Effects on rumen microbiome and milk quality of dairy cows fed a grass silage-based diet supplemented with the macroalga Asparagopsis taxiformis

The objective was to determine the effects on rumen microbiome and milk quality of reducing the methane (CH4) produced from enteric fermentation by the addition of Asparagopsis taxiformis (AT) to the diets of dairy cows. Six Nordic Red cows at 122 ± 13.7 (mean ± SD) days in milk, of parity 2.7 ± 0.52 and producing 36 kg ± 2.5 kg milk per day at the start of the trial were divided into three blocks by milk yield and assigned to an extra-period Latin-square change-over design comprising two dietary treatments. An extra period of observation was added to the Latin-square change-over design to control for carry-over effects. The dietary treatments were a diet consisting of grass silage and a commercial concentrate mixture (60:40) either not supplemented or supplemented with 0.5% AT on an organic matter intake basis. On average, daily CH4 production, CH4 yield, and CH4 intensity decreased by 60%, 54%, and 58%, respectively, in cows fed the diet supplemented with AT. Furthermore, hydrogen gas emitted by cows fed diets supplemented with AT increased by more than five times compared with cows fed a non-AT-supplemented diet. Feed intake was decreased and milk production altered, reflecting a decreased yield of milk fat in cows fed an AT-supplemented diet, but feed efficiency increased. Rumen fermentation parameters were changed to promote propionate rather than acetate and butyrate fermentation. The most prominent change in milk quality was an increase in bromine and iodine when the diet was supplemented with AT. The reduction in CH4 was associated with a shift from Methanobrevibacter to Methanomethylophilaceae in the archaeal population and a lower relative abundance of Prevotella in the bacterial population. Changes in milk fat odd-numbered and branched-chain fatty acids in the current study of AT supplementation support observed differences in ruminal archaeal and bacterial populations.


Introduction
The climate commitments made in the Paris Agreement aim to protect planet habitability by limiting global warming to 1.5°C above the pre-industrial baseline (European Commission, 2022). The environmental goals of carbon neutrality are directly related to the increased concentrations of greenhouse gases (GHGs) in the atmosphere (IPCC, 2021). Food systems contribute up to 30% of global GHG emissions (Clark et al., 2020). Emissions of methane (CH 4 ) from ruminant livestock (via enteric fermentation and anaerobic manure storage) account for 33% of all anthropogenic CH 4 sources and represent the second greatest anthropogenic CH 4 source after fossil fuels (Jackson et al., 2020;Crippa et al., 2021). Methane in the atmosphere produces around 28 times more heat on a 100-year time scale than carbon dioxide (CO 2 ). However, when CH 4 decomposes to CO 2 in the atmosphere, it will be bound to vegetation in the near future and will not contribute to further warming. This implies that reducing CH 4 emissions will have an immediate effect on the global temperature (Sterner and Johansson, 2017). Demands on food production are increasing, and the global cattle population is predicted to increase as agricultural land in the world becomes dominated by pastures and permanent meadows used for livestock (FAO, 2019).
Short-term and efficient reductions in enteric emissions of CH 4 from ruminants primarily involve feed manipulation mitigation strategies. Only a few of these feed manipulation strategies have been shown to be efficient and applicable in real farming situations. The inhibitor 3-nitroxypropanol has demonstrated CH 4 reductions of only approximately 30%, when given in very low doses to dairy cows (Hristov et al., 2015). Recently, the application of macroalgae as antimethanogenic feed additives has been thoroughly reviewed by Abbott et al. (2020). Despite green and brown macroalgae potentially possessing CH 4 -mitigating effects, which can be attributed to several bioactive compounds (e.g., peptides, carbohydrates, lipids, and phlorotannins), the effects have been small in vitro (Krizsan et al., 2022) and are likely negligible in vivo. The tropical macroalga Asparagopsis taxiformis has recently shown the greatest potential CH 4 -reducing effect, amounting to an almost 100% reduction in CH 4 production in vitro (Chagas et al., 2019), between a 55% and 65% reduction in vivo when added to the diet of dairy cows (Stefenoni et al., 2021), and at most a 98% reduction in vivo when added to the diet of steers (Roque et al., 2021), all at an inclusion level of 0.5% of organic matter (OM). A. taxiformis is a natural supplement, and the inhibitory effect is caused by the volatile compound bromoform in the last step of methanogenesis by ruminal archaea (Machado et al., 2016). Chagas et al. (2019) suggested a strong dose-dependent CH 4 -mitigating effect of A. taxiformis at the inclusion level of 0.5% of OM with the least impact on rumen fermentation parameters.
This study aimed to assess the effects on rumen microbiome and milk quality when dairy cow diets are supplemented with A. taxiformis at the inclusion level of 0.5% of OM to achieve an optimal antimethanogenic effect in dairy cows. Additionally, the study contributes to our understanding of the effects on dairy cow production performance and metabolism when applying A. taxiformis as a feed manipulation CH 4 mitigation strategy.

