See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/260999073

Rapid CO2 permeation across biological

membranes: Implications for CO2 venting from
tissue

ARTICLE in THE FASEB JOURNAL MARCH 2014

Impact Factor: 5.04 DOI: 10.1096/fj.13-241752 Source: PubMed

CITATIONS READS

5 9

2 AUTHORS, INCLUDING:

Pawel Swietach
University of Oxford
62 PUBLICATIONS 1,634 CITATIONS

SEE PROFILE

All in-text references underlined in blue are linked to publications on ResearchGate, Available from: Pawel Swietach
letting you access and read them immediately. Retrieved on: 08 March 2016
The FASEB Journal Research Communication
Rapid CO2 permeation across biological membranes:
implications for CO2 venting from tissue
Alzbeta Hulikova and Pawel Swietach1
Department of Physiology, Anatomy, and Genetics, University of Oxford, Oxford, United Kingdom
ABSTRACT The degree to which cell membranes are
barriers to CO2 transport remains controversial. Pro-
teins, such as aquaporins and Rh complex, have been
proposed to facilitate CO2 transport, implying that the
nonchannel component of membranes must have
greatly reduced CO2 permeability. To determine
whether membrane CO2 permeation is rate limiting for
gas transport, the spread of CO2 across multicellular
tissue growths (spheroids) was measured using intracel-
lular pH as a spatial readout. Colorectal HCT116 cells
have basal water and NH3 permeability, indicating the
functional absence of aquaporins and gas channels.
However, CO2 diffusivity in HCT116 spheroids was
only 24 4% lower than in pure water, which can be
accounted for fully by volume exclusion due to pro-
teins. Diffusivity was unaffected by blockers of aqua-
porins and Rh complex (Hg2, p-chloromercuribenzoic
acid, and 4,4=-diisothiocyano-2,2=-stilbene-disulfonic
acid) but decreased under hypertonic conditions (by
addition of 300 mOsm mannitol), which increases
intracellular protein crowding. Similar CO2 diffusivity
was measured in spheroids of T47D breast cells (basal
water permeability) and NHDF-Ad fibroblasts (aqua-
porin-facilitated water permeability). In contrast, diffu-
sivity of NH3, a smaller but less lipophilic gas, was
considerably slower than in pure water, as expected
from rate-limiting membrane permeation. In conclu-
sion, membranes, even in the functional absence of
proposed gas channels, do not restrict CO2 venting
from tissue growths.Hulikova, A., Swietach, P. Rapid
CO2 permeation across biological membranes: implica-
tions for CO2 venting from tissue. FASEB J. 28,
27622774 (2014). www.fasebj.org
Abbreviations: 3-D, 3-dimensional; 2-D, 2-dimensional; e,
extracellular buffering capacity; i, intracellular buffering
capacity; AMB, 4-aminomethylbenzenesulfonamide; AMC,
7-amino-4-methylcoumarin; AQP, aquaporin; ATZ, acetazol-
amide; CA, carbonic anhydrase; CBX, carbenoxolone;
cSNARF1, 5-(and-6)-carboxy SNARF-1; D, diffusion coefficient;
Dc, diffusion coefficient in cytoplasm; De, diffusion coefficient in
extracellular space; Deff, effective diffusion coefficient; DIDS
4,4=-diisothiocyano-2,2=-stilbene-disulfonic acid; DMA 5-(N,N-di-
methyl)amiloride; FS, fluorescein-5-(and-6-)-sulfonic acid; Kp,
oil:water partition coefficient; NBA 2-nitrobenzaldehyde; NHE,
Na/H exchange; PCMB, p-chloromercuribenzoic acid;
pHi, intracellular pH; Pm, membrane permeability constant;
ve, extracellular volume fraction; vi, intracellular volume
fraction
2762
Key Words: carbon dioxide metabolism excretion gas channels
CO2 is a waste product of metabolism, produced in
abundance by decarboxylation reactions (such as aero-
bic respiration) and neutralization reactions (e.g., be-
tween HCO3 ions and lactic acid). The process of CO2
venting from tissue must overcome resistances imposed
by diffusion across intracellular and interstitial com-
partments and permeation across membranes. These
resistances determine the capacity for CO2 removal,
which is important in the context of maintaining cor-
rect acid-base balance and supporting cellular metabo-
lism.
The high CO2 diffusion coefficient in water
(DCO22.5103 m2/s; ref. 1) argues for rapid trans-
port across aqueous compartments. The presence of
macromolecules in the cytoplasm (2) and the tortuosity
of interstitial spaces (3) reduce cytoplasmic and extra-
cellular DCO2 (Dc,CO2 and De,CO2), respectively. CO2
diffusion can be facilitated by a parallel flux of HCO3
