Strange India All Strange Things About India and world

Optimization of RSG incubation length

According to the manufacturer (Thermo Fisher Scientific), oxidoreductases convert RSG to a green fluorescent product inside live cells, which can be quantified using microscopy or flow cytometry. Compared with a functionally similar and previously utilized tetrazolium probe CTC44, RSG has superior fluorescence spectral properties, a shorter incubation time and no known toxic effects11. The BacLight RedoxSensor Green Vitality Kit manual (Thermo Fisher Scientific) recommends a 10 minsample incubation with RSG before proceeding with cell analyses. To test the impact of incubation length on the abundance and fluorescence intensity of marine prokaryoplankton cells, we performed a test on seawater collected from coastal GoM (43.8609° N, 69.5782° W) on 11 August 2016. Two replicate surface seawater samples were collected in sterile 50 ml polypropylene tubes and transported to the lab at in situ temperature for further processing. These samples were amended with the manufacturer’s recommended concentration of RSG (1 µM final concentration) and incubated at in situ temperature in the dark for a variable length of time before being analysed using the BD Influx Mariner flow cytometer (Becton Dickinson, formerly Cytopeia) equipped with a 488 nm laser for excitation and a 70 μm nozzle orifice. RSG-positive cells were gated on the basis of particle green fluorescence (531/40 bandpass), forward scatter and the ratio of green versus red fluorescence (692/40 bandpass; for improved discrimination of cells from detrital particles). We found that both the abundance and the fluorescence intensity of RSG-positive cells reached their maximum values and remained stable after 30–70 min of incubation (Extended Data Fig. 3f). On the basis of these results, subsequent RSG incubations were set for 30 min.

Calibration of RSG fluorescence against O2 respiration using microbial cultures

Isolates of phenotypically distinct species, Shewanella loihica45, Roseobacter denitrificans46, Pseudomonas stutzeri47, Ruegeria pomeroyi48 and Vibrio alginolyticus49, were acquired from the National Center for Marine Algae and Microbiota. Isolates of Rhodoluna lacicola50 and Oligotropha carboxidovorans51 were obtained from the German Collection of Microorganisms and Cell Cultures. Cultures were grown in specific media at varying temperatures and nutrient concentrations (medium recipes and growth conditions are provided in Supplementary Table 1) in 70 ml glass serum vials sealed with butyl rubber stoppers with atmospheric air in the headspace (starting O2 concentration 21%) on a rotating shaker table to facilitate complete gas exchange between the culture medium and vessel headspace. A headspace approach was used owing to the need to withdraw small volumes of fluids at discrete timepoints.

The oxygen concentration in all culture vessels was measured using the Firesting-O2 optical meter according to the manufacturer’s instructions (PyroScience sensor technology). For each serum bottle, the optode probe was inserted into the headspace with the temperature-normalizing probe also inserted into a different bottle headspace (with seawater in the liquid phase) kept at the same temperature as the culture undergoing analysis. The oxygen concentration was measured in the headspace and converted to dissolved oxygen concentration in the culture medium, assuming complete gas equilibration between the culture medium and the culture headspace (as all cultures were kept on a rotating shaker table) and taking into account temperature, headspace pressure and salinity of the culture medium52. This conversion was possible due to the Firesting-O2 optode method of directly measuring O2 at the tip of the optode probe (that is, there is no diffusion or consumption of O2 via this method). After measuring O2 in the headspace, the concentration present in the culture medium was calculated using the headspace equilibration technique and calculated53 using Bunsen coefficients54,55 for O2, which takes into account temperature and salinity at constant atmospheric pressure at sea level. Concentrations of O2 in the culture medium were measured every few hours on the basis of the estimated doubling time of each culture type. Measurements of O2 concentration for each culture spanned initial inoculation through the log phase of growth until the stationary phase was reached. Operational changes in microbial cell abundance were determined on the basis of measurements of optical density at a wavelength of 600 nm using a spectrophotometer. A culture was determined to be in the stationary phase once the O2 concentration no longer decreased exponentially, and when the cell abundance stopped increasing exponentially. During the stationary phase, the rates of O2 respiration were calculated as time-normalized differences in O2 concentration between measurements. To estimate the rates of O2 respiration per average cell, the abundance of microbial cells was determined by flow cytometry analysis of subsamples of the culture material taken at the same time as the O2 concentrations were measured, as described below.

