of 8
Supporting Information
van Breugel and Dickinson 10.1073/pnas.1714874114
SI Methods
Collecting.
E. hians
(Mono Lake, CA), kelp flies (Laguna Beach, CA), and the Ephydridae sp. (Santa Ana River, CA) were collected using a
large butterfly net and kept in glass vials sealed with a cotton ball a
nd refrigerated. Oil flies (La Brea Tar Pits, CA) were collected in-
dividually in glass vials.
D. virilis
(San Diego stock) and
Drosophila melanogaster
(HCS) were reared in the laboratory. Force measurements
weredonewhiletheflieswerestillalive(3
4 d from collection). MLW was collected in clean 16-
oz mason jars, stored in the refrigerator, and
used within 1
2 mo of collection. Before using the st
ored water we allowed it to return to room temperature and agitated it gently.
Force Measurements.
Flies were anesthetized with CO
2
and glued to a tungsten beam with a small drop of UV-cured glue. The beam was
bent down by 90° at its distal end so that we could easily plunge each fly into the test solution. The distal end of the horizontal section
of the beam was placed between a red LED and position-sensitive photodetector (SL5-2 UDT Sensors). The power of the LED at the
distance of the photodetector was 3.2 mW, measured using a photometer (Thor PM100D). Deflections of the beam caused its shadow
to move on the surface of the SL5-2 sensor. The sensitivity of the system could be tuned by choosing the length and diameter of the
beam. For our measurements we used a 43-mm-long, 0.26-mm-diameter beam. This combination produced a force-to-displacement
ratio of 3.15
×
10
6
N
·
μ
m
1
and a resolution of 2
10
μ
N, depending on the filter settings. Measurements were more accurate when the
fly contacted the water and viscous damping reduced oscillations of the beam.
To measure the forces required for a fly to enter the water we used a linear motor to slowly (0.3 mm
·
s
1
) raise a clear acrylic container
with 300-mm solution until the fly was fully submerged. The distance traveled was measured using an LVDT (linear variable dis-
placement transformer; Newport). The solution was then lowered until the fly was above the test solution. Force and LVDT mea-
surements were streamed to a computer with a Phidgets I/O board (
https://www.phidgets.com/
) recording data at 250 Hz.
To help reduce the effects of individual variability between flies we paired our experiments such that each fly was dipped into each of
the two to three solutions being tested. To eliminate any bias due to the order of solutions being tested we performed half of the trials in
each experiment in one order and half in the opposite order. High concentrations of CO
3
2
ions did result in lasting reductions in
hydrophobicity (Fig. 2
F
and
G
). Because of this, we did not reverse the solution order for our species comparison experiments (Fig.
3
A
). For these experiments all flies were tested in the order of pure water, MLW, and carbonate rich water.
After each sequence of solutions, a 41-mg aluminum-foil weight was balanced on the fly and the beam deflection recorded. These
calibrations served to correct for any small changes in the beam position from fly to fly.
Analysis and Statistics.
Data were analyzed using custom Python code, making use of the following packages: numpy, scipy, matplotlib,
statsmodels, pyqtgraph, and figurefirst. The force traces were manually broken into five sections (before entering the water, entering the
water, under water, exiting the water, and after exiting the water) using a custom PyQTGraph graphical user interface.
Throughout the figures we use shading to indicate the 95% confidence intervals of the means, calculated through bootstrapping with
1,000 iterations. In cases with relatively normally distributed data and similar sample sizes and variances (such as our data), distributions
with nonoverlapping 95% confidence intervals are empirically equivalent to a
P
value calculated through a permutation test of
0.02 or
less. In cases where the 95% confidence intervals are close, we explicitly calculate these probabilities through a permutation test.
Notes on Variability of Force Measurements.
There are a number of reasons behind the variability in our measurements (e.g., Fig. 2