and H ions, but the extent of this depends on the
activity of carbonic anhydrase (CA) enzymes that cata-
lyze the reversible chemical conversion of CO2 to
HCO3 and H ions (1). Indeed, a major physiological
role for exofacial CA isoforms, such as CAIV or tumor-
associated CAIX and CAXII, is to facilitate extracellular
CO2 diffusion (4, 5). Thus, transport of CO2 across
aqueous compartments of tissue depends on path
length (i.e., tortuosity and volume exclusion) and CA
activity.
In accordance to the Overton rule (6, 7), the high
oil:water partition coefficient of CO2 (Kp,CO21.5; refs.
8, 9) predicts very high CO2 permeability across the
lipid matrix of membranes. High membrane CO2 per-
meability (Pm,CO2104 m/s) has been measured in
artificial bilayers (10, 11) and native membranes (11,
12). However, several other studies have demonstrated
that at least some biological membranes are consider-
ably tighter to CO2 (1318), arguing for an unexpect-
1
Correspondence: Department of Physiology, Anatomy,
and Genetics, University of Oxford, Sherrington Bldg.,
Parks Rd., Oxford OX1 3PT, UK. E-mail: pawel.swietach
@dpag.ox.ac.uk
doi: 10.1096/fj.13-241752
This article includes supplemental data. Please visit http://
www.fasebj.org to obtain this information.
0892-6638/14/0028-2762 FASEB
edly low basal CO2 permeability (Pm,CO210 100 m/
s). Furthermore, it has been proposed that gas
channels, such as aquaporins and Rh-proteins (1720),
are physiologically important conduits for CO2 passage,
implying that CO2 permeability of the nonchannel
portion of membranes must be orders of magnitude
lower than predicted; otherwise, the membrane matrix
would short circuit the gas channel pathway (21). The
existence of CO2 channels remains contentious (7, 21),
with some (11) arguing that Pm,CO2 has been underes-
timated because of unaccounted diffusive resistance
across unstirred aqueous layers (22). Evidence for CO2
channels has been inferred from measurements of
extracellular pH (pHe) transients arising from the
passage of CO2 across membranes of heterologous
expression systems (19, 20). However, this does not
yield a calibrated measure of Pm,CO2 because the rela-
tionship between pHe transients and gas permeation is
multifactoral (23). Further evidence for CO2 gas chan-
nels was obtained from measuring influx of 18O-labeled
CO2 and HCO3 into cells (16 18). This approach can
resolve Pm,CO2 up to 103 m/s but extracts informa-
tion about a fast process from slow equilibria that
require accurate parameterization (11).
An understanding of the factors that restrict or
facilitate CO2 transport is important for demarcating
the path of least resistance for CO2 venting and iden-
tifying the rate-limiting steps. These are particularly
important to consider in poorly perfused and metabol-
ically active tissue such as solid tumors (24). If Pm,CO2
were significantly lower than the Overton prediction,
membranes would impose a substantial resistance to
CO2 removal from tissue and favor a predominantly
interstitial path for CO2 venting, i.e., permeation across
the membrane of a CO2-producing cell and diffusion
through the meandering interstitium. The alternative
transcellular pathway, running across a series of mem-
branes and adjoining cells, would require an ade-
quately high Pm,CO2. Gas channels would increase CO2
venting, particularly via the transcellular route, but
only if permeability of the nonchannel portion of
membranes were substantially lower than the Overton
prediction.
The present study evaluates CO2 permeation across
membranes and determines how this affects CO2 vent-
ing across tissue. Experiments were performed on
wild-type human cell lines (HCT116, T47D, and NHDF-
Ad) under physiological conditions of pH, salt compo-
sition, and temperature to obtain biologically meaning-
ful data. To determine whether membranes are
barriers to CO2 transport, gas diffusivity was measured
in compact, multicellular spheroids (comprising
5000 20,000 cells; ref. 4) that have adequately long
diffusion distances for resolving time delays. Effective
CO2 diffusivity across tissue (Dt,CO2) is approximated by
series resistances, as shown in Eq. 1:
(Dt,CO2)1 (Dc,CO2)1 (r Pm,CO2)1 (1) incubated overnight with mouse monoclonal anti-AQP1, rab-
where r is the spacing between permeability barriers