After the cultures reached stationary phase, 1 ml subsamples were withdrawn using a sterile needle and syringe, transferred to a 2 ml cryovial and amended with 1 µl RSG stock solution. After a gentle mix, the cryovials were incubated for 30 min at the same temperature as their source cultures. The cryovials were then amended with 5% glycerol and 1× pH 8 Tris-EDTA buffer (final concentrations), flash-frozen in liquid nitrogen and stored at −80 °C until downstream use in flow cytometry analysis. After RSG-incubated samples were taken and while the culture was still in the stationary phase, killed controls for each culture type were prepared by autoclaving 10 ml of culture and medium for 30 min at 121 °C. Samples of uninoculated medium and killed controls were also taken for use in determining appropriate gates for downstream flow cytometry analysis.

After thawing, the cryopreserved samples were analysed using the ZE5 Cell Analyzer flow cytometer (Bio-Rad). The samples were analysed with blue excitation (488 nm), and the instrument was triggered on both forward scatter and green fluorescence using a 525/35 nm bandpass filter. The analysis gate was defined on the basis of particle green (RSG) fluorescence and right-angle side scatter. The analysis gates were set to eliminate areas that contained only non-cellular particle noise from uninoculated media blanks, and also by eliminating areas that contained dead cells (killed by autoclaving). For all culture samples, an area of growing cells was clearly defined (Extended Data Fig. 2b). Cell abundance and green fluorescence were analysed in FlowJo v.10.6.2 (Becton Dickinson).

RSG fluorescence normalization

To account for day-to-day drift and differences between the ZE5 Cell Analyzer flow cytometer (Bio-Rad) and the BD Influx Mariner flow cytometer (Becton Dickinson, formerly Cytopeia) cytometry instruments and settings, a procedure was developed to normalize cell green fluorescence using one of two fluorescent particle size standard kits: NFPPS-52-4K or RCP-30-5A (Spherotech). These beads were analysed using flow cytometry each day when RSG-labelled cells were analysed, using the same instrument and settings. For each bead, the geometric mean of the 525/35 fluorescence was calculated using FlowJo. For the NFPPS-52-4K bead kit, the least-fluorescent bead was assigned a normalized fluorescence value equal to 1 and the normalized fluorescence values of the other beads (Supplementary Table 5) were calculated as the average ratio of their fluorescence relative to the fluorescence of the least-fluorescent bead. Averages of these ratios were estimated from the application of this technique on multiple days.

For each day of analysis of field and culture samples, a linear regression was performed to establish the relationship between the measured bead fluorescence and their normalized fluorescence values. Then the normalized fluorescence of each cell was estimated as follows:

$$\text{Normalized fluorescence}\,=\,m\times x+b$$

where x is a cell’s measured fluorescence, m is the experimentally determined slope of the bead linear regression from the day of cell analysis and b is the experimentally determined intercept of the bead linear regression from the day of cell analysis.

To intercalibrate the two bead sets, they were flow cytometrically co-analysed on multiple days on the BD Influx Mariner flow cytometer and the normalized fluorescence of the geometric mean of each RCP-30-5A bead was calculated using the linear regression equation described above. A standardized normalized fluorescence value was calculated by averaging the geometric mean of RCP-30-5A beads from multiple days (Supplementary Table 5 (column RCP-30-5A)). The standardized normalized fluorescence of the RCP-30-5A beads was used to perform a linear regression analysis to establish the relationship between the measured bead fluorescence and their normalized fluorescence (see above) on days on which the NFPPS-52-4K beads were not available.

Subsequently, we correlated the normalized fluorescence of microbial cultures against their average respiration rate (Fig. 1a), resulting in the following exponential relationship:

$${\rm{Cell\; respiration}}\,=\,0.004\times {e}^{0.194\times {\rm{normalized\; f}}{\rm{luorescence}}}$$

where the cell respiration is the cell-specific respiration estimate in fmol O2 per cell per h and the normalized fluorescence is the cell’s normalized RSG fluorescence (see above). This equation was transformed for ease of interpretation in Fig. 1a.

