(C) PLOS One
This story was originally published by PLOS One and is unaltered.
 . . . . . . . . . .



Supra-orbital whiskers act as wind-sensing antennae in rats [1]

['Matias Mugnaini', 'Neural Systems', 'Behavior', 'Marine Biological Laboratory', 'Woods Hole', 'Massachusetts', 'United States Of America', 'Laboratory Of Physiology', 'Algorithms Of The Brain', 'Leloir Institute']

Date: 2023-07 

We know little about mammalian anemotaxis or wind sensing. Recently, however, Hartmann and colleagues showed whisker-based anemotaxis in rats. To investigate how whiskers sense airflow, we first tracked whisker tips in anesthetized rats under low (0.5 m/s) and high (1.5 m/s) airflow. Whisker tips showed increasing movement from low to high airflow conditions, with all whisker tips moving during high airflow. Low airflow conditions—most similar to naturally occurring wind stimuli—engaged whisker tips differentially. Most whiskers moved little, but the long supra-orbital (lSO) whisker showed maximal displacement, followed by the α, β, and A1 whiskers. The lSO whisker differs from other whiskers in its exposed dorsal position, upward bending, length and thin diameter. Ex vivo extracted lSO whiskers also showed exceptional airflow displacement, suggesting whisker-intrinsic biomechanics mediate the unique airflow-sensitivity. Micro computed tomography (micro-CT) revealed that the ring-wulst—the follicle structure receiving the most sensitive afferents—was more complete/closed in the lSO, and other wind-sensitive whiskers, than in non-wind-sensitive whiskers, suggesting specialization of the supra-orbital for omni-directional sensing. We localized and targeted the cortical supra-orbital whisker representation in simultaneous Neuropixels recordings with D/E-row whisker barrels. Responses to wind-stimuli were stronger in the supra-orbital whisker representation than in D/E-row barrel cortex. We assessed the behavioral significance of whiskers in an airflow-sensing paradigm. We observed that rats spontaneously turn towards airflow stimuli in complete darkness. Selective trimming of wind-responsive whiskers diminished airflow turning responses more than trimming of non-wind-responsive whiskers. Lidocaine injections targeted to supra-orbital whisker follicles also diminished airflow turning responses compared to control injections. We conclude that supra-orbital whiskers act as wind antennae.

Funding: This work was supported by the Marine Biological Laboratory (MM, DM, FD, VS, ARM, MB, AC), a training grant from the NIMH (R25MH059472; MM, DM, FD, VS, ARM), Humboldt Universität zu Berlin, the Bernstein Center for Computational Neuroscience Berlin (MC, BG, MB), the German federal ministry of education and research (MB). AC is supported by the Simons Initiative for the Developing Brain, the University of Edinburgh and a Simons ESAT fellowship. ARM was supported by QuantOCancer and The Grass Foundation and DM was supported by The Grass Foundation to attend the Neural Systems & Behavior Course (NS&B). FD was supported by the Stanley W. Watson Education Fund to attend NS&B. MM was supported by an IBRO-USCRC Fellowship to attend NS&B. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

We find that whiskers differ markedly in their airflow responses. In particular, the supra-orbital whiskers respond distinctly when low airflow stimuli are applied, and such airflow responses reflect the specific whisker biomechanics of the supra-orbital whiskers. Micro computed tomography (micro-CT) revealed follicular differences in supra-orbital and pad whiskers. Recordings with Neuropixels probes show increased wind response in the supra-orbital versus pad barrel field. Finally, rats can sense and localize weak airflow stimuli and such abilities are diminished by selective whisker trimming of wind-sensitive whiskers or by blocking supra-orbital whiskers.

The so-called supra-orbital whiskers above the eye are of obvious interest in wind sensing due to their exposed anatomical positioning. Understanding of whisker function comes from understanding how whiskers interact in the environment [ 16 , 17 ]. Our analysis of whisker diversity in wind sensing took advantage of recent progress in automated animal tracking, specifically of the DeepLabCut toolbox [ 18 , 19 ]. We asked the following questions: (i) Which whiskers react maximally to airflow stimuli? (ii) Are whisker airflow responses dependent on whisker biomechanics and substructure? (iii) How do mechanical whisker airflow responses relate to the cortical barrel map? (iv) How do whiskers contribute differentially to airflow sensitivity?