(3). Since Dt,CO2 is a function of Pm,CO2, tracing the
CO2 TRANSPORT IN TISSUE
spread of CO2 across tissue is a means of determining
whether permeation is rate limiting. For instance,
membrane permeation cannot be rate limiting if Dt,CO2
were equal to the CO2 diffusivity in aqueous solutions
containing macromolecules (i.e., protein) to account
for tortuosity. Transport of the acidic gas CO2 can be
measured in spheroids by using intracellular pH (pHi)
changes as a spatial readout, in combination with a
carefully parameterized diffusion-reaction algorithm to
derive CO2 dynamics from pH data. The geometry and
relative homogeneity of spheroids greatly simplify the
mathematical description of CO2 fluxes.
We demonstrate that membranes of HCT116 and
T47D cells, with only basal (channel-independent)
water and NH3 permeability, are highly permeant to
CO2, even in the presence of proposed gas channel
inhibitors, and that CO2 venting is not increased across
membranes of NHDF-Ad cells that have aquaporin-
facilitated water permeability. These findings are in
agreement with the Overton rule, i.e., that membranes
are barriers to the movement of NH3, but not CO2.
Instead, the rate-limiting step for CO2 transport is its
diffusion across aqueous compartments, which is re-
stricted by tortuosity imposed by macromolecules such
as proteins. Although gas channels may increase mem-
brane CO2 permeability further, this effect would not
be physiologically important for CO2 venting from
tissue.
MATERIALS AND METHODS
Cell culture and 3-dimensional (3-D) spheroid growth
Human colorectal HCT116 and neuroblastoma MC-IXC
(ATCC, Teddington, UK), colon HT29, glioblastoma U87,
breast T47D (kind gifts from Adrian Harris, University of
Oxford, Oxford, UK), normal NHDF-Ad fibroblasts, and
InMyoFib myofibroblasts (Lonza, Slough, UK) were used.
Cells were grown in Dulbeccos modified Eagles medium
containing NaHCO3, in an atmosphere of 5% CO2 for 48 72
h until 70 90% confluent. For experiments performed on
intact 2-dimensional (2-D) monolayers, cells were grown to
confluency in Lab-Tek 4-chambered borosilicate coverglass
(Nunc; Fisher Scientific, Loughborough, UK). Spheroids
(25) were cultured using the hanging drop method (100
HCT116 cells/20 l, 2000 T47D cells/20 l, or 2000
NHDF-Ad cells/20 l) in HCO3-containing DMEM for 23
d until attainment of spherical symmetry and a radius of
100 200 m (5000 20,000 cells).
Western blotting
Confluent cells were lysed for 30 min (1% TritonX-100; 150
mM NaCl; 50 mM Tris, pH 7.5; 0.5% Nonidet P-40; 50 mM
NaF; and Complete Protease Inhibitors; Roche, Burgess Hill,
UK) on ice. Proteins were separated by electrophoresis and
transferred to PVDF membrane (Bio-Rad, Hemel Hemstead,
UK). Membranes were blocked in 5% skimmed milk and
bit monoclonal anti-AQP5, rabbit polyclonal anti-AQP0, anti-
AQP6, and anti-AQP9 antibodies (Novus, Cambridge, UK) at
4C overnight or 1 h in the presence of HRP-conjungated
2763
anti--actin antibody (Cell Signaling Technology, Hitchin,
UK) at room temperature. Membranes were then incubated
for 1 h with anti-mouse and anti-rabbit HRP-conjungated
secondary antibodies (Novus) and developed (Pierce ECL;
Thermo Scientific, Loughborough, UK).
Bulk superfusion and microstream superfusion
Imaging experiments on dissociated cells were performed in
a Perspex superfusion chamber with a coverslip base. Imaging
of 2-D monolayers was performed in Lab-Tek chambers (in
which cells were also cultured). Lab-Tek chambers were also
used for imaging 3-D spheroids, settled on the glass coverslip.
Chambers were mounted on a Zeiss Observer Z1 microscope,
coupled to a LSM 700 confocal imaging system (Carl Zeiss,
Oberkochen, Germany). Bulk superfusate, heated to 37C,
was delivered at 4 ml/min. Switching between superfusates
was achieved using a valve. For rapid solution switching or for
producing sharp partitioning of extracellular solutes, a dou-
ble-barrel pipette was used (26). Filled with 2 solutions gravity
fed from reservoirs, the pipette was maneuvered into the field
of view, no further than 600 m from a cell of interest, or
directly on top of a confluent monolayer. Flow was adjusted to
alternate between the 2 microstreams or to generate 2
parallel microstreams.
Solutions and drugs
HEPES-buffered solution contained 125 mM NaCl, 20 mM
HEPES, 4.5 mM KCl, 1 mM MgCl2, 1 mM CaCl2, and 11 mM