Materials and methods
The feeding trial was conducted at Röbäcksdalen experimental farm of the Swedish University of Agricultural Sciences in Umeå (63°4 5′N, 20°17′E) from 8 October to 9 December 2020.

Cows, experimental design, and diets
Six Nordic Red dairy cows weighing (mean ± SD) 611 kg ± 62.1 kg, being 122 ± 13.7 days in milk, of parity 2.7 ± 0.52, and producing 36.3 kg ± 2.47 kg of milk per day pretrial, were blocked by milk yield (MY) and used in the experiment. The cows were kept in an insulated free-stall barn, offered a total mixed ration (TMR) ad libitum, mixed with a feed mixer (Nolan A/S, Viborg, Denmark), and delivered with a Mullerup M2000 Free Feeder automatic feeding wagon (GEA Farm Technologies Mullerup A/S, Ullerslev, Denmark) in individual feed troughs three times per day at 07:00, 13:00, and 20:00. The cows had free access to drinking water and were milked twice per day, at 06:00 and 16:30, throughout the experiment.
The cows were randomly allocated to a dietary treatment within block, i.e., square, and assigned to an extra-period Latin-square change-over design consisting of three equal squares (Lucas, 1957). The experimental periods lasted 21 days, with the last 7 days used for data recording and sampling. All cows were fed a control diet composed of 600 g/kg dry matter (DM) of grass silage, 390 g/kg DM of a commercial concentrate (Komplett Amin 180; Lantmännen Lantbruk AB, Malmö, Sweden), and 10 g/kg DM of mineral mix (Mixa Optimal; Lantmännen Lantbruk AB, Malmö, Sweden). Dietary treatments comprised either no supplementation (control) or supplementation with the macroalga Asparagopsis taxiformis (AT) at 0.5% of organic matter (OM) intake. The daily supplementation of AT was in line with an earlier dose-response in vitro experiment conducted by Chagas et al. (2019).
The grass silage was harvested as a first cut at Röbäcksdalen experimental farm on 13 June 2020 from a second year ley of timothy grass (Phleum pratense) and some red clover (Trifolium pratense) (seed ratio 80:20; botanical analysis not made). The grass was cut with a mower conditioner (GMT 3605 FlexP; JF-Stoll A/S, Sønderborg, Denmark), wilted overnight, and harvested with a precision chop forage wagon (ES 5000 MetaQ ProTec; JF-Stoll A/S, Sønderborg, Denmark). A formic acid-based additive (ProMyr XR 630; Perstorp Holding AB, Perstorp, Sweden) was applied at a rate of approximately 3.5 L/Mg and the forage material was ensiled in bunker silos until use in the experiment.
Asparagopsis taxiformis in the gametophyte life cycle stage was harvested at Angú stias, Faial Island, Azores, Portugal (38°31′45″N, 28°37′09″W), between April and May 2020 by SeaExperts (Feteira, Ilha Do Faial, Portugal). Biomass was collected at a depth of 3 to 6 m by scuba divers and stored in a dark cooled container in a supporting boat before reaching land, where it was rinsed and immediately frozen and stored at -25°C. The frozen material was shipped to European Freeze Dry ApS (Kirke Hyllinge, Denmark) and the material was freeze-dried. The dried AT was vacuum stored in light-excluding bags and transported to the Swedish University of Agricultural Sciences in Umeå. During the experiment, the bags containing AT were stored in a dark basement storage holding room temperature, and bags were brought up and processed successively. The freeze-dried macroalga was ground using an electric hand mixer before being manually mixed with the TMR in the individual feedbunks (60% of the total daily AT dose at 07:00 and 40% of the total daily AT dose at 13:00). The delivery of the TMR was monitored and adjusted in the afternoon daily to avoid feed residues and achieve the intended provision of AT to each cow daily.