Our work was inspired by the whisker-anemotaxis shown by Hartmann and colleagues. Rather than focus on the 5 rows of mystacial whiskers, which are represented in the famous posteromedial-barrel-subfield [ 9 ], we decided to assess the role of all facial whiskers in anemotaxis. The decision to look across different whisker subfields was based on our experience that whisker subfields may have very different functional characteristics. The submandibular whisker trident, for example [ 10 ], is a three-whisker array involved in ground sensing. These whiskers appear to possess biomechanical specializations for ground sensing and may provide the animal with ego-motion information about speed and heading direction [ 10 , 11 ]. While the mystacial macrovibrissae have been studied in detail, we know little about the other approximately 300 whiskers on a rat [ 12 ]. These whiskers are organized in arrays (the upper and lower lip microvibrissae, the paw whiskers, etc.). The few studies on microvibrissae immediately suggested functional differences between macro- and microvibrissae at the behavioral level [ 13 , 14 ] and the level of cortical representation [ 15 ].

Animals can react to airflow stimuli and such wind-sensing abilities are referred to as anemotaxis. The best-studied examples of such behaviors come from insects, where anemotactic turning has been studied, among other species, in crickets [ 1 , 2 ] and in Drosophila [ 3 , 4 ]. Crickets show fast [ 1 ], highly sensitive [ 2 ], and directional escape responses to airflow stimuli. In Drosophila, the antennae are important transducers of anemotactic reactions [ 5 ]. Until recently, little was known about the anemotactic abilities of mammals, but Hartmann and colleagues showed [ 6 ] in a conditioning paradigm that rats can sense airflow. Deficits in airflow sensing after trimming of all whiskers then suggested that this form of airflow sensing is whisker-mediated. The same authors also characterized airflow mechanical responses of mystacial whiskers [ 7 ] and responses of rat trigeminal ganglion cells to airflow stimuli [ 8 ].

Results

Differential whisker displacement by airflow As a first step of our analysis, we assessed the passive displacement of whiskers by wind stimuli, using filmed heads of 5 animals, deeply anesthetized with ketamine, under 2 wind conditions. Whiskers’ end segment was tracked using DeepLabCut [19] (see also S1 and S2 Movies). We labeled the long and short supra-orbital whiskers (lSO and sSO), the straddlers and the whiskers in Arcs 1 to 4 (Fig 1A). We recorded videos of rats under low (0.5 m/s) and high (1.5 m/s) wind conditions. Fan generated turbulent airflow surpassed 80% of its total strength shortly after the stimulus onset and then rapidly reached a steady state [20] (S1A–S1D Fig; S3 and S4 Movies). We then examined the x- and y- movement of each whisker type and calculated displacement as the positive difference between the whisker end segment positions in the 2D plane and their median. We found that while a large number of whiskers exhibited substantial displacement in the high wind condition (Fig 1B), in the low wind condition, only specific whiskers showed marked displacement; these were predominantly the lSO, α, β, A1 whiskers (Fig 1C and S2 Fig). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 1. Differential displacement of rat whiskers in regard to airflow. (A) Head of a deeply anesthetized rat with whisker tips tracked by DeepLabCut. Whisker tags are color coded and whisker names are indicated. (B) Tracked X- and Y-coordinates of the whisker ending segment under low (0.5 m/s; right) and high (1.5 m/s; left) airflow conditions. Note the selective deflection of the lSO, α, β, and A1 whiskers. Whiskers are tagged as in A. (C) Bar plot depicting whiskers displacement (mean ± SEM), in the same rat as in A. Kruskal–Wallis test: H (25, 166,529) = 93,728.2, p < 0.0001. Top right: inset showing whiskers displacement in 5 rats (median ± IQR). Note that lSO whiskers show the largest displacement. (D) Scatter plot showing the paired displacement (cm) of whiskers averaged across rats in the low and high wind conditions. Right-tailed Wilcoxon test: p-value = 0.01. All data underlying the figure can be accessed through https://figshare.com/s/6a9f2304f6ca45b8a7f8. See also S1–S4 Movies. IQR, interquartile range; lSO, long supra-orbital. https://doi.org/10.1371/journal.pbio.3002168.g001 When grouping the averaged whisker displacement on each rat by whisker type, we observed that the lSO whiskers displayed the highest displacement (Fig 1C, inset). To study this pattern quantitatively, we analyzed rats individually. Kruskal–Wallis and post hoc analyses revealed that lSO whiskers moved more than any other whisker apart from α, β, and A1 whiskers. Among this whisker subset, the lSO moved more than the others in most cases (Fig 1C and S2 Fig). Only in Rats 2 and 4, another whisker (α) displaced significantly more than the lSO. Next, we characterized the whisker differential movement upon low (0.5 m/s) and high (1.5 m/s) airflow by taking the median of whisker displacement across rats and comparing their paired values on each condition (Fig 1D). Wilcoxon test indicated that there was a significant increase in the whisker displacement in the high wind condition. A more detailed analysis indicated that this effect was due to a change exhibited by nearly a dozen whiskers, most of them belonging to the supra-orbital region and the top, ocular corner of the whisker pad (S3 Fig). While the details of whisker displacements differed across video sequences, two aspects were the same: (i) lSO, α, β, and A1, whiskers as well as some closely neighboring whiskers always showed big displacements; and (ii) anterior and middle whiskers of the C, D, and E rows always showed little airflow induced displacements. In addition to the quantitatively analyzed movies shown in Fig 1, we also inspected a variety of additional rat head movies qualitatively. These movies included videos of head side views and movies of upside-down heads. All of these recordings led to similar qualitative conclusions. Notably, in all of our experiments, the lSO showed very strong and usually the maximal deflection, prompting us to further examine the function of the lSO in detail with regards to anemotaxis in rats.