Validation of RSG-based prokaryoplankton respiration rate measurements

To validate the use of RSG in quantifying respiration, we performed simultaneous RSG probing and bulk respiration measurements using traditional Winkler methodology14 on geographically diverse, coastal and open ocean prokaryoplankton samples under in situ conditions (Supplementary Table 2). For GoM samples, seawater was collected at a depth of 1 m using a Niskin bottle at the same location as the RSG samples described above, except for one sample (Damariscotta River), the location of which is given in Supplementary Table 2. The seawater was collected, passed through a 40-µm-mesh filter and used to fill eight biological oxygen demand (BOD) bottles without headspace. Each flask was overflown by at least 3 times its volume to avoid air bubbles and to omit any additional gas exchange the water would have had with the atmosphere during transfer into the vessel. Four of these bottles were immediately fixed56, while the remaining four bottles were incubated at in situ temperature in the dark for 24 h before fixation56. The oxygen concentration in all of the bottles was analysed using an amperometric titrator56.

Open ocean samples were collected directly from the Niskin bottles into BOD flasks, whereas coastal Mediterranean Sea samples were collected with buckets from the shore and transferred to acid-cleaned polycarbonate carboys. Once at the laboratory, the coastal Mediterranean Sea water was filtered through 0.8 µm polycarbonate filters and transferred to BOD flasks. Each flask was overflown by at least three times its volume. The three replicate BOD flasks were fixed immediately, and another three flasks were incubated in the dark at in situ temperature for 24 h. After fixation with the Winkler chemicals, the samples were measured using the potentiometric method56 (coastal Mediterranean and North Atlantic) or spectrophotometry57 (south and equatorial Atlantic). Winkler-based prokaryoplankton respiration rates were estimated as the difference in dissolved oxygen concentration between initial and final samples.

Immediately after collecting the field sample, subsamples were incubated with RSG at in situ temperature in the dark for 30 min, then stored at −80 °C until flow cytometry analyses and estimates of cell-specific respiration rates as described above. The RSG-based bulk respiration rates were obtained by multiplying the average, cell-specific respiration rates (described above) by the abundance of respiring cells in each sample.

Single-cell genome and cell diameter analyses of marine prokaryoplankton

For the main study, samples from the GoM were collected from the dock of Bigelow Laboratory for Ocean Sciences on the shore of East Boothbay at the mouth of the Damariscotta River (43.8609° N, 69.5782° W) on six days from April 2017 to July 2019 (Supplementary Table 2). Water samples were collected from a depth of 1 m with a 5 l Niskin bottle, passed through a 50-µm-mesh filter, transferred to acid-washed polycarbonate bottles, transported to the lab at in situ temperature and processed immediately. Open ocean samples were collected with 12 l Niskin bottles attached to a CTD rosette and transferred to acid-washed polycarbonate bottles. Subsequently, duplicate 1 ml seawater subsamples were transferred to cryovials, amended with 1 µl RSG (1 µM final concentration), incubated in the dark at in situ temperature for 30 min, amended with 5% glycerol and 1× pH 8 TRIS-EDTA buffer (final concentrations), flash-frozen in liquid N2 and stored at −80 °C. In all cases, additional 1 ml seawater subsamples were amended with 5% glycerol and 1× pH 8 TRIS-EDTA buffer (final concentrations), flash frozen in liquid N2 and stored at −80 °C without RSG addition.

To compare day and night conditions, additional GoM seawater samples were collected from the Bigelow Laboratory dock according to the same procedures as described above on 24 (14:00 EST) and 25 (02:00 EST) August 2021. Triplicate 1 ml sample aliquots were labelled with RSG and cryopreserved as described above.

For sorting of respiring cells, we used samples that were labelled with RSG in the field, as described above. For non-specific sorting of prokaryoplankton cells, the cryopreserved and thawed seawater samples were incubated with the SYTO 9 nucleic acid stain (5 µM final concentration; Thermo Fisher Scientific) on ice for 10–60 min. Flow cytometry analysis and sorting was performed using a BD InFlux Mariner flow cytometer equipped with a 488 nm laser for excitation and a 70 μm nozzle orifice (Becton Dickinson, formerly Cytopeia). The cytometer was triggered on forward scatter in most cases and on green fluorescence in August 2021. The ‘single-1 drop’ mode was used for maximal sort purity. The sort gate was defined based on particle green fluorescence (proxy to either nucleic acid content or respiration rate), forward scatter (proxy to cell diameter) and the ratio of green versus red fluorescence (for improved discrimination of cells from detrital particles). Cells were deposited into 384-well microplates containing 600 nl per well of 1× TE buffer and stored at −80 °C until further processing. Of the 384 wells, 317 wells were dedicated for single cells, 64 wells were used as negative controls (no droplet deposition) and 3 wells received 10 cells each to serve as positive controls. The accuracy of droplet deposition into microplate wells was confirmed several times during each sort day by sorting 3.46 µm diameter SPHERO Rainbow Fluorescent Particles (Spherotech) and examining their presence at the bottom of each well using microscopy. In these examinations, fewer than 2% of wells did not contain beads and <0.4% wells contained more than one bead.

