Estimating RPM to RCF in Methods from Older Papers

Estimating RPM to RCF in Methods from Older Papers

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I'm attempting to replicate a cell biology method from a 1958 Laboratory Investigation paper. The protocol is for the isolation of an extracellular matrix protein, and a key step is a centrifugation carried out for 2 hours at 2300 rpm. Apparently it was not common practice to list make, model, and rotor size of centrifuges, so I am a little lost as to what settings to use on my own centrifuge to get the same rcf.

Are there any good starting points for estimating the proper settings? e.g. who the major centrifuge suppliers were at the time or what models would be seen as 'standard' for a cell biology lab?

I am not sure about what they used in 1958 (perhaps Sorvall). However you may look at recent papers for ECM protein isolation (if that was your objective).

Have a look at this article.

From materials and methods:

ECM protein isolation.

Fibroblasts or A431 cells were removed from the surface of the culture plate with 2 mM EDTA in PBS for 3-5 min at 37⁰C. Detached cells were discarded, and remaining ECM proteins attached to the surface of culture plates were washed with the same solution. Cell removal was controlled under the microscope. After cells were completely removed, matrix proteins were covered with 5% acetic acid and incubated at 4⁰C overnight. Then, acetic acid was removed and exchanged with a buffer containing 125 mM Tris-HCl, pH 6.8, 0.1% SDS, 10% glycerol, 1% DDT, 0.05 mM PMSF, protease inhibitor cocktail (Roche, Germany), and incubated at 37⁰C for 1 h. Proteins were removed with a scraper. The procedure was repeated three times. All protein extracts, including acetic acid, were combined and ECM proteins were precipitated by five volumes of acetone. After incubation at -20⁰C overnight and centrifugation at 7000 g for 15 min, the protein pellets were dissolved in an appropriate buffer for further analysis by one- or two-dimensional gel electrophoresis.

In addition, we obtained a Triton X-100-soluble fraction of ECM proteins. Cells were removed as described above; dishes were filled with 1% Triton X-100 and incubated at 37⁰C for 30 min. The extraction was repeated three times. The following procedure of ECM extraction was the same as described above. Samples obtained from both fractions were dissolved in a buffer for one-dimensional (Laemmli) or two-dimensional (rehydration buffer containing 8 M urea, 2% CHAPS, 20 mM DDT, 0.1% Triton X-100) electrophoresis. Proteins that remained after Triton X-100 extraction were dissolved in Laemmli sample buffer.

A practical home-made microcentrifuge for teaching purposes

We report a practical home-made microcentrifuge to be used for teaching purposes. It was made using a salad spinner and two polymerase chain reaction (PCR) tubes racks. It can accommodate 2 standard size 96-well plates or 24 strips or up to 192 microfuge tube. The centrifuge is hand-operated and is ideal for short spin down purposes. Biochemistry and Molecular Biology Education Vol. 39, No. 4, pp. 298-299, 2011

Centrifugation is one of the most important and widely applied research techniques in biochemistry, cellular and molecular biology. The laboratory centrifuges are used for isolating and separating solids from liquids in a suspension. The centrifugation utilities are many, and may include sedimentation of cells and viruses, separation of subcellular organelles, and isolation of macromolecules such as DNA, RNA, proteins, or lipids. A laboratory centrifuge usually comprise a rotor containing two, four, six, or many more numbered wells within which centrifuge tubes may be placed. There are various types of centrifuges, depending on the size and the sample capacity. In a molecular biology laboratory, three types of centrifuges are common. Ultracentrifuge, that speeds up to 70,000 rpm. Large centrifuge that speeds up to about 20,000 rpm and can take tubes of various sizes, depending on the rotors and micro- centrifuges that speeds up to 12 or 13 rpm and are made for spinning 0.2 to 2 mL plastic centrifuge. These small centrifuges are very convenient to spin small amounts of material at low and medium speed.

Most of the molecular biology protocols make use of a microcentrifuge and therefore, most of the laboratory exercises have need of a microcentrifuge [ 1-3 ]. Different experiments like PCR and microarrays require a microcentrifuge that is used for the purpose of spinning down small amounts of template DNA plus other reaction components. Nevertheless, this is expensive equipment for a teaching laboratory. The cheaper microcentrifuges are tiny and handily, have a small capacity, 6–12 tubes from 0.2 to 2 mL and usually required the use of adaptors. The majority of those personal bench-top centrifuges have not interchangeable rotors or adaptors to centrifuge strips and or 96-well plates and therefore a bigger and more expensive centrifuge should be used to that aim. Having these more expensive and larger centrifuges in a laboratory is not always easy, either for economic reasons or/and because the bench top are full of materials. To overcome the lack of a proper centrifuge at the students laboratory that prevent to do certain kind of experiments, we have designed an inexpensive and a hand operating centrifuge that can be used to short spin PCR tubes, strips as well as 96-well plates.

The centrifuge was assembled using the following materials: A salad spinner, two PCR tubes racks and two rubber bands to hold the racks to the interior colander (see Figs. 1 and 2). To operate the centrifuge, just put the tubs in to the racks, close the lid and spin it around. Five or six turns will be enough to spin down liquids. The centrifuge can spin down only liquids, not solids, and is as effective as a “short spin” done in a commercial centrifuge. To test this assertion, a preliminary experiment was conducted in the research laboratory wherein a commercial centrifuge, capable of centrifuge tubes and PCR plates, is available. We did PCR amplification of microsatellites, which is the experiment that is performed more frequently in our laboratory [ 4 ]. The amplifications were performed by duplicate using tubes, strips, and plates. In one of the replicas, only the commercial spin was used where as in the other, the home-made centrifuge was used. No differences were observed when PCR amplifications were checked in agarose gel (see Fig. 3) either when microsatellites were loaded in the ABI 3100 genetic analyzer (applied Biosystems) and visualized at the computer.

Upper: General view of the hand- made centrifuge. Lower: Detail of tubes inside the centrifuge. The picture alsoshows how the rubber bands are attached to the internal colander.

Detail view of the hand- made centrifuge. It has three components, lid, interior colander where racks were attached, and outside plastic bowl.

Photographs of an agarose gel stained with ethidium bromide and visualized by UV transillumination. Upper row (lines 1 to 10): PCR amplifications done using a commercial centrifuge. Lowe row (lines 1 to 10): PCR amplifications done using the hand-made centrifuge.

We routinely use this centrifuge at teaching laboratory. One of the demonstration classes, which are performed at the laboratory, is the detection of genetically modified food. The practice is based on the works of Brinegar and Levee [ 5 ] and Thion et al. [ 6 ]. They are used to detect transgenic corn and soybeans respectably in food products. For the experiments, it is necessary to perform many PCR reactions. The reactions include DNA extracted from different kinds of food and also the positive and negative controls. We have observed that PCR amplifications are more successful when students work with tube strips instead of individual tubes because it minimizes the number of errors. In the laboratory, we have several microcentrifuges in which individual PCR tubes can be centrifuged, but not strips. The centrifuge was initially designed for this purpose. Each group of students, two or three, can have a centrifuge. We have tested that the results of PCR performed are comparable to those obtained using a commercial centrifuge. In addition to practical classes the centrifuge is also used at the research the laboratory to spin down liquids when using 96-well plates.

The centrifuge has several advantages, is very easy and very cheap to make, and do not require any special care and on the other hand, it is not susceptible to imbalance, and therefore it can be used without any special care.


The molecular characterization of extracellular vesicles (EVs) has revealed a great heterogeneity in their composition at a cellular and tissue level. Current isolation methods fail to efficiently separate EV subtypes for proteomic and functional analysis. The aim of this study was to develop a reproducible and scalable isolation workflow to increase the yield and purity of EV preparations. Through a combination of polymer-based precipitation and size exclusion chromatography (Pre-SEC), we analyzed two subsets of EVs based on their CD9, CD63 and CD81 content and elution time. EVs were characterized using transmission electron microscopy, nanoparticle tracking analysis, and Western blot assays. To evaluate differences in protein composition between the early- and late-eluting EV fractions, we performed a quantitative proteomic analysis of MDA-MB-468-derived EVs. We identified 286 exclusive proteins in early-eluting fractions and 148 proteins with a differential concentration between early- and late-eluting fractions. A density gradient analysis further revealed EV heterogeneity within each analyzed subgroup. Through a systems biology approach, we found significant interactions among proteins contained in the EVs which suggest the existence of functional clusters related to specific biological processes. The workflow presented here allows the study of EV subtypes within a single cell type and contributes to standardizing the EV isolation for functional studies.