Data recording, sampling, and calculations
Individual feed intake was recorded daily by Roughage Intake Control feeders (Insentec B.V., Marknesse, The Netherlands) and daily MY was recorded using a gravimetric milk recorder (SAC, S.A. Christensen and Co Ltd, Kolding, Denmark). The body weight (BW) of the cows was measured after morning milking for 3 days before the experiment and for the last 3 days of each period.
Mass fluxes of CH 4 , carbon dioxide (CO 2 ), and oxygen (O 2 ) were recorded daily by the GreenFeed emission monitoring (GEM) system (GreenFeed; C-Lock., Rapid City, SD, USA) as described by Huhtanen et al. (2015). Gas calibrations (N 2 and a mixture of CH 4 , O 2 , and CO 2 ) were performed once a week and CO 2 recovery tests were conducted every second week throughout the entire experiment. The average recovery (mean ± SE) was 100% ± 0.6%. The air filters of the GEM units were cleaned twice a week in order to maintain the airflow above 26 L/s. The same commercial concentrate (Komplett Amin 180; Lantmännen Lantbruk AB, Malmö, Sweden) as in the TMR was given to the cows in order to ensure regular visits by the cows to the GEM system. The program was set to allow each cow to visit at a minimum of 5-h intervals and to give eight drops of 50 g of concentrate during each visit. The interval between drops was set at 40 s.
The DM concentration was determined twice a week for the silages and once a week for the concentrates. All feed samples were oven-dried at 60°C for 48 h. The diets were adjusted twice weekly to account for changes in DM content. The dried feed samples were ground to pass through a 1-mm sieve in a cutter mill (SM 2000; Retsch Ltd., Haan, Germany) to provide samples for chemical analysis, or through a 2-mm sieve to provide samples for indigestible neutral detergent fiber (iNDF) analysis. Additionally, silages were sampled for 3 days in the last week of each period to obtain a composite sample for fermentation quality analysis. The silage samples were stored frozen at -20°C before and after grinding. Frozen silage samples were ground in a cutter mill (SM 2000) to pass through a 20-mm sieve.
Milk samples were collected during days 19-21. Pooled milk samples were made on day 19, weighted according to MY from morning and afternoon milking, for sensory testing and bromoform analysis. The samples for sensory testing were stored at 5°C in glass bottles that were provided by the dairy co-operative (Norrmejerier, Umeå, Sweden). The sensory analysis was conducted by the same two trained persons throughout the whole trial directly after the last milk sample collection. Sensory taste was compared with an equal test of the milk from cows in the experiment before the start of the experimental diets. The milk sample for bromoform analysis was stored frozen at -20°C until analysis. Samples for analysis of fat, protein and lactose, and urea composition of the milk were taken during four consecutive milking occasions in each period, and stored in plastic bottles with preservative (Bronopol) at 5°C until analysis. From the same four milking occasions a pooled sample of milk, weighted according to morning and afternoon MY, was made for each cow in each period for analysis of fatty acid composition, and stored at -20°C until analysis.
Rumen fluid was collected from all cows on day 19 in each period. The collection and preparation of rumen fluid was carried out in accordance with Chagas et al. (2021). In addition, 5-mL subsamples of filtered rumen fluid were transferred to Falcon tubes and preserved with 0.5 mL of sulfuric acid for ammonia determination and 1 mL of metaphosforic acid (25%) for volatile fatty acid (VFA) analysis. Samples were immediately frozen on dry ice and kept at -80°C in a freezer until analysis.
Fecal spot samples (approximately 250 g) from all cows were collected in connection with morning and afternoon milking on days 19-21 in each period. After the last collection of each period, the fecal samples were oven dried at 60°C for 48 h and milled to pass through a 1-mm sieve in a cutter mill. Fecal samples used for the analysis of iNDF were milled using mortar and pestle to pass through a 2.5-mm sieve. Fecal samples were pooled by cow within period.
The chemical composition of diets was calculated based on the intake, dietary ingredient composition determined from fresh weight proportions, and ingredient chemical composition. Total intake was calculated as TMR intake plus concentrate intake from the GEM system. The apparent digestibility of nutrients was calculated using iNDF as an internal marker in feeds and feces (Huhtanen et al., 1994). Potentially digestible neutral detergent fiber (pdNDF) was calculated as neutral detergent fiber (NDF) -iNDF. The metabolizable energy content was calculated based on coefficients from feed tables (LUKE, 2019). Milk constituent concentrations were calculated as a weighted mean of the combined morning and afternoon milk yields. Daily energy-corrected milk (ECM) yield was calculated in accordance with Sjaunja (1990). Feed efficiency was calculated as the daily yield of ECM/daily amount of dry matter intake (DMI), and milk nitrogen efficiency (MNE) as the ratio of nitrogen (N) milk yield in grams to N intake in kilograms.