Differential whisker biomechanics determine airflow responses We wondered how the differential responses of whiskers to airflow arise. To address this question, we first visually inspected whiskers with differing airflow responses. Differential characteristics were readily visible and immediately noted, namely that the lSO whisker was unusually thin for its length (Fig 2A). Such differences were confirmed when we acquired micrographs of full whiskers and their shafts (Fig 2B). Total whisker length and base diameter were measured in wind and non-wind-engaged whiskers (Fig 2C). We computed a Pearson correlation to examine the relationship between whisker length and base diameter and found a positive correlation between the two variables [r (26) = 0.83, p < 0.001] (Fig 2C). lSO whiskers were relatively thin and short among the long whiskers (Arc 1, 2, and the straddlers) and display a clear difference with respect to the small supra-orbital and the shorter whiskers (Arc 3 and 4). We computed a heatmap of the ratio between whisker length and base diameter and found that lSO has the highest ratio (Fig 2D). We grouped the different whisker types according to a semicircular arrangement and compared their fold change for that ratio with respect to the lSO whisker. Semicircles were found to minimize the mean variance of the ratio along the whisker pad when compared to other possible arrangements using shuffling statistics. Further statistical analysis confirmed that lSO exhibits the highest ratio (S4A Fig). This result suggests that optimal wind engaging occurs within a length-base diameter range that includes supra-orbital and top semicircle whiskers. To further explore this possibility, we performed Pearson correlations of whisker length, base diameter, and their ratio, against the whiskers displacement under low wind (0.5 m/s; Fig 2E and S3B Fig). Results indicated that only the whisker length and the ratio exhibited a significant correlation with displacement (r = 0.4, p = 0.04 and r = 0.64, p < 0.0001, respectively). Up to this point, our data on whisker biomechanics shows that in spite of the fact that the lSO whisker has a moderate length (27 mm; distribution median = 24.7 mm and mean = 25mm), it displays the highest proportion between length-base diameter ratio and wind-induced displacement. These results are in line with previous work showing that whiskers with larger ratios (α >A2 >E2 >C2 >D5), which are not necessarily the longest or thinnest, have higher vibration magnitudes at a given wind speed [8]. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 2. Differential biomechanics determine rat whisker airflow responses. (A) Head of a deeply anesthetized rat. Note the thin whisker diameter of the lSO whisker. (B) Micrograph of the initial segments of lSO and E1 whiskers (top). Photograph of lSO and E1 whiskers (bottom). Scale: 1 mm. Scale: 100 μm. (C) Whisker length plotted against whisker base diameter. Color coded by arcs, inside which length varies the least. Each data point represents the mean length or diameter of each whisker type (n = 4). Pearson correlation indicated. (D) Heatmap of the ratio between whisker length and base diameter. Note that lSO has the highest ratio (see S1 Fig). (E) Pearson correlation between the whisker length to base diameter ratio in regards to whisker displacement. lSO whiskers indicated in red. Rho and p-value indicated. (F) Whisker bending while blowing wind onto extracted whiskers ex vivo. Color-coded curves fitting 75% of the total whisker length were employed to trace a radius (dashed lines) centered at the base of the whisker to calculate the bending angle. Maximal lSO bending is shown. Approximate wind direction (black arrow). Scale: 2 mm (black line, bottom left). (G) Bending angle for each whisker type (color coded). Each dot represents the deflection that a given whisker reached when itself or other whisker type reached its maximal bending. Kruskal–Wallis test on whisker type [H (5, 42) = 36.45, p < 0.0001]. Dunn’s post hoc test indicated that the lSO bending angle significantly differed from every other whisker (all ps < 0.02) except from A1. Meanwhile, A1 differed from C3 and E1 (ps < 0.01). Black crosses indicate the mean ± SEM. All data underlying the figure can be accessed through https://figshare.com/s/3601b5d34cecd4df8f25. See also S5 Movie. lSO, long supra-orbital. https://doi.org/10.1371/journal.pbio.3002168.g002 In order to further evaluate if these biomechanical properties are sufficient to determine differential airflow responses, we performed ex vivo experiments on extracted whiskers (Fig 2F). To this end, we inserted the base of a similar sample of wind and non-wind-engaged whiskers in clay on a linear array with similar orientation. We calculated the maximal bending of the whiskers during low wind flow with respect to the curvature at rest and took the bending angle (Fig 2F and 2G; see Materials and methods). A Kruskal–Wallis test on whisker type showed a significant effect [H (5, 42) = 36.45, p < 0.0001]. Dunn’s post hoc test indicated that only comparisons involving lSO and A1 whiskers yielded significant differences. Particularly, the bending angle of lSO significantly differs from every other whisker (all p-values <0.02) except A1, which was another wind-sensitive whisker found in our previous in vivo assay. A1 differed from C3 and E1 (p-values <0.01). Taken together, our results identify whisker biomechanics as crucial determinants of airflow responses.