Implementation of a method for isolation of mtDNA from small tissue samples

To investigate the full mtDNA mutation spectrum in small mouse brain regions and avoid the inherent bias present in PCR amplification, we implemented a method for the enzymatic depletion of nuclear DNA (nDNA) from total DNA [20] extracted from brain tissue. We enzymatically depleted nDNA from total DNA by treatment with exonuclease V, an enzyme that targets the free ends of linear DNA essentially leaving circular mtDNA intact [20]. We used mtDNA-enriched samples directly for library preparation for next-generation sequencing (Fig. 1a).

Ageing increases the load of both SNVs and deletions in mtDNA across all brain regions. a Schematic illustration of the workflow from mouse to prepared library. Briefly, brain regions of interest were rapidly sampled and total DNA was extracted. Linear DNA was enzymatically degraded by exonuclease (ExoV), and non-linear DNA is purified and used for library preparation. FL: full-length mtDNA molecule, ∆: mtDNA molecule with deletion. b Overview of the analysis workflow to optimise mtDNA variant detection. Shortly, after quality filtering, reads were mapped to mm10 without the mitochondrial chromosome (MT). Unmapped reads were then re-mapped to a modified MT reference (dMT: two MT references in tandem) and variants called. c Overview of mouse mtDNA. Green: rRNA encoding genes blue: protein-coding genes red: tRNA-encoding genes orange: non-coding region (NCR). d Schematic showing the areas isolated as the cortex (COR), caudate putamen (CP), dorsal raphe (DR), nucleus accumbens (NAc), paraventricular nucleus of the thalamus (PVT), and substantia nigra (SN). e DNA stored before and after ExoV digestion was subjected to qPCR to determine the relative levels of three mtDNA and three nuclear targets before and after digestion (shown for two different mice, A and B). Mouse C was treated as A and B but without the addition of ExoV. Bars show the mean of target signals and the standard deviation is indicated. †: nDNA after ExoV treatment was not detected or only detected at a very low level by qPCR and may not be visible in the bar plot. f Dot plot illustrating the age-dependent increase in the load of SNVs (left) and deletions (right) across the investigated brain regions (as indicated by the colour legend). All samples have been normalised to the mean of the variants at 10 weeks. Grey diamonds indicate the mean of all regions at the indicated age, and the 95% confidence interval is shown. Three-way ANOVA showed age, not region or animal, significantly (p < 0.01) contributed to SNV and deletion levels. Tukey’s test was used post hoc to determine p values between each age group

We allowed for split-read mapping to identify deletions with BBMap [21] (Fig. 1b) using a custom mtDNA reference composed of two mm10 MT references in tandem (dMT). In a two-round mapping approach, we removed residual nDNA-derived sequencing reads, especially due to the presence of nuclear mitochondrial DNA segments (Numts), i.e. mtDNA-like sequences in the nuclear genome. Due to the circularity of mtDNA (Fig. 1c), deletions may span the “ends” of the mtDNA reference, that is linear in nature, which will interfere with deletion calling. By using dMT, we circumvented this and were able to reliably detect variants at any position in mtDNA. As sequencing reads generated by Nextera are well known to exhibit GC bias in the first bases of the read, we trimmed these bases and excluded an additional 5 bp at the read ends during variant calling (see the “Methods” section). Based on this, we have no reason to believe that variant calling is influenced by transposase sequence bias. For identification of mtDNA variants, we sampled the sensory cortex (COR), caudate putamen (CP), dorsal raphe (DR), nucleus accumbens (NAc), paraventricular thalamic nucleus (PVT), and substantia nigra (SN) (Fig. 1d) during mouse ageing and confirmed nDNA depletion by qPCR before library prep and sequencing (Fig. 1e).

Ageing-related accumulation of SNVs and deletions across all brain regions

We initially mapped the ageing-related changes in mtDNA mutations across the brains of 10-, 50-, and 80-week-old wild-type (WT) mice (Fig. 1f). As expected, the number of SNVs in 10-week-old mice was very low but rose on average 10-fold in 50-week-old mice and remained relatively unchanged at 80 weeks (Fig. 1f, left). Similarly, deletions also reached a plateau at 50 weeks after increasing 2.5–3-fold from 10 weeks (Fig. 1f, right). The plateau reached in both SNVs and deletions at 50 weeks indicates a restriction in the load of mtDNA mutations. This may be imposed by loss of mitochondria function, thus limiting its propagation or triggering mitophagy. Alternatively, selective replication of mtDNA molecules may keep the mutation load from further increasing. In all, both SNVs and deletions appeared homogenously accumulated across the examined brain regions during ageing, but may be influenced by different pathological settings.

Polg D181A expression causes brain region-specific SNV accumulation with ageing

Having established that our method could be used to map ageing-induced mtDNA mutations, we turned to our Polg D181A model mice to investigate the brain region-specific influence of proof-reading deficiency. We sequenced mtDNA from the six brain regions of interest from Polg D181A mice and identified the SNVs in each region at 10, 50, and 80 weeks of age (Fig. 2a).

SNVs heterogeneously accumulate across brain regions in Polg D181A mice and cause mtDNA position-specific mutational patterns. a Dot plot illustrating the age-dependent increase in the load of SNVs in Polg D181A mice across the investigated brain regions (as indicated by the colour legend) normalised to the mean of WT samples at 10 weeks. Grey diamonds indicate the mean of WT-derived brain region samples for reference (same as in Fig. 1f). Red diamonds indicate the mean of Polg D181A -derived brain region samples and the 95% confidence interval is shown. Three-way ANOVA (age, region, and animal) of Polg D181A -derived samples showed that age significantly contributed to SNV levels (p values of post hoc Tukey’s test are shown). Three-way ANOVA showed a significant contribution of all variables (age, genotype, region). p values of post hoc Tukey’s test comparing WT and Polg D181A at each age are shown. For region contribution, we found a significant contribution of COR, NAc, and PVT to SNV levels in Polg D181A mice using a linear model for main effects. b SNVs were counted in 10-bp non-overlapping bins for WT (grey) and Polg D181A (red) mice at 10, 50, and 80 weeks, and the number of regions with SNV in each bin calculated. Note that in the case that one region has more than one SNV in a bin, it is only counted as one instance of an SNV. The overlap was visualised for non-overlapping bins (“1”), bins shared across two or three regions (“2–3”), and bins shared across four to six regions (“4–6”). c SNVs were counted in 10-bp non-overlapping bins for WT (grey) and Polg D181A (red) mice at 50 weeks, and the number of individual animals with SNVs in each bin calculated. Note that in the case that one animal has more than one SNV in a bin, it is only counted as one instance of an SNV. The overlap was visualised for non-overlapping bins (“1”), bins shared across two or three animals (“2–3”), and bins shared by four or more animals (“≥ 4”). d Cumulative percentage of SNVs detected in each examined brain region (thin lines) for both WT (grey) and Polg D181A (red) at 10, 50, and 80 weeks old. Bold lines indicate the smooth conditional mean for each genotype. e The relative average SNV allele frequency for each region for WT (grey) and Polg D181A (red) mice at 10, 50, and 80 weeks as indicated shown as boxplots. p values of two-sided t tests are shown. f SNVs across brain regions were pooled for each genotype at each age and divided into 100-bp bins across the mtDNA reference and the allele fraction for SNVs in each bin summed and normalised (i.e. highest peak set to 1). Grey areas indicate mtDNA regions where peaks are found across all variables (α), peaks that are ageing-dependent (β), and ageing-induced Polg D181A -dependent peaks (γ)

In young mice, there was no change in SNV levels between WT and Polg D181A in 10-week-old mice. At 50 and 80 weeks, we observed a significant increase in SNV levels which was especially prominent in COR, NAc, and PVT (Fig. 2a). This demonstrated that the heterogeneity of the mitochondrial response to proof-reading deficiency is present across an organ and not only between organs [6, 16, 22, 23].