Index sort data were collected using the BD Sortware software. The following laboratory cultures were used in the development of a cell diameter equivalent calibration curve: Prochlorococcus marinus CCMP 2389; Microbacterium sp.; Pelagibacter ubique HTCC1062; and Synechococcus CCMP 2515. The average cell diameters of these cultures were determined using the Multisizer 4e Coulter Counter (Beckman Coulter). The average forward scatter of each of the four cultures was determined using the same flow cytometry settings used during environmental sample sorting, repeated on each day of single-cell sorting. We observed a strong correlation between cell diameters and forward scatter (FSC) among these cultures16. Taking advantage of this correlation, the estimated diameter of the sorted environmental cells (diameter, in µm) was estimated from a log-linear regression model:

$${\rm{diameter}}={10}^{(a\times {\log }_{10}[{\rm{FSC}}]-b)},$$

where a and b are empirically derived regression coefficients from each day of sorting16, and FSC is the measured forward scatter. This calculation was repeated for all flow cytometry cell sorting sessions. The proxy cell volume was calculated assuming a spherical shape for all lineages.

Before genomic DNA amplification, cells were lysed and their DNA was denatured by two freeze–thaw cycles and the addition of 700 nl of a lysis buffer consisting of 0.4 M KOH, 10 mM EDTA and 100 mM dithiothreitol, and a subsequent 10 min incubation at 20 °C. The lysis was terminated by the addition of 700 nl of 1 M Tris-HCl, pH 4. Single-cell whole-genome amplification was performed using WGA-X16. In brief, the 10 µl WGA-X reactions contained final concentrations of 0.2 U µl−1 Equiphi29 polymerase (Thermo Fisher Scientific), 1× Equiphi29 reaction buffer (Thermo Fisher Scientific), 0.4 µM each dNTP (New England BioLabs), 10 µM dithiothreitol (Thermo Fisher Scientific), 40 µM random heptamers with two 3′-terminal phosphorothioated nucleotide bonds (Integrated DNA Technologies) and 1 μM SYTO 9 (Thermo Fisher Scientific). These reactions were performed at 45 °C for 12–16 h, then inactivated by a 15 min incubation at 75 °C. To prevent WGA-X reactions from being contaminated with non-target DNA, all cell lysis and DNA amplification reagents were treated with ultraviolet light in a Stratalinker (Stratagene)58 system. An empirical optimization of the ultraviolet exposure was performed to determine the length of ultraviolet exposure that is necessary to cross-link all detectable contaminants without inactivating the reaction. Cell sorting, lysis and WGA-X set-up were performed in a HEPA-filtered environment conforming to Class 1000 cleanroom specifications. Before cell sorting, the instrument, the reagents and the workspace were decontaminated for DNA using ultraviolet irradiation and sodium hypochlorite solution59. To further reduce the risk of DNA contamination, and to improve accuracy and throughput, Bravo (Agilent Technologies) and Freedom Evo (Tecan) robotic liquid handlers were used for all liquid handling in 384-well plates.