Mapping of supra-orbital whisker barrels and relation of whisker airflow displacement to the cortical barrel map The differential mechanical airflow responses of whiskers point towards a role of the supra-orbital whiskers in airflow sensing. We therefore mapped the location of cortical barrels representing the supra-orbital whiskers in extracellular receptive field mapping experiments and prepared cytochrome oxidase sections of layer IV of the barrel cortex (Fig 4A) [22]. We consistently (in 4 out of 4 mapping experiments) observed supra-orbital whisker responses in brain regions posterior to the A1 and α whisker response areas. Also, the stereotaxic coordinates of supra-orbital whiskers were highly consistent (6.26 ± 0.01 mm lateral and 3.75 ± 0.20 mm posterior to bregma, mean ± SEM). These observations led us to a suggestion for the location of the supra-orbital whisker barrels in relation to the rest of the barrel field (Fig 4B). Next, we wondered how mechanical airflow responsiveness relates to the cortical barrel field and we color coded it and superimposed it on the barrel map (Fig 4C). Quantitative tracking data for whisker displacement was not available for all whiskers (hence, the empty barrels), but it was nonetheless clear that wind-responsive whiskers (with large airflow displacements) cluster in the posterolateral barrel map. We also inspected the putative supra-orbital whisker barrels in many (n = 10) additional barrel maps that we obtained from other purposes in previous studies [23,24]. We made the following observations: (i) the exact position and orientation of putative supra-orbital whisker barrels relative to the posteromedial-barrel-subfield is somewhat variable and more variable relative to the position and orientation of the mystacial barrels to each other. (ii) Putative supra-orbital whisker barrels are elongated. (iii) Putative supra-orbital whisker barrels are always close (see also Fig 4A and 4B). (iv) The septum separating putative supra-orbital whisker barrels is weaker than the septum separating mystacial barrels (see also Fig 4A and 4B). The latter 2 observations support the idea that the short and long supra-orbital whiskers are functionally related. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 4. Localization of supra-orbital whisker barrels and relation of whisker airflow displacement to the cortical barrel map. (A) Cortical barrels in a tangential section through layer 4 of rat barrel cortex revealed staining for cytochrome oxidase reactivity; dark brown color indicates high reactivity. a = anterior, l = lateral. (B) Drawing of cortical barrels (from A) with the positions of supra-orbital whisker barrels. sSO and lSO whisker barrels were identified in 4 receptive field mapping experiments, in all cases posterior rather than lateral to α/A1 whisker responses. Note that some anterior barrels (A4 and B4) and microvibrissae barrels are missing due to sectioning. (C) Whisker displacement under low airflow conditions was quantified, normalized to the maximal response, color coded and superimposed to the barrel map drawn in (B). Data comes from an airflow whisker displacement experiment on the head of an anesthetized animal analogous to the data shown in Fig 1F. Quantitative tracking data for whisker displacement were not available for all whiskers (hence, the empty barrels). Qualitative assessment of D- and E-row whiskers suggested they show little airflow whisker displacement similar to the data of whisker D4 (also see S1 Movie). Wind-responsive whiskers (with large airflow displacements) cluster in the posterolateral barrel map. All data underlying the figure can be accessed through https://figshare.com/s/e563353889ea06181807. lSO, long supra-orbital; sSO, short supra-orbital. https://doi.org/10.1371/journal.pbio.3002168.g004