Importantly, we found no relationship between the expression of the Polg D181A transgene in the investigated brain regions and the level of detected SNVs (Additional file 1: Fig. S5c).

SNVs are excessively shared between brain regions

We wondered whether the increase in SNVs with both ageing and Polg D181A expression was affecting the same positions in mtDNA across brain regions. Indeed, looking in 10-bp non-overlapping intervals, we found an increase in shared SNV positions with ageing which was enhanced by Polg D181A expression (Fig. 2b). The overlap between individual animals was most prominent in PVT and NAc (Fig. 2c). As Polg D181A expression introduced a shift of the SNV distribution to the right side of the distribution plot (i.e. towards the NCR) compared to WT (Fig. 2d), we examined where shared SNVs were located. We found that shared SNV positions were significantly different from non-shared SNV positions in Polg D181A mice (t test, p < 1 × 10 −5 ) but not in WT (t test, p = 0.254) when looking across all ages and shifted towards the 3′ region (i.e. towards the NCR) (Additional file 1: Fig. S1a).

The increase in shared SNVs was accompanied by a significant increase in SNV frequency with Polg D181A expression (Fig. 2e) which was driven by high frequency SNVs in specific mtDNA regions (Fig. 2f). These regions appeared highly context-dependent, i.e. peaks that are found across all samples (“α” on Fig. 2f), only in very aged mice (“β”), or are Polg D181A -specific (“γ”). This was mimicked in the Pearson correlation, where most brain regions from 50- and 80-week-old Polg D181A mice form a distinct cluster and most samples from 10-week-old mice form a distinct cluster (Additional file 1: Fig. S1b) and we found specific SNV hotspots in 10-week-old animals independent of genotype (Additional file 1: Fig. S1c). In addition, there was a significant overlap of the specific positions at which SNVs are present in COR, NAc, and PVT (the brain regions most sensitive to Polg D181A expression) at both 50 and 80 weeks in Polg D181A mice (Additional file 1: Fig. S1d, right). For WT mice, the overlap is only pronounced at 80 weeks and p values do not reach similar levels of significance (Additional file 1: Fig. S1d, left).

We found an increase of SNVs in the NCR and complex III genes (Additional file 1: Fig. S1e), while transitions and transversions (Additional file 1: Fig. S1f) were comparable to those of previous studies, and we saw no indication of either ageing- or Polg D181A -induced oxidative mutations [19, 24], together with no change in the types of mutations (Additional file 1: Fig. S1g).

Together, these data demonstrated a brain region-specific ageing-dependent Polg D181A -induced mtDNA SNV spectrum, where COR, NAc, and PVT are regional hotspots. In addition, certain mtDNA positions are highly sensitive to SNVs and seem to function as context-dependent mutational hotspots.

Ageing-dependent Polg D181A -induced deletions accumulate in the same brain regions as SNVs

We next turned our attention to the influence of Polg D181A expression on the accumulation of deletions. We found a significant ageing-induced accumulation of deletions in both 50- and 80- compared to 10-week-old Polg D181A mice, but we observed no significant differences between WT and Polg D181A at any age (Fig. 3a). However, deletion accumulation in response to Polg D181A showed a prominent region specificity. While CP, DR, and SN Polg D181A SNV levels were only slightly elevated compared to WT mice, COR, NAc, and PVT showed a very high accumulation of deletions at 50 and 80 weeks. We found a significant difference in deletion levels between these regions compared to the other regions in Polg D181A when pooling data from 50- and 80-week-old mice (p = 0.002, one-way ANOVA). Similar to SNVs, we found no indication that expression levels of the Polg D181A transgene were the major driver of deletion levels in the Polg D181A mice (Additional file 1: Fig. S5c).

Accumulation of deletions induced by Polg D181A expression are brain region-specific and ageing-dependent. a Dot plot illustrating the age-dependent increase in the load of SNVs in Polg D181A mice across the investigated brain regions (as indicated by the colour legend) normalised to the mean of WT samples at 10 weeks. Grey diamonds indicate the mean of WT-derived brain region samples for reference (same as in Fig. 1f). Red diamonds indicate the mean of Polg D181A -derived brain region samples and the 95% confidence interval is shown. Three-way ANOVA (age, region, and animal) of Polg D181A -derived samples showed that age significantly contributed to deletion levels (p values of post hoc Tukey’s test are shown). p values of three-way ANOVA (age, genotype, region) with post hoc Tukey’s test are shown for each age group. For region contribution, we found a significant contribution of COR, NAc, and PVT to deletion levels in Polg D181A mice using a linear model for main effects. b Chord diagrams indicating the deletions accumulated at 10, 50, and 80 weeks in DR and PVT from WT and Polg D181A mice. Data is normalised pr. brain region, and the width of each gene indicates the summed allele fraction of deletions spanning the indicated gene(s). The colour of the chord indicates the gene in which the breakpoint 5′ position is located. Plots were made using circlize

The differences across brain regions with ageing of WT and Polg D181A mice can be appreciated by chord diagrams showing the span of all deletions at each time point (Fig. 3b). Where PVT showed both an ageing-induced and a clear ageing-dependent Polg D181A -induced accumulation of deletions, DR only showed an ageing-induced accumulation of deletions, highlighting the brain region-specific mtDNA sensitivity to a setting of replication instability. Pearson correlation indicated similarities in the deletions found with ageing of Polg D181A mice (Additional file 1: Fig. S2a), indicating that Polg D181A expression induces a specific landscape of mtDNA deletions.

Deletions share characteristics independent of genotype

The positions at which deletions start and end are termed breakpoints and based on the co-occurring deletions between brain regions from Polg D181A mice, we hypothesised that breakpoints must be shared between different samples. We looked in 100-bp bins along the mtDNA and found that shared breakpoints cluster in very distinct locations (Additional file 1: Fig. S2b). Some shared breakpoints are age- and genotype-independent (

5 kb) whereas others are genotype-dependent (

15 kb). Sizes of deletions themselves follow a bimodal distribution independent of age and genotype and can be roughly divided into those < 100 bp and those > 1 kb, with few observations in the intermediate range (Fig. 4a). Even though the number of deletions in 10-week-old animals is low, they still follow this distribution, though the fraction of very small deletions is high compared to aged animals.

Characteristics of deletions change with age, but not the expression of Polg D181A . a Density plot of deletion sizes for WT (grey) and Polg D181A (red) for 10- (dotted line), 50- (dashed line), and 80-week-old (full line) mice. b The shortest average distance from 5′ and 3′ deletion breakpoint pairs to a direct repeat pair in the mitochondrial genome for the observed deletions (darker colour) and a random in silico generated deletion length-matched library (lighter colour) for both WT (left, in grey) and Polg D181A (right, in red) using pooled data from all ages and brain regions examined for each genotype. p values of two-sided t tests are shown. c Needle identity score calculated in a ± 10 bp window at the 5′ and 3′ deletion breakpoints as a function of deletion size after pooling of WT and Polg D181A samples. Correlation for each age is indicated by the full lines and correlation data indicated in the same colour code

Molecular determinants of deletions

A previous study has suggested that the majority of mtDNA deletions in Parkinson’s patients occur at direct repeats [17], a proposed [25] though highly debated [26] feature of human mtDNA deletions. To investigate the influence of direct repeats in breakpoint formation in the mouse brain, we identified direct repeats ≥ 8 bp in mtDNA (Additional file 1: Fig. S2c). After pooling deletions per genotype, we identified the direct repeat pair with the shortest average distance from the 5′ and 3′ breakpoints of WT and Polg D181A deletions as well as for in silico generated, deletion length-matched deletion libraries for each genotype (see the “Methods” section). The shortest average distance was shorter for experimentally derived deletions than randomly generated deletions for both WT and Polg D181A mice (Fig. 4b), but there was no difference between WT and Polg D181A (t test, p = 0.905). This indicates that direct repeats may contribute to at least a part of the identified deletions. We found this to be the case at 10 and 50 weeks but not 80 weeks (Additional file 1: Fig. S2d), as deletions are significantly closer to direct repeats than the in silico deletion libraries for both WT and Polg D181A mice.