Libraries for SAG genomic sequencing were created with Nextera XT (Illumina) reagents according to the manufacturer’s instructions except for the purification steps, which were performed using column clean-up kits (QIAGEN) and library size selection, which was performed with BluePippin (Sage Science) with a target size of 500 ± 50 bp. DNA concentration measurements were performed using Quant-iT dsDNA Assay Kits (Thermo Fisher Scientific) according to the manufacturer’s instructions. Libraries were sequenced with either the NextSeq 500 (Illumina) system in 2 × 150 bp mode or the NextSeq 2000 (Illumina) system in 2 × 100 bp mode. The obtained sequence reads were quality-trimmed using Trimmomatic (v.0.32)60 using the following settings: -phred33 LEADING:0 TRAILING:5 SLIDINGWINDOW:4:15 MINLEN:36. Reads matching the Homo sapiens reference assembly GRCh38 and a local database of WGA-X reagent contaminants16 (≥95% identity of ≥100 bp alignments) as well as low-complexity reads (containing <5% of any nucleotide) were removed. The remaining reads were digitally normalized using kmernorm v.1.05 ( using the settings -k 21 -t 30 -c 3 and then assembled with SPAdes (v.3.0.0)61 using the following settings: –careful –sc –phred-offset 33. Each end of the obtained contigs was trimmed by 100 bp and only contigs longer than 2,000 bp were retained. Contigs matching the H. sapiens reference assembly GRCh38 and a local database of WGA-X reagent contaminants16 (≥95% identity of ≥100 bp alignments) were removed. The quality of the resulting genome assemblies was determined using CheckM (v.1.0.7)62 and tetramer frequency analysis63. This workflow was evaluated for assembly errors using three bacterial benchmark cultures with diverse genome complexity and percentage GC, indicating no non-target and undefined bases in the assemblies and the following average frequencies of misassemblies,indels and mismatches per 100 kb: 1.5, 3.0 and 5.0, respectively (ref. 16). The 153 SAGs in which potentially contaminating DNA sequences were detected were removed from the dataset. This resulted in 7,518 curated SAG genome assemblies.

Pairwise SAG average amino acid identity was estimated using CompareM64. The 16S rRNA gene regions longer than 500 bp were identified using local alignments provided by BLAST against CREST’s65 curated SILVA reference database SILVAMod (v.128)66 and classified using a reimplementation of CREST’s last common ancestor algorithm. Phylogenetic trees were generated from near-complete (>1,400 bp) SAG 16S rRNA genes by first producing alignments using the SINA aligner (v.1.2.11)67 and then inferring maximum-likelihood phylogenetic relationships in MEGACC (v.10.2.4)68 with 100 bootstraps. The trees were annotated and visualized in iTOL69. Taxonomic assignments of SAGs were obtained with GTDB-Tk (v.1.4.1)70.

Functional annotation was first performed using Prokka71 with the default Swiss-Prot databases supplied by the software. Prokka was run a second time with a custom protein annotation database built from compiling Swiss-Prot72 entries for Archaea and Bacteria. Taxonomic assignments were obtained using GTDB-Tk (v.1.4.1)70. Viral sequences (whole or partial contigs) were independently identified using three pre-existing bioinformatic viral identification tools: ViruScope73 (virus_prob > 0.95), VirSorter274 (category 1 and 2 only) and DeepVirFinder75 (P > 0.95). The results of these three tools were combined. False-positives (P < 0.95) were removed using CheckV76.

Metagenomics and metatranscriptomics

Microbial biomass from 100–200 ml seawater aliquots was collected on Supor 0.2 µm sterilized filters (Pall Corporation) by vacuum filtration (max pressure of 15 psi), then flash-frozen in liquid nitrogen and stored at −80 °C until nucleic acid extraction. MTST5 and MTST6 RNA standards were added to each filter before RNA extraction to an estimated concentration of 0.1–1% total yield, as previously described31. DNA and RNA were extracted in parallel using the Zymobiomics DNA/RNA Miniprep kit (Zymo Research) and stored at −80 °C until further processing. RNA samples were further cleaned and concentrated using the RNA Clean & Concentrator kit (Zymo Research). cDNA libraries for Illumina shotgun sequencing were generated with the KAPA RNA HyperPrep (Roche Sequencing). Libraries for DNA shotgun sequencing were generated using the Nextera XT kit (Illumina) according to the manufacturer’s instructions. Both library types were size-selected with BluePippin (Sage Science) set for 370 bp ‘tight’ and quantified using the Tapestation (Agillent). These libraries were sequenced using the NextSeq 500 (Illumina) system in 2 × 150 bp mode.