Neurons in the supra-orbital whisker representation respond more strongly to wind stimuli than E/D-row barrel cortex neurons Next, we wondered if the cortical supra-orbital whisker representation differed from barrel cortex neurons in their responses to wind stimuli. We applied wind stimuli to urethane-anesthetized rats, while recording simultaneously with Neuropixel probes from the supra-orbital whisker region at the coordinates identified in our mapping experiments and from the whisker pad region aiming towards E/D-row barrel cortex (Fig 5A). We histologically confirmed recording locations to the supra-orbital cortical region and the whisker pad barrel cortex near E/D-row (Fig 5B). Judging by the population peristimulus time histogram (PSTH), there was not much of a wind-evoked response in recordings from E/D-row barrel cortex. In contrast, there was a clear excitatory response in the supra-orbital whisker region (Fig 5C). Plots of the z-scored responses of individual neurons revealed either no, weak, or inhibitory responses to wind stimuli in E/D-row barrel cortex. In the supra-orbital whisker region instead, we observed strong excitatory responses in single cells (Fig 5D). The differences in the firing rate response to either low (0.5 m/s) or high (1.5 m/s) wind between these cortical regions were highly significant (Fig 5E) and were distributed differently in time across response categories (Fig 5F). We found that the SO region exhibited the highest percentage of excited neurons, surpassing the 25% of recruitment 1 second after the stimulus onset in both wind conditions. In contrast, pad region neurons displayed a balance between being excited and inhibited during low wind and only recruited 12% of neurons at its peak during high wind. This pattern of response was further explained by an analysis on the response latency, which showed that neurons reached their maximum response 1 second after the stimulus onset (S6A Fig). These differences suggest that wind responses map to the supra-orbital whisker barrel. To further confirm this, we calculated for each cell the mutual information of the firing rate given a wind stimulus (S6B Fig). Results indicated that only firing rate activity of neurons in the SO region significantly informed about wind stimuli. Moreover, in line with our results regarding the percentage of recruited activity and the response latency, mutual information peaked during second 2 (low wind, p = 0.0007; high wind, p < 0.0001). PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 5. Supra-orbital whisker cortex responds more strongly to wind stimuli than D/E-row barrel cortex. (A) Schematic of the experimental setup. Posteriorly and anteriorly placed Neuropixels probes were aimed to the supra-orbital and the whisker pad regions of the barrel cortex, respectively. Simultaneous, contralateral recordings of single units were made while blowing wind. Low (0.5 m/s, blue) or high (1.5 m/s, red) wind epochs (10 s) were blown in alternating order from a frontal fan placed 12.5 cm apart from the rat’s head. Top right: schematic of the wind epochs in time (12–30 total wind epochs per rat). Scales: x: 10 s; y: 1.5 m/s. (B) Left: representative histology showing the 2 recording sites on the whisker pad and supra-orbital regions of the barrel cortex. Scale: 500 μm. Right: schematic reconstruction of the barrel cortex from successive flattened brain slices. (C) Representative examples of peri-wind stimulus firing rate of 2 single units recorded at the whisker pad (left) or supra-orbital (right) regions in the low (blue) and high (red) wind conditions. Black dash lines and color code step lines on top indicate stimuli onset. (D) Heatmap of z-scored firing rate around wind stimuli (low wind, top; high wind, bottom) of single units recorded at the whisker pad (left) or supra-orbital (right) regions. Positive z-scores indicate excitation (black). Negative z-scores indicate inhibition (white). (E) Firing rate for the difference between post vs. pre-wind stimulation in single units recorded at the whiskers pad (yellow) or the supra-orbital (lilac) regions for low (left) and high (right) wind conditions. (F) Percentages of excited and inhibited pad (left) and supra-orbital (right) regions across time after low (top row) and high (bottom row) wind conditions, calculated using a generalized linear model (GLM). Note that response percentage peaks at second 2. All data underlying the figure can be accessed through https://figshare.com/s/ef2e783e590dc552cf08. https://doi.org/10.1371/journal.pbio.3002168.g005