Restriction of sequence similarity to direct repeats is a rigorous criterion. We therefore calculated the sequence identity scores in a 20-bp window surrounding all 5′ and 3′ breakpoints (i.e. 10 bp on each side of the breakpoint). We found a negative correlation between sequence identity score and deletion length in 50- and 80-week-old Polg D181A mice (Fig. 4c) and further saw a significant difference between the identity scores of deletions < 100 bp and > 100 bp at 50- and 80-week-old Polg D181A as well as WT mice (Additional file 1: Fig. S2e). These data imply a differential contribution of non-direct repeat sequence similarity to short and long deletions.

Abundant NCR multimers are exclusive to Polg D181A -expression

As expected, the sequencing coverage exhibited some variability, likely associated with a slight sequence specificity of the transposase used for library preparation [27,28,29]. However, in specific brain regions from the 50- and 80-week-old Polg D181A mice, we observed an increased coverage in the 15 kb+ region including at least a part of the NCR (Additional file 1: Fig. S3a). Localised increased coverage is often thought to be associated with duplicated regions. As mitochondrial DNA is circular, it is not possible to distinguish small duplications from very long-range deletions (VLRDs) (Additional file 1: Fig. S3b). We therefore wondered if our data of mtDNA deletions could support the presence of multimers. By classifying VLRDs as deletions > 15 kb, we found that VLRDs are specifically enriched in NAc and PVT from 50- and 80-week-old Polg D181A mice (Fig. 5a) and enriched in the 15 kb+ region (Fig. 5b and Additional file 1: Fig. S3c), supporting the idea that mtDNA multimers including at least part of the NCR accumulate in a brain region-specific and Polg D181A -dependent manner with age. We also found an increase in discordant reads in 50-week-old Polg D181A mice, which further supports the presence of genomic rearrangements such as multimers (Fig. 5c and Additional file 1: Fig. S3c).

Putative NCR multimers are Polg D181A -specific and accumulate with age in a highly brain region-specific manner. a Dot plot illustrating the age-dependent increase in the load of VLRDs in Polg D181A mice across the investigated brain regions (as indicated by the colour legend). All samples have been normalised to the sample with the lowest number of detected variants. b Cumulative percentage of 5′ position of VLRDs (i.e. start position of the putative multimeric sequence) summed across brain regions for WT (grey) and Polg D181A (red) for 10- (dotted line), 50- (dashed line), and 80-week-old (full line) mice. c Mean number of discordant reads as extracted by samtools at 10, 50, and 80 weeks for WT (grey) and Polg D181A (red) and standard deviation is indicated. p values of two-sided t tests are shown. d Summed analysis of mtDNA breakpoints of VLRDs from Polg D181A mice in the NCR and surrounding region. 5′ (purple) and 3′ (dark turquoise) breakpoints are summed at each position across all brain regions at either 50 (left) or 80 (right) weeks old, and smooth conditional means are plotted. The lower panel shows the phastCons conservation score via the UCSC genome browser in the same region. p values of two-sided t tests are shown. e Boxplot showing the shortest average distance to a direct repeat of all Polg D181A VLRDs separated by VLRD 5′ position into < 15 kb (light blue) or > 15 kb (dark blue). p values of two-sided t tests are shown. f Boxplot showing the Needle identity score of WT and Polg D181A -derived VLRDs pooled across ages and brain regions examined split into VLRDs with 3′ position < 15 kb (light blue) and > 15 kb (dark blue). p values of two-sided Wilcoxon tests are shown

These putative multimers appear to form in a quite restricted region of mtDNA as their 5′ and 3′ “breakpoints” (indicating the end and the start of the duplicated sequence, respectively), accumulate at rather discrete positions (Fig. 5d) spanning a region with a low conservation score across mammals (Fig. 5d, bottom panel). Previous data suggested the presence of multimers in the brain from the mutator mouse [16]. We used the same PCR approach as Williams et al. and validated the Polg D181A -specific presence of multimers (Additional file 1: Fig. S4a, top and middle panel). An alternative PCR setup that would only yield a product in the presence of multimers confirmed these results (Additional file 1: Fig. S4a, bottom panel) and subsequent data also indicated the presence of inversions (Additional file 1: Figs. S4b,c).

Together, these data support the presence of highly brain region-specific ageing-induced Polg D181A -dependent multimers which are highly specific to a partial NCR-containing segment of mtDNA specifically in NAc and PVT.

Direct repeats may be involved in NCR multimer formation

Multimers can be formed by several mechanisms. One mechanism is by strand slipping during replication which may be influenced by the local environment surrounding the NCR, which is known to interact with the inner mitochondrial membrane [9]. Another mechanism is mediated by the DNA sequence surrounding the start and end positions of the multimer region. In support of the idea of strand slipping, we find VLRDs in the 15 kb+ region to be closer to direct repeats compared to multimers in other parts of the mtDNA (Fig. 5e), though the overall sequence similarity surrounding breakpoints is not different (Fig. 5f). SNVs were enriched near VLRD 5′ breakpoints as well as

7 kb upstream with a mean distance of 75 ± 462 bp to the nearest SNV (Additional file 1: Fig. S5a). Fifteen percent of VLRD breakpoints co-position with SNVs, a number which is not influenced by discordant reads. SNVs were not enriched within the putative multimeric region (Additional file 1: Fig. S5b).

Transgene expression level does not drive variants

The expression of transgenes are often not similar across tissues, which is also true for Polg D181A expression [14]. To confirm that the expression differences were not driving the differences we observed in the accumulation of mutations in response to Polg D181A expression, we evaluated the expression levels of endogenous Polg and transgenic Polg D181A . Importantly, we were interested in the relative expression levels of the two transcripts, as endogenous and transgenic Polg will be competing for access to mtDNA during replication. As presented in previous sections, we find no correlation between mtDNA mutation levels and relative Polg D181A /Polg levels at any age (Additional file 1: Fig. S5c). Together, this demonstrates that transgene expression levels were not the major driver of brain region specificity to proof-reading deficiency in mitochondria.

MtDNA variants cluster together along genomic regions

Throughout our analysis of the mutation spectrum of mtDNA from both WT and Polg D181A mice, it became increasingly clear that different types of variants often were found in specific mtDNA regions. To further investigate this, we plotted all variants analysed—SNVs, deletions, VLRDs (i.e. multimers)—across mtDNA in a circular plot (Fig. 6a). Visual inspection of this plot showed that different types of variants are enriched in the vicinity of each other. We found a strong, positive correlation between SNV and deletion load at the gene level which is independent of ageing and genotype (Fig. 6b), indicating positional sensitivity to the accumulation of mutations which may reveal underlying genomic instability in specific regions or be caused by higher order structures.

