore, we cannot establish the limit of aerial dispersal of the fungus. Other wind-borne pathogens have shorter dispersal distances, such as H. fraxineus, the causal agent of the ash dieback disease, whose spores can be detected up to 50–100 km from the disease front (Grosdidier et al. 2018). Our results support the idea that C. corticale is an efficiently dispersed pathogen. However, the high spore detection could have been remarkably favoured by the fact that the sampled years were affected by heat waves and the number of SBD cases was inherently high. Extended time-series aerobiological analyses are needed to further understand the epidemiology and dispersion pattern of the SBD.

The relatively abundant number of spores of C. corticale detected in the surveyed air samplers, placed in cities, reveals a potential risk to human health. The spores of C. corticale cause the MBD (Emanuel et al. 1962), currently called Maple Bark Stripper Lung (WHO 2022), a hypersensitivity pneumonitis, allergic asthma, flu-like infections and interstitial pneumonia (Braun et al. 2021). With drought predicted to increase in the Mediterranean basin and western and central Europe in the following decades (IPCC 2021), a further expansion and intensification of the SBD can be expected. Thus, the risks to human health and the environment are intertwined, as the One Health approach states. Surveillance of spore levels in the air is crucial to assess disease risk. We have presented here a suitable methodology, including the use of aerobiological samples to monitor the evolution of the SBD, that can also provide data to assess the potential risk of MBD to humans. Assessing SBD epidemiology at a continental scale implies access to harmonized databases of both pathogen occurrence and host density. We seemingly succeeded in our joint effort to homogenise molecular protocols to reduce the bias in molecular detection of C. corticale in aerobiological samples. However, there is a need to homogenise data from forest national inventories and disease reporting at the European level. That would allow models including larger geographical areas, which would provide a better understanding of disease dynamics.