After sequencing, reads from both RNA and DNA libraries were initially trimmed with Trimmomatic (settings: -phred33 LEADING:0 TRAILING:5 SLIDINGWINDOW:4:15 MINLEN:75). Reads matching the H. sapiens reference assembly GRCh38 and a local database of WGA-X reagent contaminants16 (≥95% identity of ≥100 bp alignments), as well as low-complexity reads (containing <5% of any nucleotide) were removed. The remaining reads were merged with bbmerge using the following settings: k = 40, extend2 = 60, iterations = 5, loose=‘t’, qtrim2=‘t’. The merged reads were taxonomically and functionally annotated using the Kaiju classifier77, after replacing the underlying reference genome database with SAGs from this study as previously described3. To identify SSU rRNA gene transcripts, metatranscriptomic reads were mapped on the SILVA SSU library66 (v.132) using Bowtie278 at 95% identity across the length of the read. The abundance of each genus in a sample (GAi; cells per ml) was calculated as follows:


where e is a normalization factor for genera that did not generate SAGs, and is defined as:

$$e=\mathop{\sum }\limits_{i=1}^{N}\left[(\left(\frac{{a}_{i}}{{b}_{i}}\right)\frac{1}{c})\frac{1}{{d}_{i}}\right],$$

where ai is the count of metagenome reads recruited to genus i SAG, bi is the average fraction of nucleotides in genus i SAG assemblies that encode proteins, c is the total count of reads in the metagenome, di is the average estimated genome size of the genus i SAG and f is the total abundance of prokaryoplankton, in cells per ml. This calculation was repeated for every sampling day.

The average count of transcripts per cell (TCi) was calculated for each genus and gene as follows:

$${{\rm{TC}}}_{i}={a}_{i}\times \left(\frac{b}{c}\right)\times \left(\frac{1}{d}\right)\times \left(\frac{1}{{{\rm{GA}}}_{i}}\right),$$

where ai  is the count of reads mapped on the gene of interest, b is the number of MTST5 and MTST6 RNA transcripts added to the sample, c is the number of reads mapped on MTST5 and MTST6 and so the ratiob/is the number of transcripts per read, d is volume of sample collected on the filter in ml and GAiis the genus abundance (cells per ml; see above). This calculation was repeated for every sampling day.

Genus-specific respiration estimates

The average cellular respiration rate (ACRi; amol O2 per cell per h) of each genus in each GoM seawater sample was estimated as follows:

$${{\rm{ACR}}}_{i}=\left(\frac{{a}_{i}R+0.5\times z[{{\rm{GA}}}_{i}-R]}{{{\rm{GA}}}_{i}}\right)$$

 where R is defined as:

where ai is the average respiration rate of a given genus’ cells recovered in SAGs with estimated respiration rates exceeding the detection limit (4 amol O2 per cell per h), bi is the count of SAGs classified as belonging to genus i (that is, the number of active cells in genus i), c is the count of all RSG SAGs (also known as the number of sequenced active cells in sample), so bi/c is the fraction of active cells in genus i, d is the total concentration of active cells with an estimated respiration rate above the detection limit (cells per ml; Supplementary Table 2) and z is equal to the minimum detected cell-specific respiration rate, in amol O2 per cell per h.

The quantity R is the number of active cells of genus i per ml. The quantity 0.5 × z[GAi − R] is a correction factor for the cells below the detection limit. Owing to different sensitivities of the measurement methods, it was possible for R to occasionally be slightly greater than GAi. When this happened, R was set equal to GAi making GAi − R = 0. This calculation was repeated for every sampling day. This calculation was only made for those genera and samples for which at least 3 RSG-labelled SAGs were identified.

Water chemistry analyses

For major nutrient concentration measurements, 100 ml subsamples of GoM seawater were collected from the same samples that were used in single-cell genomics and respiration analyses. These samples were immediately filtered through a nucleopore track-edge membrane 25 mm, 0.4 µm pore-size filter (Whatman) and stored at −80 °C until analysis of NO3, NO2, NH4+ and PO43 on a SEAL AA3 (Seal Analytical Limited). Moreover, triplicate seawater subsamples, 100 ml each, were collected from each field sample and vacuum-filtered onto glass microfiber filters (Whatman, GF/F) within 2 h. Each GF/F filter was extracted in 10 ml of 90% acetone at −20 °C for 24–48 h. The extracts were analysed on the 10-AU fluorometer (Turner Designs) before and after the addition of 50 µl of 10% HCl, and the concentrations of chlorophyll and phaeophytin were determined as described previously79

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Source link


Leave a Reply

Your email address will not be published. Required fields are marked *