Levels of different variants correlate across mtDNA genes. a Identified variants plotted across the mtDNA reference for WT (top) and Polg D181A (bottom) samples across all regions at 10, 50, and 80 weeks as indicated. Tracks from outside to inside: (1) mtDNA gene names (note that tRNA gene names are not shown), (2) mtDNA genes by length, (3) the relative level of VLRDs (e.g. multimers and inversions), (4) SNVs detected across all regions with height indicating log2-transformed allele frequency, and (5) deletions plotted as lines connected to start and end positions. Note that the start and end points of deletions are not indicated. b Load of SNV and deletion for each mtDNA gene divided by the gene length are plotted for WT (grey) and Polg D181A (red) for 10-, 50-, and 80-week-old mice. Data was scaled (from 0 to 1) before plotting for clarity. Pearson correlation and significance of the correlation is shown below each plot for both WT (grey) and Polg D181A (red)


2.1 Modelling the required carbon input to the soil

(1) where SOC is a dimension 4 vector with four components, each component referring to the organic carbon content in one of the four dynamic compartments of the RothC model: the resistant plant pool (RPM), the decomposable plant pool (DPM), the microbial pool (BIO) and the humic pool (HUM). is a dimension 4 vector that represents the C amounts incorporated in the four dynamic pools, and F is a 4 × 4 matrix representing SOC mineralization and carbon flows between pools. The type of vegetation influences the distribution of C inputs into the RPM and DPM pools hence, the DPM:RPM ratio typically depends on the vegetation type. In RothC, four vegetation types are considered: croplands, improved grasslands, unimproved grasslands and forests with a DPM:RPM ratio of 1.44, 1.44, 0.67 and 0.25, respectively. For a given total carbon input and mineralization rate, land use with lower values of the DPM:RPM ratio will exhibit higher total SOC stocks.

where , hereafter called the equilibrium or steady state, can then be easily calculated and depends on organic carbon (OC) input rates and the F matrix: (2) where is the identity matrix of dimension 4 × 4 and is the vector of constant carbon inputs. Conversely, if estimates of and exist, can, in turn, be estimated, assuming that the soil has reached equilibrium and that climatic conditions are constant. Note that SOC refers to SOC that is subject to SOC dynamics. For the RothC model, this dynamic SOC is a fraction of the total SOC.

where is the total SOC, which can be measured, and IOM the inert SOC, which is, according to RothC, constant over time. IOM is usually estimated using the Falloon et al. ( 1998 ) equation. Equation (2) gives: (3)

Like in Equation (1), is a dimension 4 vector. For the purpose of simplicity, (and later on ) hereafter refers to total carbon inputs into the soil, that is, the sum of the four components of this vector.

In the present study, we used the RothC model to estimate (i) the inputs of SOC that would be needed to maintain current SOC stocks, hypothesizing that these are at steady state and (ii) the increase in SOC inputs needed to reach SOC stocks in 30 years from now, assuming a constant yearly 4‰ rate of increase in SOC. Note that the steady-state hypothesis is not supported by any data since a robust dataset is not yet available for mainland France. It was tested at a later stage of our work when we compared carbon input levels estimated by RothC under this hypothesis and NPP levels obtained independently from RothC (see Section 2.3).

  1. Compute SOC 0 as (Falloon et al., 1998 )
  2. Using Equations (2) and (3)
    1. Split among , , and
    2. Calculate needed to have the observed
    Step 4 was done using a differential evolution optimization algorithm (Ardia et al., 2016 ). From the estimate, the increase in SOC input was calculated as (4)

    Additionally, to assess the effect of climate change on SOC inputs needed to maintain or increase SOC stocks, RothC was run with two climatic datasets: observed data (1980–2010) and simulated data taking into account climate change (RCP 8.5). This scenario was selected because it predicts the highest increase in temperature and cumulative CO2 emissions, with potentially important consequences on SOC dynamic which is affected by both the increase in temperature and the increase in C inputs due to the CO2 fertilization effect (Meinshausen et al., 2011 Wieder et al., 2015 ). Furthermore, Schwalm et al. ( 2020 ) showed that, looking at mid-century and sooner, RCP 8.5 is clearly the most useful choice: it is consistent with historical total cumulative emissions for the present period, and given current and stated policies, it gives the most plausible cumulative emissions for the 2030–2050 period. Results for the observed 1980–2010 climatic conditions are presented first, and those for the 8.5 RCP scenario (1980–2010 and 2020–2050) are further used to discuss the effect of climate change.

    The RothC model was implemented within the RothC R package (Martin, 2018 ), GIS operations using GRASS GIS software (GRASS Development Team, 2018 ) and statistical analysis using R software (R Core Team, 2015 ).

    2.2 Data for estimation of with RothC

    RothC needs several input variables to simulate carbon dynamics and SOC mineralization. SOC mineralization is a function of soil clay content (which drives SOC stabilization in different pools and soil moisture), temperature, precipitation, potential evapotranspiration and soil cover (bare or covered). Soil cover drives mineralization both directly and indirectly through soil moisture content. Temperature drives mineralization directly, and precipitation and potential evapotranspiration drive mineralization indirectly through soil moisture content. Additionally, in our framework, we needed maps of SOC in the top 23 cm of soil (as RothC is parameterized to model SOC stocks in the 0–23 cm soil layer) as inputs for Equation (3), and the share of total C inputs between plant residues and organic fertilization since these two categories of C inputs is characterized by specific parameters.

    2.2.1 Soil data

    We used the data from the recently produced GlobalSoilMap products for France (Mulder et al., 2016 ), for both clay and SOC. GlobalSoilMap products provide estimates for, among others, the 0–5, 5–15, 15–30 cm depth layers at 90 m resolution for France. Clay and SOC datasets were aggregated to the 0–23 cm layer using weighted averages of estimates for GlobalSoilMap layers. SOC stocks were calculated assuming a constant rock fragments content (2%) and estimating bulk density with a pedotransfer function (see Meersmans et al., 2012 for details). For French soils, GSM estimates of bulk density are not yet available mainly because of data scarcity and issues related to measurement methods. The same applies for rock fragments content. Soil depth estimates (Lacoste et al., 2016 ) were used to truncate soil profiles on pixels. For instance, where soil depth d is less than 23 cm, stocks are computed on 0-d cm instead of on 0–23 cm. All calculations were first performed on a 90 m resolution grid, and later aggregated to a 1 km resolution grid. All subsequent work was done on this 1 km resolution grid to reduce computation time. A finer resolution was not required for our study, as the maps produced were only used to analyse regional and national patterns.

    2.2.2 Climate data

    Monthly rainfall (mm month −1 ), reference evapotranspiration (PET, mm month −1 ) and temperature (monthly averages, in degrees Celsius) were averaged over the 1980–2010 period from the French SAFRAN reanalysis in order to yield a reference year representing current climate in France on 8 × 8 km 2 pixels (Quintana-Segui et al., 2008 ). PET was calculated using the FAO Penman–Monteith method (Allen et al., 1998 ). Climate projections data are from the French climate model ALADIN (CNRM-CM5/CNRM-ALADIN53) for the CO2 concentration scenario RCP 8.5. Climate projections have been debiased with the SAFRAN reanalysis on each pixel using a Statistical Downscaling Method based on a Quantile Mapping approach. PET was again calculated with downscaled climate data (temperature, relative humidity, solar radiation and wind speed).

    2.2.3 Landcover

    Landcover was estimated using ecoclimap data (Faroux et al., 2013 ), at 1 km resolution (see Figure 1 Supporting Information S1). Ecoclimap predictions were grouped into four main categories (i.e. croplands, grasslands, forests and others). Simulations were only performed on croplands, permanent grasslands and forests. We distinguished improved permanent grasslands from unimproved permanent grasslands using data on grassland types from a previous study (Tibi & Therond, 2018 ), which relied on a classification proposed by Devun and Legarto ( 2011 ).