Compositional analysis
The DM, ash, and NDF concentrations of feed and fecal samples, and fermentation quality of silages, were analyzed as described by Pang et al. (2021). Total N in the samples was analyzed with the Kjeldahl method (Nordic Committee on Food Analysis, 1976), degrading the sample in a 2520 Digestor adding Kjeltabs CK (Foss Analytical A/S Hilleröd, Denmark), and distilling and titrating using a Kjeltec 8400 Analyser unit equipped with an 8460 sampler unit (Foss Analytical A/S Hilleröd, Denmark). Crude protein (CP) concentration was calculated as N × 6.25. The concentration of indigestible NDF (iNDF) in feed ingredients and feces was analyzed as described by Krizsan et al. (2015).

Microbial analysis
DNA isolation and 16S rRNA gene amplicon sequencing Bacterial library preparation and sequencing were performed by Novogene with total DNA extracted using TIANamp Stool DNA Kit and primer pair 515F (GTGBCAGCMGCC GCGGTAA)/805R (GACTACHVGGGTATCTAATCC). For the archaeal library, total DNA was extracted with the FastDNA ™ Spin kit for soil (MP Biomedicals) with an additional wash step with 500 mL of 5.5 M guanidine thiocyanate (Sigma-Aldrich, Missouri, USA) for humic acid removal (Singh, 2020). A two-step PCR protocol with the primer pair 516F (TGYCAGCCGCCGCGGTAAHACCVGC)/915R (GTGCTCCCCCGCCAATTCCT) was used as described by Müller et al. (2016). Archaeal amplicon sequencing was performed using the SNP&SEQ Platform (National Genomics Infrastructure, Uppsala, Sweden). Illumina MiSeq sequencing with v3 chemistry and paired end reads (2 × 300 bp) was used for both bacterial and archaeal amplicon libraries.

Statistical analysis
Experimental data were subjected to analysis of variance using the generalized linear model (GLM) procedure in SAS (SAS Inc. 2002-2003.4 SAS Inst., Inc., Cary, NC, USA) by applying the following model: where Y ijkl is the dependent variable, μ is the mean of all observations, P j is the fixed effect of period i, C j (S) k is the fixed effect of cow j within square k, T l is the fixed effect of experimental dietary treatment l, and ϵ ijkl~N (0, s 2 e ) is the random residual error. Potential carry-over effects on all traits were evaluated by including a residual effect in the model described above. There were no significant (p ≥ 0.22) residual effects for any production traits evaluated, and the effect was not included in the final analysis. The residual effect was only significant (P = 0.05) for cis-11-octadecenoic acid (cis-vaccenic acid) among all milk fatty acids analyzed.

Ingredient and diet composition
The grass silage displayed a low DM content and high lactic acid concentration, but was well preserved, with low pH and low levels of ammonia-N and butyric acid ( Table 1). The concentrations of metabolizable energy and indigestible NDF indicated a relatively early harvest, and grass silage CP was high. The grass silage was fed with a commercial concentrate product, which resulted in a diet composition reflecting the high CP of both feed ingredients (Table 2).