D
J
).
Three of the largest contributors are as follows:
i
) The forces are correlated with body size (Fig. 3
B
), and individuals vary in size. We explored an analysis that controlled for body
size by normalizing each group of experiments to the recovered work of one of the solutions. This alternate analysis does produce
stronger statistics, but we chose to present the simpler analysis that makes no such assumptions because the results were already
interpretable without introducing more complicated methods.
ii
) We made our force measurements with briefly anesthetized flies, but many of them awoke during the course of our experiment.
These flies wiggled their legs and abdomen, creating additional forces. In our first attempts, we worked with frozen and defrosted
flies. The freezing process, however, appeared to have damaged the hairs that keep the flies dry, and they were more prone to
wetting than fresh flies.
iii
) Both the work to submerge flies and the recovered work were subject to similar variances. We chose to focus on the recovered
work because it is best correlated with our observations of wetting, so that the quantitative results can be intuitively connected with
real world phenomenon of flies getting stuck at the water surface.
SEM Imaging.
Before preparing the flies for imaging they were stored at
20 °C in sealed glass vials. We prepared the samples using a
critical point drying protocol. The samples were soaked in ethanol for 24 h and placed in a pressure chamber cooled to 15 °C. The
chamber was filled with liquid CO
2
(800 psi) as the ethanol was flushed out for several minutes. The drain valve was closed and CO
2
allowed to permeate the samples for 20 min. After flushing the system again, we waited another 20 min. Finally, the temperature was
raised to 35 °C and the pressure rose to
1,200 psi. At this point, the liquid CO
2
is in a supercritical liquid state. The outlet valve was
opened and pressure relieved at a rate of 100 psi
·
min
1
, allowing gaseous CO
2
to slowly leave the samples. We mounted the samples to
SEM stubs with Pelco conductive graphite (Ted Pella) and sputter-coated with Pt/Pd to a thickness of 10 nm.
Hairiness Analysis.
To calculate the hairiness of a fly we used image processing to count the luminance peaks across several image transects
of SEM images for different body parts of the fly. For thorax, abdomen, and wing images we chose four nonoverlapping quadrants. For
each quadrant we calculated the number of hairs crossing 12 transects rotated about the quadrant center at 30° increments and chose the
largest value, which corresponds to the transect perpendicular to the mean hair orientation. We then took the average of these values for
the four quadrants. For the oil fly
s abdomen we manually specified two quadrants, since the image did not fill the frame. For leg images,
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
1of8
we manually specified four transects for each fly to ensure that the transects spanned the leg and were perpendicular to the hair ori-
entation. For some images with a preponderance of thick hairs this algorithm failed (finding two peaks instead of one). For these images
we manually counted the number of hairs across the specified transects. Based on this approach,
hairiness
is defined as the average
number of hairs crossing a transect, which is correlated with both hair density and hair length. For example, suppose we have two flies
with the same hair density but with different hair lengths. The fly with longer hairs will yield more hair crossings for a given transect.
Extraction and Analysis of Cuticular Hydrocarbons.
To extract the cuticular hydrocarbons for each species we gently agitated 1
10 flies
(depending on the number of specimens collected) for 1 min in enough
hexane to cover the flies and tran
sferred the solution to a clean
Teflon-capped glass vial. The quantity of flies tested was as follows:
E. hians
(5), blue kelp (5), black kelp (3), Ephydridae sp. (5), oil fly (1),
D. melanogaster
(10), and
D. virilis
(10). We did not verify the sex of the flies tested except for
D. melanogaster