    2.2.4 Management and input data scenario for RothC

    When run in inverse mode, that is in order to estimate required carbon input levels to reach a given SOC level, RothC solely needs two input data related to management. The first one is the number of months where soils are left bare. This input variable, only applicable for croplands, was set to 4 months except when soils belonged to nitrate vulnerable zones where it was set to 2 months since cover crops are mandatory in these zones. The second input variable is the proportion of carbon inputs to the soil that consists of organic amendments. OC inputs to the soil are mainly the result of plant residues and of additions of animal manure and other organic products (hereafter referred to as organic amendments). In RothC, the fate of carbon provided by plant residues and organic amendments is specific, reflecting their difference in terms of decomposability. Therefore, in order to use RothC with the framework presented above, to estimate the amount of C input needed to sustain a given carbon stock, or to increase it, one needs to decide the share between plant residues and organic amendments in C inputs. This proportion was thus prescribed, based on previous studies having estimated it at the regional scale for France (Tibi & Therond, 2018 ). Furthermore, we considered that the only type of organic amendment was farmyard manure, for which RothC provides a default composition and which represents about 60% of organic amendments spread on agricultural soils in France (Houot et al., 2014 ). In RothC, the OC in this organic material is split into the RPM, DPM and HUM pools in the following proportions: 49%, 49% and 2%. Although another parameterization has been proposed for RothC (Peltre et al., 2012 ), with higher HUM fraction for organic amendments, taking it into account was not possible because spatial data on the nature of organic amendments were not available for our study. Organic amendments were only allowed on croplands and grasslands, not on forests. We also considered that an increase in total OC inputs during the 4‰ carbon storage period (i.e. between t0 and t0 + 30 years) was only possible through an increase in inputs of plant residues, and not through increments of organic amendments. In mainland France, all animal manures are already spread on agricultural soils so that it is not possible to increase the availability of this resource (Houot et al., 2014). Moreover, because of the low social acceptability of spreading urban and industrial organic products, increasing the amount spread on agricultural soils is unlikely. Note that other information about management practices in croplands, grasslands and forests were used in this study, but at a later step in the process, that is, for estimating independently from RothC the available NPP flowing to the soil (see also Figure 1), which is presented in the next section.

    2.3 Available NPP and ecosystems' carbon balance

    (5) where NPP is the net primary productivity, is the amount of NPP allocated to increase plant biomass, are exports due to human activity, the carbon returned through animal faeces and manure application, and is the proportion of total C inputs into the soils allocated to the 0–23 cm soil layer only (which is the layer considered by the RothC model, and which matches, for France, the average depth of the plough layer Arrouays et al., 2001 ). was derived from recently published estimates of belowground NPP flows (Balesdent et al., 2018 ), depending on land use, clay content, mean annual temperature and the mean ratio of annual precipitation to potential evapotranspiration. For grasslands and croplands, was assumed to be equal to 0 considering that, based on a yearly average, the amount of carbon in aboveground and belowground plant biomass remains constant in the long term, contrary to non-mature forest systems. We defined the carbon balance of a given soil as the difference between available NPP flowing to the considered soil layer () and the soil carbon input (), as estimated with the RothC model, needed to maintain current SOC levels or to reach the 4‰ target SOC stocks. (6) with equal to or (see Section 2.1 about the modelling framework), yielding, respectively, and .

    NPPsurf and were evaluated with separate methods and data sources, and computing the balance between them addressed the following question: is the carbon input to the soils required to sustain current stocks or to reach the 4‰ target available? Figure 2 summarizes this approach based on a comparison between the required carbon input, represented by the variable, and the available carbon input is represented by the NPPsurf variable, estimated here using data about plant productivity and human activities. Studying the sign of may lead to different conclusions depending on the hypothesis on the stationarity of SOC stocks. If differs from zero, the steady-state hypothesis is currently not valid. indicates that is not sufficient to sustain existing SOC stocks which are on a declining trend. If , SOC stocks might be on an increasing trend. Alternatively, when RothC is used to compute , that is, carbon required to reach the 4 target after 30 years, negative values indicate that, given NPP levels and human activity, there is indeed insufficient carbon to reach this target. Positive values indicate that higher targets could, indeed, be reached. Note that calculations were performed using a current database, thus based on current plant productivity and existing agricultural and forestry systems. Moreover, the proposed framework ignores some of the relatively minor fluxes that result in C inputs and outputs. These include erosion, fires, dissolved organic carbon leaching, methane emissions and volatile organic carbon emissions (see Soussana et al., 2019 for a full accounting of these components). This choice was made because of available data and for the sake of simplicity. However, as all these fluxes result in SOC losses, it might induce a global overestimation of the carbon balances proposed here. RothC itself does not represent some output fluxes including vertical transport in the deep layer due to bioturbation or advection or lateral transport due to erosion. This may also contribute to carbon balance overestimation but these non-represented fluxes are considered to be of second order compared to heterotrophic respiration (Jagercikova et al., 2014 Naipal et al., 2018 Warner et al., 2019 ) and are particularly difficult to model at regional scales.

    Diagram of the proposed approach. Main input variables are current soil organic carbon stocks (), available net primary productivity () and clay content. Two cases are considered. The first one deals with current SOC stocks and the second one with the 4‰ target. Both the carbon balances (, Equation 6) and the carbon saturation deficit (, Equation 7) are used to assess the status of current SOC stocks and the feasibility of the 4‰ target. Note that the clay soil input variable is used for RothC computations for both and . The steady-state hypothesis enables to estimate using Equation (3), the right-hand side of which is called here for the sake of conciseness. The expected trends of current stocks and the feasibility of the 4 do not account for possible future changes in SOC mineralization rates

    Several datasets were combined to yield spatial estimates of , , and on our 1 km × 1 km grid, depending on land use of the grid cells. Comparison of the results yielded by the different sources of input data was then used to discuss the uncertainty of our results. These datasets included NPP estimates derived from MODIS for the 2001–2012 period (Zhao et al., 2005 ), NPP and its human appropriation for the year 2006 (Plutzar et al., 2016 , using the LPJml model for forests and grasslands and a model based on regional yield statistics for croplands). LPJml is a global model (GM), meaning that although it accounts for plant productivity variations, in managed and unmanaged areas, it does so less specifically than do domain models (DM), for example, models dedicated to grasslands and croplands. French inventory data were used to estimate human appropriation in forests (IGN, 2018 ). NPP and human appropriation data yielded by previous simulations (Tibi & Therond, 2018 ) with the STICS (a crop model, Brisson et al., 1998 , 2003 ) and PASIM (a pasture simulation model, Riedo et al., 1998 ) domain models were also used. Lastly, for the variable, we used data assembled from national inventories to estimate organic fertilization (Tibi & Therond, 2018 ). How the different data sources were combined depending on the land use is detailed in Table 1 of Supporting Information S1.

    2.4 Estimates of SOC saturation levels and saturation deficit

    Several studies have assessed C sequestration potential or C saturation deficit (i.e. additional SOC that can be stabilized in the fine fraction) across different land uses (croplands, grasslands, forests) at large extent (Angers et al., 2011 Chen et al., 2018 Wiesmeier, Hübner, Spörlein, et al., 2014 ). We estimated the SOC saturation level () based on concentrations of mineral fine fractions and applied the equation proposed by Hassink ( 1997 ). Fine fraction content comprises the particle of size <20 μm (%). To estimate fine silt (2–20 μm) fractions for each of our 1 km × 1 km grid cells, we combined GlobalSoilMap predictions of clay (<2 μm) with estimates of clay:fine silt ratio based on the French soil monitoring network data (Réseau de Mesures de la Qualité des Sols: RMQS, Jolivet et al., 2006 ). The sum of clay and fine silt fractions was then used in Hassink's equation to estimate SOC saturation levels. From the SOC content at saturation, SOC stock at saturation was estimated using the same method as previously explained for predictions based on the GobalSoilMap SOC content. The SOC saturation deficit was then estimated as (7) where (Mg ha −1 ) is the SOC stock at saturation and and are the SOC stock in the HUM and IOM pools as predicted by RothC for the current SOC stock. In so doing, we assumed that RothC's RPM and DPM pools are the same as particulate organic matter, by definition not included in soil organic matter bound to the fine fraction (Stewart et al., 2007 ), and further, we assumed that carbon in the BIO pool was negligible compared to that in the HUM and IOM pools.

