Immune responses following the first dose of the Sputnik V (Gam-COVID-Vac)

As the first dose of Gam-COVID-Vac, is currently used as a single dose vaccine in some countries, we investigated the immunogenicity of this at 4 weeks (327 naïve individuals). 88.7% seroconverted, with significantly lower seroconversion rates in those over 60 years (p = 0.004) and significantly lower than previously seen with AZD1222 (p = 0.018). 82.6% developed ACE2 receptor blocking antibodies, although levels were significantly lower than following natural infection (p = 0.0009) and a single dose of AZD1222 (p < 0.0001). Similar titres of antibodies were observed to the receptor binding domain of WT, B.1.1.7 and B.1.617.2 compared to AZD1222, while the levels for B.1.351 were significantly higher (p = 0.006) for Gam-COVID-Vac. 30% developed ex vivo IFNγ ELISpot responses (significantly lower than AZD1222), and high frequency of CD107a expressing T cells along with memory B cell responses. Although single dose of Gam-COVID-Vac was highly immunogenic, administration of a second dose is likely to be beneficial.

www.nature.com/scientificreports/ Although the seroprevalence of adenovirus 26 is less than the seroprevalence of adenovirus 5, the seroprevalence rates of these viruses vary widely in different populations 11,12 . Therefore, based on the seroprevalence rates of adenovirus 26 in a given population, the immunogenicity of a single dose vaccine using a human adenovirus vector may change.
Currently, while Gam-COVID-Vac is used in Sri Lanka, many individuals are only given the first dose of Gam-COVID-Vac. As there are no data regarding the seroprevalence of adenovirus 26 virus in Sri Lanka, it would be important to evaluate the immunogenicity of the first dose (rAd26-S) in a real-world scenario. Therefore, we studied antibody responses to the SARS-CoV-2 virus, antibodies to the receptor binding domain (RBD) of the ancestral Wuhan variant and other VOCs, ex vivo T cell responses and their functionality, and memory B cell responses in a large cohort of Sri Lankan individuals who received the first dose of the Gam-COVID-Vac. In addition, in order to compare the immunogenicity of a single dose of the rAd26-S with another adenovirus vector vaccine, we compared the immunogenicity of this vaccine with previously published data of AZD1222 single dose responses at 4 weeks following vaccination in Sri Lankan individuals 13 .

Results
Seroconversion rates to the first dose of Gam-COVID-Vac. Of the 388 individuals at 4 weeks post vaccination, 61 (15.7%) were seropositive and therefore, they were excluded from the analysis of seroconversion rates. During this period of 4 weeks post vaccination, none of the baseline seronegative individuals reported a symptomatic infection. Of the 327 individuals who were seronegative at baseline, 203 (62.1%) were females. The mean age was 50.3 years (range 20 to 83 years). The demographic details and comorbidities of these individuals are shown in Supplementary Table 1.
The overall seroconversion rates following a single dose of the vaccine was 88.7% (95% CI 85.2 to 92.1%). Seroconversion rates and the median antibody titres (given as the antibody index) are shown in Table 1. The seroconversion rates were significantly lower in older individuals (≥ 60 years of age) when compared to younger individuals (20-39 years of age) (Kruskal Wallis Chi Square = 11.137, p-value = 0.004). The antibody titres, as indicated by the antibody index, were significantly different in different age groups (p = 0.008, Fig. 1A) and significantly and inversely (Spearman's r = −0.21, p = 0.0001) correlated with the age (Fig. 1B). The overall seroconversion rates for AZD1222 (93.2%) 14 were significantly higher (p = 0.018) compared to Gam-COVID-Vac (88.7%). The seroconversion rates in the ≥ 60 age group (81.8%) were similar (p = 0.97) to levels seen following a single dose of AZD1222 at 4 weeks (81.6%) 14 .