, in which case we used females.
Hexane extracts were concentrated as needed under a gentle stream of nitrogen then analyzed by coupled GC-MS using an H-P 5890
GC (Hewlett-Packard, now Agilent) interfaced to an H-P 5973 mass selective detector, with electron impact ionization (70 eV). The GC
was fitted with a DB-17 column (30-m
×
0.25-mm i.d., 0.25-
μ
m film; J&W Scientific) and was programmed from 100 °C/0 min, 20°/min to
160°/0 min, 4°/min to 280°, hold for 20 min. Injections were made in splitless mode with injector and transfer line temperatures of 280°.
Compounds were identified by a combination of their retention indices relative to straight-chain hydrocarbon standards and in-
terpretation of their mass spectra as described by Carlson et al. (1).
We tried to coat a different species (
D. virilis
) with cuticular hydrocarbons (CHCs) extracted from the alkali flies but did not
measure a change in performance. However, this negative result is difficult to interpret as we do not know if we were successful in
transferring enough CHC in the correct orientations and distributions onto the test flies.
Bond Polarizabilities.
The polarizability of a molecule can be calculated by summing the individual bond polarizabilities for each bond in
the molecule (2). Using C
C bond polarizability
=
5.3
×
10
41
C
2
m
2
J
1
and C
H bond polarizability
=
7.2
×
10
41
C
2
m
2
J
1
we
calculated the total polarizability of C25 (24 C
C bonds, 52 C
Hbonds)tobe5.0
×
10
39
and calculated the total polarizability of tetra-
methylated C30 (29 C
Cbonds,66C
Hbonds)tobe6.3
×
10
39
.
Contact Angle on Shrimp Shell.
To determine the effect of carbonate ions on the hydrophobicity of smooth surfaces we compared the
contact angle of pure water and 0.5 M carbonate solution on clean, flattened shrimp shell using the static sessile drop technique. A flat
and homogenous piece of shrimp shell was cleaned and rinsed with hexane and then fixed to a glass slide with superglue. We then placed
2-
μ
L droplets of the solution on the shrimp shell surface and photographed the droplet using a Canon 60D equipped with a Canon
MPE-65 lens. Contact angles were measured manually using the image processing module in Fiji (ImageJ).
Dunk and Escape Tests with Sunscreen.
Sunscreen was applied to wooden popsicle sticks and allowed to sit for 15 min. The sticks were then
swirled in 400 mL pure water for 1 min. For these experiments, we used 16-oz wide-mouth glass mason jars (surface area of 44 cm
2
). Flies
were briefly anesthetized with CO
2
, held by one leg with forceps, dunked under water, and released. Shortly after floating to the
surface the flies started to regain consciousness and attempted to take off. The number of flies trapped on the surface was counted
after 15 min. We followed a similar protocol for the dimethicone experiments but pipetted the liquid onto the surface rather than using
popsicle sticks as an intermediate.
Sunscreen Ingredient Analysis.
To determine which ingredients are most deleterious to the flies
ability to escape the water surface we analyzed
the ingredients across all six sunscreens. The wo
rd cloud was generated us
ing python-wordcloud (
https://github.com/amueller/word_cloud
).
Electric Potential Order of Magnitude Calculations.
To provide a simple order of magnitude estimate of the molarity of Na
2
CO
3
required to
generate an electric field potential of 22 V between the water
s surface and the flies
cuticular hydrocarbons we m
odeled the sodium and car-
bonate ions as two superimposed electric fields. The electric f
ield generated by an infinite flat plate with a charge density
σ
=
ze
=
L
2
is defined as:
E
=
σ
2
«
0
«
,
[S1]
where
z
is the net atomic charge,
e
is the charge of a single electron (1.6
×
10
19
C),
L
is the width of the sheet,
«
0
is the permittivity of free
space (8.85
×
10
12
F/m), and
«
is the static dialectric permittivity of the intervening space (1 for air). In the case of our problem, however,
an infinite plate is too much of an oversimplification, so we will instead use the electric field generated by a charged square sheet:
E
=
σ
π
«
0
«
arctan
0
B
@
1
2
ð
z
=
L
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4