Anemotaxic turning in rats To assess the behavioral capacities for wind sensing in rats, we developed an airflow-sensing paradigm. We placed a rat in a box with 3 compartments separated by wire-mesh in total darkness. The rat was placed in the middle compartment and 2 experimenters performed repetitive hand-flaps or cardboard-flaps, in either one of the 2 lateral compartments (Fig 6A and 6C). Airflow measurements of hand- and cardboard-flap stimuli were on average ≤0.3 m/s and 0.5 m/s, respectively. The reactions of rats to hand-flap stimuli were assigned by forced choice to one of 3 categories: either no reaction, turning towards the stimulus, or turning away from the stimulus (Fig 6B). Even though rats often showed no reaction, when they did, the animals appeared to be able to distinguish the side where the hand-flap was delivered. Accordingly, rats turned significantly more often towards hand-flaps than away from them (Fig 6B; p < 0.001, χ2 Test; “Turn to” (31 trials) versus “Turn away” (7 trials)). Next, using the same behavioral paradigm, we changed the wind delivery method to utilize a cardboard piece, which induced stronger airflow than the hand-flap (Fig 6C). Again, the animals consistently showed a higher percentage of responses towards the stimuli side when compared to turning away responses (Fig 6D; p < 0.001, χ2 test). When comparing the “Turn to” responses in the 2 wind delivery methods, we observed a stronger reactivity of the animals to cardboard-flap than to hand-flap stimuli (Fig 6C and 6D; p = 0.0036, Fisher’s exact test). Our results show that rats can not only sense, but also turn towards airflow stimuli. The strength of the reactions differed between weak (hand-flap) and strong (cardboard-flap) stimuli. Since we carefully avoided noises associated with hand-flap or cardboard-flap stimuli and conducted experiments in total darkness, it is likely that animals indeed sensed airflow. The whisker trimming and lidocaine injection effects described below show that the turning responses observed were indeed at least partially, if not entirely, tactile reactions. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 6. Anemotaxic turning in rats. (A) Schematic of the behavioral arena for wind-sensing behavior in response to hand-flapping. See Materials and methods for more details. (B) Behavioral responses of rats (n = 7) to hand-movement stimuli (0.5 s post stimulus) were assigned by forced choice to one of 3 categories: either no reaction, turning towards the stimulus, or turning away from the stimulus. Rats were strongly biased to turn towards the hand-movement stimuli (p < 0.001, χ2 test). (C) Cardboard-flaps are used to apply stronger airflow stimuli than the hand-flaps. (D) Rats react to the cardboard-flap movement stimuli from (C) and scoring is done as in (B). Rats were strongly biased to turn towards the cardboard-flap stimuli (p < 0.001, χ2 test). Also, rats turn towards cardboard-flaps more frequently than to hand-flaps (p = 0.0036, Fisher’s exact test). All data underlying the figure can be accessed through https://figshare.com/s/07af3c1099b5b785acff. https://doi.org/10.1371/journal.pbio.3002168.g006

[END]
---
[1] Url: https://journals.plos.org/plosbiology/article?id=10.1371/journal.pbio.3002168

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/