Intake, milk production, and digestibility
Total intake of DM and silage decreased (p< 0.01) by 2.8 and 1.8 kg per day during the experiment for the cows that were supplemented with AT (Table 3). Yield of ECM decreased (P = 0.05) by 1.9 kg per day as a result of decreased (p = 0.03) fat yield for cows fed AT. However, the most prominent change in quality composition was the far greater (p< 0.01) concentrations of bromine (more than an eightfold increase) and iodine (more than a 15-fold increase) in the milk of the cows that were fed antimethanogenic diets with AT. Generally, feed efficiency increased (p< 0.01) and cows were smaller (p = 0.02) when their diet was supplemented with AT. The sensory panel could not detect any odd flavors or changes in sensory quality from pre-trial milk samples in the milk samples from cows fed diets with AT (results not presented). Feeding AT had no effect on the total concentration of saturated fatty acids (SFAs) or monounsaturated fatty acids in milk fat, averaging 74 and 22 g per 100 g of fatty acids, respectively (Supplementary Table S1). However, the proportion of several short-and medium-chain SFAs of 6-12 carbon atoms in milk fat was increased by AT supplementation (p ≤ 0.01). Feeding AT also affected the concentration of several odd-numbered and branchedchain FAs of 5-15 carbon atoms (p< 0.05). Treatment had no effect on milk fat cis-9 18:1, 18:2n-6, or 18:3n-3 concentrations, but AT enriched trans-11 18:1, cis-9,trans-11 18:2, and trans-11, cis-15 18:2 in milk fat at the expense of 18:0 (p = 0.01).
Daily intake of OM and NDF was 2.6 kg and 1.2 kg lower, respectively (p< 0.01), when cows were supplemented AT, but no measures of total tract diet digestibility were affected by the antimethanogenic feed supplementation (Table 4). There were no differences (p = 0.19) in fecal bromoform concentration between the cows fed diets supplemented with AT and those fed nonsupplemented diets (results not presented).
( Table 6) Total VFA decreased (p = 0.03) when AT was included in the diets. Molar proportions of individual VFAs were altered (p ≤  Determined in situ according to Krizsan et al. (2015). 6 Potentially digestible NDF (NDF -iNDF). DM, dry matter; iNDF, indigestible neutral detergent fiber; NDF, neutral detergent fiber; pdNDF, potentially digestible neutral detergent fiber. 0.04) toward more propionate and valerate, and less acetate and butyrate for cows fed AT-supplemented diets. All rumen fluid samples were all below detection level (0.09 ug bromoform/g sample; results not presented).

Microbial data
The structure of the rumen archaeal and bacterial community in the dairy cows was characterised by sequencing the V4 region of archaeal and bacterial 16S rRNA gene, respectively. After quality control of raw data and filtering of untargeted taxa ("NA" at kingdom level, bacterial reads from archaeal data and vice versa) a total of 362,798 and 1,896,528 readsrepresented the 1,349 and 8,139 ASVs of archaeal and bacterial community, respectively.

Archaea
The archaeal community was represented by two different phyla, Thermoplasmatota and Methanobacteriota, where Thermoplasmatota and Methanobacteriota represented on average 53% ± 32% (mean ± SE) and 44% ± 33% in relative abundance, respectively. Orders were represented by Methanomassiliicoccales (53% ± 33%) and Methanobacteriales (44% ± 33%). Methanobacteriota was represented only by the genus Methanobrevibacter and Thermoplasmatota mainly by unclassified Methanomethylophilaceae with genus identity UBA71 (43% ± 29%) and MX-02 (6.0% ± 10%). Analysis using the GLM procedure to evaluate any differences between archaeal sequences was performed and revealed no significant differences at phyla or genus level. The AT-treated group had a numerically lower relative abundance of Methanobacteriota (29.9%) than the control group (58.8%) with a tendency to significance [p = 0.05, standard error of the mean (SEM) = 8.80] (Figure 1). Thermoplasmatota was numerically higher in the AT group (66.2%) than in the control group (40.3%), with a tendency to significance (p = 0.08, SEM = 9.10). At genus level, Methanobrevibacter was numerically higher in the control group (p = 0.05, SEM = 8.87), and unclassified UBA 71 belonging to the family Methanomethylophilaceae had a tendency of difference in higher relative abundance for AT treatment (p = 0.09). Differences at species level between treatments were observed only in species UBA71_sp900767505_RS_GCF_9007675 (Table 7), which was mainly present in the AT group, and there was a tendency (p = 0.06) for the relative abundance of the species Methanobrevibacter_A_sp001548675 to be higher in the control group.
The 10 most abundant phyla accounted for > 98% of the sequences generated from the bacterial communities of both the control and AT-supplemented groups. Analysis using the GLM procedure revealed statistically significant differences in the relative abundance of Bacteroidota (p = 0.04, SEM = 3.35), with a higher abundance in the control treatment group than in the AT treatment group, with abundances of 45.6% and 33.9%, respectively ( Figure 2). Firmicutes had a tendency to difference (p = 0.07, SEM = 6.65), with a numerically higher abundance in the AT group than in the control group, with abundances of 54.5% and 46.5% respectively. Proteobacteria was more abundant in the AT group (5.1%) than in the control group (3.1%) (p = 0.04, SEM = 0.58).