ACE2 receptor blocking antibodies at 4 weeks following a single dose of Gam-COVID-Vac. ACE2
receptor blocking antibodies were determined using a surrogate neutralizing antibody assay (sVNT) as previously described 15 . A positive cut off value of 25% was considered as positive response for this assay in the Sri Lankan population, as previously described by us 16 . We measured ACE2 receptor blocking antibodies at paired samples at baseline uninfected individuals (n = 69) and baseline infected individuals (n = 62), at 4 weeks following vaccination. The individuals who were included in these analyses, were those who consented to give an additional volume of blood (12 ml vs 5 ml) as a larger volume was required for all the antibody and T cell assays. 57/69 (82.6%) individuals who were baseline uninfected gave a positive response with median values of 42.17% (IQR 27.3 to 64.9% of inhibition) (Fig. 1C). In comparison, 66/68 (97.1%) who received a single dose of AZD1222 gave a positive response to sVNT assay, with median antibody levels of 69.42 (IQR 54.09 to 81.54% of inhibition) as previously published 13 . We also compared the ACE2 receptor blocking antibody levels with baseline antibodies of those who were found to be infected, and in another cohort of 36 individuals who had varying severity of natural infection 17 . While the ACE2 receptor antibody levels were similar in the two naturally infected cohorts, the ACE2 receptor blocking antibodies were significantly lower in the naïve individuals who received 1 dose of the Gam-COVID-Vac compared to baseline seropositives (p = 0.0003), and those who were naturally infected (p = 0.0009) (Fig. 1B). In contrast, those who had a single dose of AZD1222 had similar levels of ACE2 receptor blocking antibodies as those following natural infection (p = 0.7). Those who had a single dose of AZD (median 69.4, IQR 54.1 to 81.5) had significantly higher levels (p < 0.0001) than those who received a single dose of Gam-COVID-Vac.
Of the baseline infected individuals, 9/62 (14.5%) did not have ACE2 receptor antibodies at baseline. However, the ACE2 receptor blocking antibodies significantly increased (p < 0.0001) from baseline (median 67.1, IQR 43.9 to 85.4% of inhibition) to 4 weeks following the vaccine (median 98.7, IQR to 95.5 to 99.3% of inhibition (Fig. 1B). www.nature.com/scientificreports/ However, 5/9 individuals who were found to be previously infected (gave a positive response by the Wantai total antibody assay), failed to develop ACE2 receptor blocking antibodies following vaccination.
Antibodies to the RBD of the ancestral SARS-CoV-2 virus and variants of concern. We measured antibodies to the RBD of the ancestral variant (WT), B.1.1.7, B.1.351 and B.1.617.2 as previously described using the HAT assay 18 . A HAT titre of 1:20 was considered as the cut-off value for a positive response as previously described by us 19 . The assay was carried out on baseline uninfected individuals (n = 69) and baseline infected individuals (n = 62); these were the same individuals used to assess ACE2 receptor blocking antibodies.
The levels were compared with our previously published data 13 for those who received the AZD1222 at 4 weeks and additional assays were done for RBD antibodies for B.   The SARS-CoV-2 total antibody levels (antibody index values) were measured using the WANTAI SARS-CoV-2 Antibody ELISA assay in those between 20 to 39 years (n = 71), 40 to 59 years (n = 168) and 60 years (n = 88) (A) and the antibody levels were correlated with the age of individuals (n = 327), Spearman's r = −0.21, p = 0.0001 (B). ACE2 receptor blocking antibodies were measured by the surrogate neutralizing antibody assay in a sub-cohort of baseline uninfected individuals (n = 69, blue) at 4 weeks and in a cohort of baseline infected individuals (n = 62, red), who were found to be seropositive for SARS-CoV-2 at the time of recruitment (C). The differences in the total antibody titres between different age groups was determined by the Kruskal-Wallis test. The differences in ACE2 receptor blocking antibodies at baseline uninfected and infected individuals were assessed by the Wilcoxon matched pairs signed ranked test. All tests were two sided. The error bars indicate the median and the interquartile ranges.

Ex vivo IFNγ ELISpot responses and intracellular cytokine responses to Gam-COVID-Vac. IFNγ
ELISpot responses were done in 40 individuals (who were seronegative at baseline) at 4 weeks from receiving the first dose of Gam-COVID-Vac. Since the cut-off value of a positive response was considered to be mean ± 2SD, the threshold for a positive response was set at 220 spot forming units (SFUs)/1 million PBMCs. Accordingly, 12/40 (30%) had a positive response to S1 pool of peptides and 11/40 (27.5%) had a positive response to S2. The positive threshold for the AZD1222 was set at 302 SFU/1 million PBMCs. The proportion of individuals who responses to AZD1222 was higher for S1, with 42/68 (61.7%) of individuals S1 pool of peptides, while the responses to S1, 21/68 (30.8%) was similar.