r
L

2
+
2
q
1
C
A
,
[S2]
where
r
is the distance from the sheet. This electric field will induce dipole moments in nearby uncharged, but polarizable, molecules
such as cuticular hydrocarbons or chitin. The electric field of each induced dipole can be written as
E
r
=
2
u
ind

4
π
«
0
«
r
3
,
[S3]
where
u
ind
=
α
E
,
[S4]
and
α
is the molecular polarizability, determined by summing the bond polarizabilities of the full compound (ref. 2, p. 93). Because
α
is
quite small for a cuticular hydrocarbon [for a C25,
α
=
5
×
10
39
C
2
/(m
2
·
J
1
); see above], the contribution of the electric field
generated by the induced dipoles is negligible compared with that of the charged surface, so for simplicity we will omit it.
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
2of8
Because the sodium and carbonate ions settle at different distances from the air
water interface based on their hydration shells (Fig. 7
B
)
we modeled the system using two superimposed electric fields generated by two charged sheets, a negative one positioned at the top
surface of the carbonate ions and a positive one at the top surface of the sodium ions. To calculate the voltage potential between a 1-
μ
m
2
patch of interfacial water and the center of the hydrocarbons on the surface of the insect cuticle (2 nm from the air
water interface) we
integrate the two fields with respect to r and sum the resulting voltage potentials due to the cation and anion fields (V
c
due to Na
+
and V
a
due to CO
3
2
).
Because the interaction takes place partly across the hydrocarbon
ð
«
=
2
Þ
and partly across water
ð
«
=
80
Þ
we use values of
«
c
=
2
*
2
+
80
*
.116
=
2.116
=
6.3 for the sodium field and
«
a
=
2
*
2
+
80
*
.076
=
2.076
=
4.8 for the carbonate field:
V
c
=
σ
c
π
«
0
«
c
Z
E
c
=
σ
c
π
«
0
«
c
Z
r
0
=
2.116
×
10
9
arctan
1
2
ð
z
=
L
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4

r
L

2
+
2
q
dr
=
σ
c
π
«
0
«
c
*
8.78056
×
10
7
.
[S5]
V
a
=
σ
a
π
«
0
«
a
Z
E
a
=
σ
a
π
«
0
«
a
Z
r
0
=
2.076
×
10
9
arctan
1
2
ð
z
=
L
Þ
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
4

r
L

2
+
2
q
dr
=
σ
a
π
«
0
«
a
*
8.78119
×
10
7
.
[S6]
Now we replace
σ
with the following expressions, where
N
is the number of molecules spread across the 1-
μ
m
2
sheet (
L
=
1
×
10
6
m):
σ
c
=
2
N*
1
*e
L
2
and
σ
a
=
N*
2
*e
L
2
.
[S7]
We than add
V
c
and
V
a
, substitute in the expressions for
σ
and
«
, and set the sum equal to 22 V:
V
c
+
V
a
=
22
ð
V
Þ
=
2
N*
1
*e
L
2
π
«
0
«
c
*
8.78056
×
10
7
+
N*
2
*e
L
2
π
«
0
«
a
*
8.78119
×
10
7
.
[S8]
Solving for
N
yields
N
=