Discussion
In line with the results of this study, several in vivo experiments have confirmed the considerable potential of AT as an antimethanogenic feed supplement across several ruminant species (Li et al., 2016;Muizelaar et al., 2021;Roque et al., 2021;Stefenoni et al., 2021). However, studies investigating the effect of feeding AT to Least-squares means of relative abundance of archaea at phylum level for cows with a diet without (control) or with supplementation with Asparagopsis taxiformis (AT). NA, not applicable. cows on the rumen microbial community are rare. It is known that Prevotella spp. may redirect excess H 2 to propionate when CH 4 is inhibited in the rumen (Aguilar-Marin et al., 2020). However, propionate levels were reduced by AT supplementation, which is in line with the inhibition of Prevotella in this study. Propionate and acetate are both fermentation products of Prevotella, which makes the relationship between enteric CH 4 emission and Prevotella more complex to interpret. To obtain a holistic understanding of the effect on the microbial population, metagenomics or metatranscriptomics studies are required. Stefenoni et al. (2021) reported decreased feed intake, milk yield, and yield of milk fat and protein when dairy cows' diets were supplemented with AT at the same inclusion level as in this study. The induced changes in milk fat yield in this study could be explained by the altered rumen fermentation pattern, promoting lower proportions of lipogenic VFA. Milk fatty acid composition was similar to previous reports on grass silage-based diets (Lindmark Månsson, 2008). However, the enrichment of the biohydrogenation intermediates of 18-carbon unsaturates in milk fat at the expense of the final product, 18:0 (Shingfield et al., 2010), suggests that ruminal biohydrogenation of feed fatty acids was less complete in cows fed AT. Milk fat odd-numbered and branched-chain fatty acids are largely synthesized by rumen microbiota and reflect their relative abundance (Vlaeminck et al., 2006). Subtle but systematic changes in milk fat odd-numbered and branched-chain fatty acids in the current study of AT supplementation support observed differences in ruminal archaeal population.
The most prominent change in milk quality was the increase in bromine and iodine when the cows' diets were supplemented with AT, in line with the results of Stefenoni et al. (2021). Stefenoni et al. (2021) observed an increase in bromine of equal magnitude in the milk of cows given AT supplementation and those not given AT supplementation, but a much lower change in milk iodine than in our study. However, in absolute values, Stefenoni et al. (2021) reported a concentration of milk bromine comparable to that in our study, but a far higher concentration of milk iodine. Excess consumption of iodine by humans can potentially cause thyroid dysfunction (Leung and Braverman, 2014), but harmful amounts would depend on total milk consumption, assuming that all retail milk was from AT supplemented cows (Antaya et al., 2015), and on age (Zimmermann et al., 2005). The recommended daily intake (RDA) of iodine differs according to age, but in general ranges from 90 to 150 mg/day, with additional intake recommended for those who are pregnant or lactating (Russell et al., 2001;EFSA, 2017). The upper limit for iodine consumption varies from two to four times higher than the RDA, i.e., between 200 and 600 mg per day (EFSA, 2017). From the results of our study, intake of a cup of milk (225 mL would provide approximately 470 mg of iodine per day, which would exceed the upper limit for children aged up to either 8 years (Russell et al., 2001) or 14 years (EFSA, 2017). For bromine, there are fewer recommendations available. However, in a study by Van Leeuwen et al. (1987), the provisional acceptable daily intake for bromine was 0.4 mg/kg body weight. This means that if milk contained 43.2 mg/L of bromine, as in our study, a child that weighs 10 kg should not exceed a daily intake of 10 mL of milk while an adult weighing 60 kg could consume 0.5 L of milk without exceeding the acceptable daily intake.

Data availability statement
The datasets presented in this study can be found in online repositories. The name of the repository and accession numbers can be found below: NCBI; PRJNA911204 and PRJNA911215.

Ethics statement
The use and handling of animals in this experiment was approved by the Swedish Ethics Committee on Animal Research (Dnr A 32-16), represented by the Court of Appeal for Northern Norrland in Umeå, and the experiment was carried out in accordance with laws and regulations governing experiments performed with live animals in Sweden.