The responses to S1 pool of peptides significantly increased (p < 0.0001) from baseline to 4 weeks (median 137.5, IQR 48.7 to 298.8 SFU/1 million PBMCs), while a significant increase (p = 0.23) was not seen for the S2 pool of peptides (median 87.5, IQR 41.3 to 240 SFU/1 million PBMCs) (Fig. 3A). The ELISpot responses for www.nature.com/scientificreports/ neither S1 nor S2 correlated with the age of the individuals. Individuals had significantly higher (p < 0.0001) ex vivo IFNγ ELISpot responses to the S1 pool of peptides following the AZD1222 13 compared to the first dose of Gam-COVID-Vac, and for the S2 pool of peptides (p = 0.01). Intracellular cytokine assays (ICS) was carried out in a subset of the above individuals (n = 32) in whom IFNγ ELISpot responses were carried out due to limitations in the number of cells. A relative high frequency of CD107a expressing CD4+ (median 0.069, IQR 0.022 to 0.18% of CD4+ T cells) and CD8+ T cells (median 0.021, IQR 0 to 0.045% of CD8+ T cells) were seen with CD4+ T cells expressing significantly more CD107a (p < 0.0001) (Fig. 3B). IFNγ production was also predominantly by CD4+ T cells (median CD4+ T cells 0.01 vs 0 from CD8+ T cells), which was significant (p = 0.001). Spike protein specific CD4+ T cells expressed significantly more CD107a (p = 0.007) than IFNγ. No comparison could be carried out with AZD1222 as we had not carried out ICS for those who received the first dose of AZD1222 at the 4 weeks' time point. Ex vivo IFNγ ELISpot assays were carried out at 4 weeks (individuals who were vaccinated from Gam-COVID-Vac) (n = 40) in baseline uninfected individuals for overlapping peptides of spike protein, which were in two pools S1 and S2 (A). Intracellular cytokine staining was used to determine CD107a expression and IFNγ production at 4 weeks (n = 32) in CD4+ and CD8+ by flow cytometry (B). The number of antibody secreting B cells (ASCs) were determined at 4 weeks (n = 25) by doing B cell ELISpot assays (C). Wilcoxon matched pairs signed ranked test was used to find out the differences in ex vivo ELISpot responses for S1 and S2, and differences in CD107a expression and IFNγ production by CD4+ and CD8+ T cells and for B cell ELISpot assays. All tests were two sided. The error bars indicate the median and the interquartile ranges. www.nature.com/scientificreports/ Memory B cell responses to the Gam-COVID-Vac. Antibody secreting memory B cell responses were assessed by B cell ELISpot assays for S1, S2 and N recombinant protein at 4 weeks (n = 25) following vaccination in baseline seronegative individuals. A positive response was defined as mean ± 2 SD of the background responses. Accordingly, a cut-off of 22.1 antibody secreting cells (ASCs)/1 million cells was considered as the positive threshold for S1 protein, 25.2 for S2 and 10.3 for N protein. 23/25 individuals had a positive response to S1, 11/25 and 19/25 gave a positive response to S2 and N protein respectively, in those who were baseline seronegative. However, the responses for S1 (median 60, IQR 45 to 72.5 ASCs/1 million cells) were significantly higher than (p = 0.0004) for S2 (median 25, IQR 15 to 45 ASCs/1 million cells) and significantly higher than (p < 0.0001) for N (median 20, IQR 12.5 to 30 ASCs/1 million cells) (Fig. 3C).