22

2
*
e
L
2
π
«
0
«
c
*
8.78056
×
10
7
2
*
e
L
2
π
«
0
«
a
*
8.78119
×
10
7






=
4.66
×
10
16
molecules
per m
2
.
[S9]
Now we calculate the molar concentration necessary to achieve the above. First we assume that the molecules are evenly distributed
through the volume and that the thickness of the sheet of ions below the water surface that contributes to
E
c
and
E
a
is equal to the
diameter of the hydrated carbonate sphere (
0.5 nm):
N
molecules
m
2
=
M
mol
L
*
1,000
L
m
3
*
6.022
×
10
23
molecules
mol
*
0.5
×
10
9
ð
m
Þ
.
[S10]
Solving for the molarity yields
M
=
0.15 mol
·
L
1
.
Bond Number.
To confirm that the changes in recovered work we found in Fig. 2
D
are primarily a result of surface tension related effects,
as opposed to being a function of the changes in density between solutions, we calculated the Bond number (also known as the Eötvös
number) (3). The Bond number compares the potential energy associated with buoyancy with the surface energy associated with the
air/water interface. Small Bond numbers (less than 1) indicate that the system is dominated by surface tension. The Bond number is
calculated as
Bo
=
Δ
ρ
gL
2
σ
,
where
Δ
ρ
is the change in density between the liquid and solid phase,
g
is the gravitational acceleration,
L
is the characteristic length,
and
σ
is the surface tension. Here, we calculate the Bond number of a 0.5 M Na
2
CO
3
solution. A 25-mL 0.5 M Na2CO3 solution has a
mass of 26.477 g, and after freezing, the volume increases to 27.5 mL. Thus,
Δ
ρ
=
96 kg
=
m
3
. For the characteristic length we chose the
average length of an alkali fly, 6 mm. For the surface tension, we used 73.15 mN
·
m
1
(4). These values yield a Bond number of 0.46,
confirming that the forces on a submerged alkali fly in solutions like Na
2
CO
3
are dominated by surface forces, not buoyancy forces.
The Bond number for pure water
ð
Δ
ρ
=
80 kg
=
m
3
;
σ
=
72.8 mN
=
m
Þ
is slightly smaller, 0.39.
1. Carlson DA, Bernier UR, Sutton BD (1998) Elution patterns from capillary GC for methyl-branched alkanes.
J Chem Ecol
24:1845
1865.
2. Israelachvili JN (2011) Intermolecular and surface forces (Academic, Waltham, MA), 3rd Ed.
3. Vogel S (2003)
Comparative Biomechanics: Life
s Physical World
(Princeton Univ Press, Princeton).
4. Ozdemir O, Karakashev SI, Nguyen AV, Miller JD (2006) Adsorption of carbonate and bicarbonate salts at the air
brine interface.
Int J Miner Process
81:149
158.
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
3of8
Fig. S1.
Entering water with a vertical orientation minimizes peak forces. (
A
) Data plotted as in Fig. 2
B
for experiments done on alkali flies in the field.
Working in the field meant that the flies were subject to more handling, likely compromising some of their hydrophobic properties, thus the slightly r
educed
forces on exiting the water. (
B
) Peak force to enter the water vs. fly orientation for the data shown in
A
.(
C
) Same as
B
, for flies exiting the water.
Fig. S2.
SEMs of various body parts for seven fly species.
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
4of8
Fig. S3.
Oils such as natural fish oil and artificial dimethicone saturate alkali flies
hairy cuticle, causing them to drown. (
A
) Alkali fly walking on MLW (Movie S10).
(
B
) Alkali fly walking on MLW coated in a thin surface film of fish oil (Movie S11). (
C
) Word cloud of ingredients from the six sunscreens tested in Fig. 6
E
.Wordsizeis
positively correlated with a mortality metric based on the results from Fig. 6
E
. We assigned mortality values of 3, 2, and 1 to the Neutrogena, Banana Spray, and
Banana Baby sunscreens and
1 to the other three. The mortality value for each ingredient was then calculated by summing the mortality values for each sunscreen
that contained that ingredient. For example, an ingredient present in the Neutrogena sunscreen and the Bullfrog sunscreen yields a mortality value o
f3
1
=
2.
Movie S1.
E. hians
entering Mono Lake on a piece of partially submerged tufa. A brine shrimp provides an additional obstacle.
Movie S1
Movie S2.
E. hians
entering Mono Lake on a piece of partially submerged tufa.
Movie S2
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
5of8
Movie S3.
E. hians
surfacing after being brought under water in a test tube.
Movie S3
Movie S4.
E. hians
naturally surfacing in Mono Lake.
Movie S4
Movie S5.
Synchronized movie of an alkali fly being dipped into MLW and the resulting forces (in arbitrary units) on its body. This recording was made in the
field with the fly in a vertical orientation, which is why the forces exiting the water are not quite matched with those entering the water. The bright ob
ject that
swims past the fly is a brine shrimp.
Movie S5
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
6of8
Movie S6.
E. hians
being dipped into MLW (8
×
real time).
Movie S6
Movie S7.
Coelopa vanduzeei
being dipped into 0.5 M Na
2
CO
3
solution (8
×
real time).
Movie S7
Movie S8.
Movies of droplets of deionized water coming into contact with the trailing edge of nine different
M. domestica
wings. Playback speed is
3
×
actual speed and adjusted in postproduction to be synchronized to the time when the droplets are farthest to the left.
Movie S8
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
7of8
Movie S9.
Movies of droplets of 2 M Na
2
CO
3
solution coming into contact with the same nine trailing edges of
M. domestica
wings as in Movie S10. Playback
speed is
3
×
actual speed and adjusted in postproduction to be synchronized to the time when the droplets are farthest to the left.
Movie S9
Movie S10.
E. hians
walking on MLW.
Movie S10
Movie S11.
E. hians
walking (and drowning) on MLW coated with a thin film of fish oil (cod liver oil).
Movie S11
van Breugel and Dickinson
www.pnas.org/cgi/content/short/1714874114
8of8