    2.5 Uncertainty analysis

    Our uncertainty analysis consisted in estimating the variance of the 4 carbon balance () and of the saturation deficit (). We included all variables and parameters used in their calculation and for which information about variance was available. We used the uncertainty estimates of the clay and SOC estimates attached to the GlobalSoilMap data. Although uncertainty attached to the NPP products was not available, we explicitly included in the analysis the variance resulting from using different sources of NPP data. This variance was considered to be representative of the uncertainty associated with the current knowledge of NPP levels. We also included parameter-related uncertainty when it was available, that is, for parameters of the psurf function (Balesdent et al., 2018 ), for the parameters of the function used to estimate the amount of inert organic carbon (Falloon et al., 1998 ) and for the parameters of the equation used to estimate the carbon saturation deficit (Hassink, 1997 ). The propagation of the uncertainty attached to these input variables and parameters was based on an analytical formulation of and , taking advantage, among others, of the simplicity of the RothC model and the availability of explicit solutions of Equation (1) (see Supporting Information S3). We applied a first-order Taylor analysis to calculate the variance of intermediate functions of and , an approach which was previously applied in soil sciences (Heuvelink et al., 1989 Román Dobarco et al., 2019 ). Using the Taylor analysis enables one to approximate the variance of any continuously differentiable function of a set of variables or parameters. This approach has the major advantage to reduce computation time (compared to Monte-Carlo approaches) and to facilitate the identification of the various sources of uncertainty. Details of our procedure are presented in Supporting Information S2.


    Reproducibility of repeated single measurements

    There were no systematic absolute or relative differences in VO2 and HR between the first and second measurement occasion in the laboratory (Table 3).

    Positioning work rates for the HR-VO2 relations in the laboratory

    The three submaximal work rates, used in both models of HR-VO2 regression equations, induced mean levels of HR ranging from on average 97 ± 8 to 139 ± 18 beats per minute for the males, and from 98 ± 8 to 150 ± 10 for the females (Table 3). For maximal HR and other descriptive aspects of the work rates used, see Tables 3 and 4.

    HR levels from commuter cycling used for estimating levels of VO2

    The mean values of the 20% lowest, intermediate and highest heart rate segments during the commuter cycling and their mean HR values are described in Table 5.

    The corresponding levels of percent of heart rate reserve and percent of HRmax are also given (mean ± SD).

    The mean levels of all HR (not shown) were somewhat lower than the intermediate 1/5 of the size ordered HR. This is since the lowest 1/5 of HR is clearly further away from the intermediate 1/5 than the distance to the highest 1/5 of HR.

    Reproducibility of HR-VO2 regression equations and estimated levels of VO2 (model 1)

    The test and retest HR-VO2 regression equations and estimated levels of oxygen uptake from three levels of HR are presented in Tables 6 and 7. There was a tendency towards a lower y-intercept and a greater slope in the regression equations at the retest compared to the test (Table 6). Based on calculations of all subjects, there were no systematic differences in estimated absolute levels of VO2 between test and retest. The relative differences between test and retest were 0.99 ± 11.0 (n.s.), 2.67 ± 6.48 (p<0.1) and 3.57 ± 6.24% (p<0.05) based on estimations from the lowest to the highest levels of HR (Table 7). The individual data for all tables (Tables 6–11) related to evaluations of the HR-VO2 relations are given as Supporting Information S1 Results. The 95% limits of agreement for the individual variations in the differences in estimated VO2 between test and retest varied between -0.3155 and 0.2923) (L · min -1 ) for the low HR, -0.3922 and 0.2764 for the middle HR, and -0.4735 and 0.3029 for the high HR (Fig 2).

    The y-axes show absolute differences in VO2 against the mean values of the estimations from the repeated measurements on the x-axes.

    Reproducibility of HR-VO2 regression equations and estimated levels of VO2 (model 2)

    The test and retest HR-VO2 regression equations and estimated levels of oxygen uptake from three levels of HR are presented in Tables 8 and 9. There were no significant differences between test and retest in the constituents of the regression equations (y-intercept, slope and r-value)(Table 8). Based on calculations of all subjects, there were no systematic differences in estimated absolute levels of oxygen uptake between test and retest. The relative differences between test and retest, based on estimations from three different levels of HR, were 1.09 ± 10.6, 1.75 ± 6.43 and 2.12 ± 5.92% (all n.s.)(Table 9). The 95% limits of agreement for the individual variations in the differences in estimated VO2 between test and retest varied between -0.2894 and 0.2684)(L · min -1 ) for the low HR, -0.3233 and 0.2539 for the middle HR, and -0.3649 and 0.2722 for the high HR (Fig 3).

    The y-axes show absolute differences in VO2 against the mean values of the estimations from the repeated measurements on the x-axes.

    Comparisons in regression equations and estimated VO2 between the HR-VO2 relations in model 1 and 2

    The differences between the two HR-VO2 models in the y-intercept, slope, r-value as well as in the three levels of estimated VO2 at test and retest were compared for all subjects (Tables 10 and 11). All differences between the models were small and non-significant. The mean absolute and relative differences in VO2 varied from 0.00 ± 0.04 to -0.04 ± 0.10 liter/min (all n.s.) and 0.10 ± 3.39 to -1.46 ± 3.30% (all n.s.), respectively.


    The study area was located in the natural park of El Estrecho, in Tarifa (southern Spain) on the northern shore of the Strait of Gibraltar (Fig. 1 36°07′−36°06′N, 5°45′−5°46′W). This area was the most southern protected area in Europe. It was a maritime–terrestrial park along 54 km of coastline in Andalusia and an Important Bird Area (Guerra García et al. 2009 , BirdLife International 2017 ). In this area, Ferrer et al. ( 2012 ) reported the greatest collision rates ever published for birds (1.33/turbine/yr) with the Griffon vulture being the most frequently killed species (0.41 deaths/turbine/yr). There were several Griffon vulture colonies in the area, consisting of approximately 320 breeding pairs in total. We focused on a colony at an escarpment running north–south, 4 km from the Strait of Gibraltar with approximately 65 breeding pairs (Del Moral 2009 ). Our analysis was constrained to the space used by one tagged Griffon vulture the space encompassed an area of 152 km 2 and included 20 wind farms with 269 operational turbines (Table 1).

    Model Color on Fig. 1 No. of turbines Hub height Blade length Total height
    ECOTECNIA ECO-74 34 70 35.5 105.5
    ENERCON E-70 20 84 33.5 117.5
    GAMESA G-80 30 67 40.0 107.0
    GAMESA G-87 11 78 42.3 120.3
    MADE AE-56 43 60 27.3 87.3
    MADE AE-59 55 60 28.8 88.8
    VESTAS V-72 4 78 36.0 114.0
    VESTAS V-80 6 78 40.0 118.0
    VESTAS V-90 66 80 44.0 124.0


    As described by 2016, Enquist et al. ( 2016 , in prep.), the generation of the BIEN database consists of a linked workflow that (1) standardizes taxonomy by correcting spelling of species names and updating synonyms to currently accepted names via the Taxonomic Name Resolution Service or TNRS (Boyle et al., 2013 ) (2) detects and flags observations with erroneous geographic coordinates and (3) flags cultivars and non-native species via the Native Species Resolver ( Coordinates were flagged as erroneous if they fell outside the specified political regions, if the latitude was exactly 0 or 90 degrees, the longitude was exactly 0 or 180 degrees, or if the point fell in the ocean. The detection of cultivars and non-native records relies on native species lists, which are not available throughout the New World, so this filtration is imperfect.

    Range maps were built for each species using a method determined by the number of observations of that species. A species with a single record was assigned a range that included only the 100 km 2 cell where it was found. Ranges for species with 2–3 records were rectangular bounding boxes with the limits set by the minimum and maximum latitude and longitude of all occurrences. Ranges for species with 4–9 records were built with convex hulls (the minimum-fitting polygon that encompasses all occurrences of that species). For species with >9 records, we built species distribution models using the Maxent algorithm (Phillips, Anderson, & Schapire, 2006 ). Only one occurrence record per cell (in cases of multiple records) was used for Maxent model building. Maxent model building generally followed the recommendations outlined in (Merow, Cory, & Silander, 2014 Merow, Cory, Smith, & Silander, 2013 ). Model settings were chosen to balance overfitting, which underestimates range sizes, with underfitting, which results in excessively smooth models that over predict range size. Only linear, quadratic, and product features were used and regularization was set at the default value.