Discussion
We have investigated the immune responses of the first dose (rAd26-S) of the Gam-COVID-Vac, which is also known as Sputnik light, in a Sri Lankan population, 4 weeks after receiving the first dose and compared the immune responses with our previously published data following AZD1222, which is another adenovirus vector vaccine. The first dose of Gam-COVID-19 vaccine was shown to induce an overall seroconversion rate of 88.7%, with significantly lower seroconversion rates (81.8%) and SARS-CoV-2 specific antibody levels in individuals over 60 years of age. It induced ACE2-receptor blocking antibodies in 82.6% of individuals and the levels were significantly lower than levels seen following a single dose of AZD1222 and those following natural infection. Therefore, the overall antibody responses following a single dose of Gam-COVID-Vac appear to be significantly lower than those following a single dose of AZD1222, especially in older individuals in a Sri Lankan population 13 . Older individuals are known to have impaired innate and adaptive immune responses to new antigens due to multiple factors such as compromised antigen presentation, T cell exhaustion and a limited B cell receptor repertoire in memory B cells leading to development of antibodies with poor affinity and avidity 20 . Therefore, due to the reduced immunogenicity of Gam-COVID-Vac, and due to the rapid decline in immune responses to some COVID-19 vaccines 21 , different dosing schedules and booster dosing schedules may have to be implemented in older individuals.
In the phase 1/2 studies, it was shown that the vaccine only induced neutralizing antibodies (Nabs) in 61.1% of individuals at 28 days although the antibody titres were comparable to those following natural infection 6 . Although the ACE2 receptor blocking antibodies are not a direct measure of Nabs, they have shown to be a surrogate indicator of Nabs 15,22 . Nabs have been shown to be highly predictive of protection against SARS-CoV-2 infection 23 , and those with lower Nabs have shown to associate with breakthrough infection in vaccinated health care workers 24 . Although the AZD1222 reported an efficacy rate of 76% up to 90 days following a single dose of the vaccine for the ancestral SARS-CoV-2 virus 25 , the efficacy of a single dose reduced to 30% for B.1.617.2 26 . Therefore, given that the ACE2 receptor blocking antibody levels were significantly lower than following a single dose of AZD1222 and since the proportion of individuals who developed ACE2 receptor blocking antibodies were also significantly lower, especially in the older age groups. A single dose of Gam-COVID-Vac may possibly have a reduced efficacy than AZD1222 in Sri Lanka, and so may benefit from administration of the second dose.
At the end of 4 weeks following the single dose of Gam-COVID-Vac, 81.2% had antibodies to the RBD of the WT detected by HAT, similar to levels seen following a single dose of AZD1222 13 . However, the proportion of individuals who developed antibodies were less than those reported in the phase 1/2 study, in which 100% of the study participants developed antibodies to the RBD by 21 days 6 . These differences could be due to the differences in the assays used, ages of the individuals included (we had included individuals > 60 years of age), sample size (n = 18 in the published study compared to n = 69 in this study) and possible pre-existing antibodies to Ad26 reducing vaccine immunogenicity in Sri Lanka 27 . The lower RBD antibodies to the WT and ACE2 receptor blocking antibodies in our cohort of patients, especially in those who were > 60 years of age, in comparison to the published data, could be due to the differences in the seropositivity rates for adenovirus 26 in different age groups in different populations 12 . It was previously shown that individuals who were seropositive for adenovirus 5, had a higher risk of acquiring HIV following immunization with an HIV vaccine with adenovirus 5 28 . This increased risk of acquiring HIV was thought to be due to reduced HIV specific CD4+ T cell responses and a reduced breadth of CD8+ T cell responses, in vaccine recipients who were seropositive for adenovirus 5 29 . As pre-existing anti-vector immunity is likely to dampen the immune responses to vaccines, it would be important to study the seroprevalence of antibodies to the adenovirus vector used in such vaccines, in order to further understand the effect of a vaccine when rolling it out in a population.
The antibody responses to the RBD of the WT, B.1.1.7 and B.1.617.2 following the first dose of Gam-COVID-Vac were similar to that of AZD1222, but significantly lower than antibody levels seen following natural infection, although antibodies to the RBD of B.1.351 were significantly higher than AZD1222. As AZD1222 was shown to have a reduced efficacy for B.1.351 30 , given the higher magnitude and frequency of antibodies to the RBD of B.1.351, Gam-COVID-Vac may be more effective than AZD1222 in countries, which have outbreaks due to B.1.351, although further clinical studies are required. However, since many of the SARS-CoV-2 variants such as B.1.1.7 and B.1.351 are on the decline, the immune responses to the new emerging sub lineages of delta should be continuously monitored.