    Environmental predictors for SDMs were obtained from the WorldClim current (1960–1990) climate data at 10 arc-minute resolution (Hijmans, Cameron, Parra, Jones, & Jarvis, 2005 ) and resampled to a 10 km resolution. Predictors included mean annual temperature, mean diurnal temperature range, annual precipitation, precipitation seasonality, precipitation in warmest quarter/(precipitation in warmest quarter + precipitation in coldest quarter), and five spatial eigenvectors (De Marco, Diniz-Filho, & Bini, 2008 ). The spatial eigenvectors essentially captured large-scale regional differences in occurrence and primarily served as broad-scale dispersal limitation of species ranges, limiting predictions far in geographic space from presence locations (Blach-Overgaard, Svenning, Dransfield, Greve, & Balslev, 2010 ).

    Maxent's continuous predictions were converted to binary presence/absence predictions by choosing a threshold based on the 75th percentile of the cumulative output.

    We constructed a phylogeny of 18,641 species based on a standardised list of New World species and the gene regions atpB-rbcL, ndhF, psbA, psbA-psbH, rbcL and trnT-trnL-trnF marker regions, using the software PHLAWD (Smith, Beaulieu, & Donoghue, 2009 ). The phylogeny was constructed with RAxML (7.3.0 Stamatakis, 2006 ) with an unconstrained ML search and divergence times were estimated using penalised likelihood and the tree PL software package (Smith & O'Meara, 2012 ). Additional details on the methodology used to extract these data from GenBank and align them are presented in Hinchliff and Smith ( 2014 ). Onto this phylogeny, we grafted the additional taxa from the BIEN dataset using taxonomy (genus membership) as a guide for the remaining c. 72,000 species. This grafting was repeated to create a set of 100 phylogenies to account for uncertainty in placement of species without genetic information. Additional information on the BIEN phylogenies is available online (


    3.1 PRF maintains viability and increases proliferation of macrophages

    To investigate the impact of soluble extracts of PRF membranes on the cell viability, an MTT assay reflecting the NAD(P)H-dependent formazan production was carried out. Concentrations below 10% of PRF notably increased formazan production in primary macrophages and RAW264.7 cells (Figure 1A). Further, 30% PRF lysates did not affect the viability of freshly isolated murine bone marrow cells (data not shown). Cell viability was further confirmed by live-dead staining in RAW264.7 cells (Figure 1B). In addition, PRF further enhanced BrdU incorporation indicating an increased proliferation of RAW264.7 cells (see Supporting Information Table 2 in online Journal of Periodontology). Moreover, PRF tended to reduce the expression levels of pro-apoptotic Bax and caspase-3 along with the anti-apoptotic marker gene B cell lymphoma-2 (BCL2L1) (see Supporting Information Table 3 in online Journal of Periodontology). This tendency was further supported by exposing RAW264.7 cells to 30% PRF. PRF lysates suppressed the basal levels of cleaved caspase-3 (see supplemental Figure 1 in online Journal of Periodontology) and caused a weak reduction of caspase-3 activity (see Supporting Information Figure 1 in online Journal of Periodontology). Altogether, these results reveal that soluble extracts of PRF membranes maintain viability and increase proliferation of primary macrophages.

    3.2 PRF reduces the expression of TRAP and Cathepsin K

    To determine the most appropriate experimental condition, we evaluated the effect of various concentrations of PRF on gene expression. Murine bone marrow cells were incubated with different concentrations of soluble extracts of PRF membranes in the presence of RANKL and M-CSF. Dose-response curves revealed a suppression of the osteoclast marker genes TRAP and Cathepsin K by the addition of PRF (Figure 2A). Next, murine bone marrow cells were incubated with the same concentrations of PRF but in the presence of RANKL, M-CSF, and TGF-β. Again, PRF reduced the expression of the osteoclast marker genes TRAP and Cathepsin K in a dose-dependent manner (Figure 2B). Taken together, these results suggest that 50% of soluble extracts of PRF membranes is a suitable concentration to substantially reduce the aforementioned osteoclast marker genes.

    3.3 PRF reduces osteoclast differentiation in vitro

    To further examine the potential role of PRF in osteoclast differentiation, murine bone marrow cells were grown in the presence of 50% soluble extracts of PRF membranes with RANKL and M-CSF. We report here that PRF strongly decreased the number of multinucleated cells staining positive for TRAP (Figures 3A and 3B) and also the respective number of nuclei per cells (see Supporting Information Table 4 in online Journal of Periodontology). In line with this observation, PRF decreased the expression of TRAP and Cathepsin K, both enzymes required for bone resorption (Figure 3C). In addition, the other osteoclast marker genes DCSTAMP, NFATc1 and OSCAR were also suppressed by soluble extracts of PRF membranes (Figure 3C). Altogether, these observations indicate that PRF can inhibit osteoclastogenesis.

    3.4 PRF reduces osteoclast differentiation induced by RANKL, M-CSF, and TGF-β

    To trigger osteoclastogenesis even further, TGF-β was added to RANKL and M-CSF as previously described. 20 As expected, the addition of TGF-β markedly increased osteoclastogenesis compared to the RANKL and M-CSF cultures. 20 Most notably, soluble extracts of PRF membranes at 50% concentration significantly reduced osteoclastogenesis under these conditions as indicated by the reduced number of multinucleated TRAP positive cells (Figures 4A and 4B) and reduced number of nuclei per cells (see Supporting Information Table 4 in online Journal of Periodontology). Moreover, PRF lysates decreased the expression of TRAP and Cathepsin K as well as DCSTAMP, NFATc1, and OSCAR (Figure 4C). This suppression of osteoclastogenesis by PRF lysates was further validated by pit formation assay (see Supporting Information Figure 2 in online Journal of Periodontology). Taken together, these findings suggest that PRF suppresses osteoclastogenesis independent of the bioassay.

    3.5 PRF cannot reverse osteoclastogenesis at later stages

    Considering that PRF reduced the differentiation of osteoclasts from their progenitors, the question then arises whether PRF can reverse this process. To answer this question, murine bone marrow cells were grown in the presence of RANKL and M-CSF. After 72 hours, PRF was added to the cells for another 72 hours. As shown in Figures 5A and 5D, large multinucleated cells stained for TRAP were observed regardless of whether PRF was added. Furthermore, PRF was not able to reduce the number of osteoclasts (Figures 5D and 5F) neither the numbers of nuclei per cell (data not shown). Moreover, PRF was unable to significantly change the gene expression of the osteoclast marker genes (Figure 5C), also in the presence of TGF-β (Figure 5F). These findings indicate that PRF is unable to reverse osteoclastogenesis when the process has started.

    3.6 Growth factors naturally released by PRF decrease osteoclastogenesis

    Finally, to simulate the natural release of growth factors from PRF membranes, membranes were transferred into culture medium. After 24 and 72 hours the conditioned medium was collected. 18 Our data show that PRF conditioned medium, independent of the harvesting time, decreased osteoclastogenesis in the presence of RANKL, M-CSF, and TGF-β, indicated by TRAP staining (not shown) and the expression of TRAP and Cathepsin K (Figures 6A and 6B). Altogether, these observations suggest that PRF releases an activity that decreases osteoclastogenesis.

    Vesicular- or vacuolar-type adenosine triphosphatases (V-ATPases) are multi-component, ATP-driven proton pumps, which play important roles in many physiological processes by acidifying intracellular vesicles, organelles, and the extracellular milieu. Long-standing challenges in purifying mammalian V-ATPases have limited the biochemical and structural study of mammalian V-ATPase. Here, we provide a protocol for purifying milligrams of human V-ATPase and detail procedures for the reconstruction of its structure by cryo-EM. Our method can be applied to any biochemical and biophysical study of human V-ATPase.

    For complete details on the use and execution of this protocol, please refer to Wang et al. (2020).

    Watch the video: RCF and RPM in Centrifugation. Explained Mathematically (February 2023).