Although Nabs are thought to be correlated with prevention of breakthrough infections following vaccination and an important correlate of protection 23 , a robust T cell response is also associated with development of milder clinical disease 31 . Gam-COVID-Vac induced a high frequency of ex vivo IFNγ ELISpot responses to spike protein peptides in 30% of individuals, and a high frequency of CD107a producing CD4+ and CD8+ T cells along with memory B cell responses. Interestingly, both CD107a and IFNγ production was predominantly by the CD4+ T cell subset of T cells, which is similar to what was seen following a single dose of the AZD1222 32  www.nature.com/scientificreports/ the spike protein overlapping pool of peptides was significantly lower than for AZD1222 13 . The Gam-COVID-Vac induced ASCs, responding to S1, S2 and N recombinant proteins. Although 19/25 individuals responded to the N protein of SARS-CoV-2, these responses were of very low frequency and were several folds lower than the responses to S1 subunit. The frequency of N protein specific ASCs were also several fold lower than what we previously observed for the Sinopharm/BBIBP-CorV, which is an inactivated vaccine and therefore, induces immune responses to the N protein 17 . Therefore, these low frequency memory T cell responses are likely to be due to pre-existing cross reactive memory B cell responses to other seasonal coronaviruses and are unlikely to be SARS-CoV-2 specific 33 .
In summary, the Gam-COVID-19 first dose (rAd26-S) induced seroconversion rates in 88.7% of individuals 4 weeks following the vaccine, with significantly lower seroconversion rates in the elderly in a Sri Lankan population. ACE2 receptor blocking antibody responses were seen in 82.6% of individuals, with levels significantly lower than after a single dose of AZD1222 and following natural infection. While the antibody responses to the RBD by HAT were similar to AZD1222, the ex vivo IFNγ ELISpot responses were significantly lower. The findings suggest that the two Gam-COVID-19 doses using heterologous adenoviruses may be of particular importance in certain populations. This data further highlights the importance of studying the prevalence of different adenoviruses and how pre-existing immunity affects the immune responses to different adenovirus vector vaccines. This information is viral in order to determine timing and targeting additional doses of COVID-19 vaccines.

Methods
Study participants. 388 adult individuals from Sri Lanka, were recruited at the time they received the first dose of the vaccine following informed written consent. Blood samples were obtained at baseline (at the time of recruitment) to determine SARS-CoV2 seropositivity, and at 4 weeks from receiving the first dose of the vaccine. Individuals who were between the ages of 20 to 39 years were considered as younger individuals, while those who were > 60 years of age, were considered as older individuals. The presence of comorbid illness such as diabetes, hypertension was recorded. The SARS-CoV-2 specific antibody assays were carried out in all the individuals, while certain assays were done in a sub cohort (see below). We also compared the ACE2 receptor blocking antibodies, antibodies to the RBD of different variants detected by HAT and ex vivo IFNγ ELISpot responses in individuals 4 weeks after a single dose of AZD1222 (Covishield) vaccine (n = 69) 13 , and with those who were naturally infected (n = 36) as previously reported by us 17 . Overall seroconversion rates for Gam-COVID-Vac first dose was also compared with AZ first dose at 4 weeks, using our previously published data 14 .
Ethics approval was obtained from the Ethics Review Committee of University of Sri Jayewardenepura (COVID 01/21). All research was performed in accordance with relevant guidelines/regulations, and informed written consent was obtained from all participants.

SARS-CoV-2 specific total antibodies, ACE-2 receptor blocking antibodies and antibodies to the RBD of SARS-CoV2 variants.
Antibodies to SARS-CoV-2 were detected by Wantai SARS-CoV-2 antibody ELISA (Beijing Wantai Biological Pharmacy Enterprise, China), which detects IgM, IgG and IgA antibodies. Surrogate virus neutralization test (sVNT) 15 was used to detect ACE2 receptor blocking antibodies as previously described 16 . Inhibition percentage ≥ 25% in a sample was considered as positive for ACE2 blocking antibodies. This assay was found to be 100% specific for measuring ACE2 blocking antibodies in the Sri Lankan population 16 .

Haemagglutination tests for detection of antibodies to the RBD in WT and SARS-CoV-2 variants.
For the HATs, sera were doubling-diluted in 50 μl PBS in V bottomed 96 well plates, 50 μl of ~ 1% v/v O-ve red cells were added, followed by 50 μl of the relevant IH4-RBD reagent diluted to 2 μg/ml (100 ng/well). Plates were incubated for 1 h at RT, tilted for ~ 20 s to allow a red cell "teardrop" to form, photographed and read by eye. The RBD-specific antibody titre for the serum sample was defined by the last well in which the complete absence of "teardrop" formation was observed. The HAT titration was performed using 7 doubling dilutions of serum from 1:20 to 1:1280, to determine presence of RBD-specific antibodies. A titre of 1:20 was considered as a positive response, as previously determined by us 19 . The IH4-RBD reagents for each VOC were standardized by titration with the monoclonal antibody EY-6A 18,34 that binds to a conserved epitope common to all variants. A 20 μg/ml solution of EY-6A titrated equally with a standard (2 μg/ml) solution of each of the new IH4-RBD reagents. All were therefore added as 50 μl from a 2 μg/ml stock solution (100 ng/well) as described 18 .
Ex vivo ELISpot assays. Ex vivo IFNγ ELISpot assays were carried out using freshly isolated peripheral blood mononuclear cells (PBMC). Two pools of overlapping peptides named S1 (peptide 1 to 130) and S2 (peptide 131 to 253) covering the whole spike protein (253 overlapping peptides) were added at a final concentration of 10 µM and incubated overnight as previously described 35,36 . 100,000 cells/well were added, PHA was included as a positive control of cytokine stimulation and media alone was applied to the PBMCs as a negative control. All peptide sequences were derived from the wild-type consensus and were tested in duplicate.
Briefly, ELISpot plates (Millipore Corp., Bedford, USA) were coated with anti-human IFNγ antibody overnight (Mabtech, Sweden). The plates were incubated overnight at 37 °C and 5% CO 2 . The cells were removed, and the plates developed with a second biotinylated Ab to human IFNγ and washed a further six times. The plates were developed with streptavidin-alkaline phosphatase (Mabtech AB) and colorimetric substrate. The spots were enumerated using an automated ELISpot reader (AID Germany). Background (PBMCs plus media alone) was subtracted and data expressed as number of spot-forming units (SFU) per 10 6 PBMCs. A positive response was defined as mean ± 2 SD of the background responses. An example of a ex vivo ELISpot assay for S1, S2, is shown in Supplementary Figure 1.  Figure 2). The proportion of cells expressing S pool of peptide specific CD107a or producing IFNγ, was determined by subtracting the expression levels/production levels in the unstimulated wells, from the peptide stimulated wells.
B cell ELISpot assays. Briefly, freshly isolated PBMCs were stimulated in a 24 well plate using IL-2 and R848 (a TLR 7/8 agonist) in RPMI supplemented with 10% fetal bovine serum, 1% penicillin streptomycin and 1% glutamine at 4 million cells/well and incubated at 37 °C with 5% CO 2 for 3 days. They were then washed and rested overnight and 100,000 cells/well were added. 50,000 cells/well were added to the positive control wells. A Human IgG ELISpot kit (Mabtech 3850-2A) was used according to the manufacturer's instructions to quantify IgG-secreting cells specific to SARS-COV2 S1, S2 and N recombinant proteins, which were coated at 2 µg/ml in phosphate buffered saline (PBS). All experiments were carried out in duplicate and anti-human IgG monoclonal capture antibodies, was used as a positive control, and media alone as a negative control. The spots were enumerated using an automated ELISpot reader (AID Germany). A positive response was defined as mean ± 2 SD of the background responses. An example of a B cell ELISpot assay for S1, S2 and N recombinant proteins is shown in Supplementary Figure 3.
Statistical analysis. 95% confidence intervals for each category were calculated using the R software (version 4.0.3) and R-studio (version 1.4.1106). Pearson Chi Square Association tests were performed at a confidence level of 95% using the R software in order to identify the statistically significant associations of the age categories and the sex of study participants with antibody responses at 4 weeks. GraphPad Prism version 6 was used for other statistical analysis. In instances when the data were not paired, the Mann-Whitney U test (two tailed) was used and the Wilcoxon matched-pairs signed rank test was used when comparing paired data. The Kruskal-Wallis test was used to compare the differences of the antibody levels in different age groups. Spearman rank order correlation coefficient was used to evaluate the correlation between variables including the association between RBD antibodies, SARS-CoV-2 antibodies, SARS-CoV-2-specific T cell responses and age. www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.