glucose, pH 7.4; osmolarity 300 mOsm/kg. Hypoosmotic
solution had [NaCl] reduced by 150 mM. Hyperosmotic
solution had 300 mM d-mannitol added. NH3/NH4-contain-
ing solutions had NaCl replaced with NH4Cl. CO2/HCO3-
buffered solutions had HEPES replaced with 22 mM
NaHCO3, and the solution was bubbled with 5% CO2 (for pH
7.4) or 20% CO2 (for pH 6.8). Pretreatment with PCMB was
conducted for 15 min at 2 mM in alkaline solution (pH 7.8)
to increase solubility. Chemicals were obtained from Sigma-
Aldrich (Gillingham, UK).
Confocal imaging and photolytic uncaging
Fluorescence was recorded confocally. Monolayers and
spheroids were scanned in x-y-t mode (acquisition 2 Hz).
Linescan mode was used for fast (400 Hz) acquisition. The
following wavelength settings were used: 5-(and-6)-carboxy
SNARF-1 (cSNARF1): excitation (ex) 555 nm, emission
(em) 580/640 nm; calcein: ex 488 nm, em 525 nm;
fluorescein-5-(and-6)-sulfonic acid: ex 405/488 nm, em 525
nm; and 7-amino-4-methylcoumarin or Hoechst-33342: ex
361 nm, em 450 nm. Dyes were AM loaded into cells in
suspension or 2-D monolayers (cSNARF1 or calcein) for 5
min; into 3-D spheroids (cSNARF1) for 30 min. cSNARF was
calibrated by the nigericin technique (25). Dyes were pur-
chased from Life Technologies (Paisley, UK).
RESULTS
Measuring membrane water permeability in
dissociated cells
The extent to which membranes are barriers to CO2
transport can be most readily appreciated in cells that
lack functional activity of postulated gas channels.
2764 Vol. 28 July 2014 The FASEB Journal www.fasebj.org
Water-conducting aquaporins are the principal candi-
dates for gas channels (13, 18, 27); therefore, cells with
only basal water permeability (Pf) are most likely to
have unfacilitated CO2 permeability. Pf was measured
in intact cells from the rate of dilution of cytoplasmic
calcein fluorescence upon decreasing extracellular os-
molarity from 300 to 150 mOsm. Rapid solution change
was achieved by alternating flow between 2 mi-
crostreams (300 and 150 mOsm) emerging from a
square-bore double-barrel pipette (Supplemental Fig.
S1A and ref. 26) placed upstream of a dissociated
colorectal HCT116 cell (radius 6.750.16 m). From
the time course of cell volume (Fig 1Ai), Pf was 25 1
m/s (20), which is equal to the intrinsic water perme-
ability of membranes lacking aquaporins (refs. 21, 28,
29; see Supplemental Data). A similar result (221
m/s) was obtained in T47D breast cells (radius
8.370.19 m). Aquaporin inactivators p-chloromer-
curibenzoic acid (PCMB; pretreatment with 2 mM for
15 min) or Hg2 (1 mM added to both microstreams)
did not reduce Pf, further arguing for the absence of
functional aquaporins in HCT116 and T47D cells (Fig
1Aii). To eliminate the possibility that cell dissociation
affects aquaporin activity, hypotonic swelling was mea-
sured in HCT116 spheroids (radius 1434 m), which
retain normal cell-cell contacts. Hg2 (1 mM) did not
reduce the initial rate of hypotonicity-induced swelling
(Fig. 1B), confirming the single-cell observations.
According to the gas channel hypothesis, cells with
aquaporin-facilitated water permeability may also have
raised CO2 permeability. In 6 cell lines tested for
aquaporin-1 immunoreactivity, the highest expression
was found in NHDF-Ad fibroblasts (Fig 1Ci). Functional
measurements on NHDF-Ad cells confirmed a substan-
tially raised water permeability (Pf625 m/s) that
was PCMB/Hg2 sensitive (Fig. 1A) and may be attrib-
uted to aquaporin-1 as well as other aquaporin iso-
forms. Previously, aquaporins 0, 1, 5, 6, and 9 have been
proposed to facilitate CO2 transport (27), and their
expression was tested in HCT116, T47D, and NHDF-Ad
cells (Fig 1Cii). Aquaporins 1 and 9 were detected in
NHDF-Ad fibroblasts but not in HCT116 and T47D
cells (Fig 1Cii). In summary, HCT116 and T47D mem-
branes have negligible aquaporin-dependent water per-
meability. In contrast, water permeates NHDF-Ad mem-
branes via aquaporins that include postulated CO2
channels.
Measuring membrane NH3 permeability in
dissociated cells
Many of the proposed CO2 gas channels, including
aquaporins (27) and Rh proteins (30), are also pro-
posed to conduct NH3. Compared with CO2, the NH3
molecule is smaller but substantially less lipophilic, and
therefore, its basal membrane permeability is expected
to be low. HCT116 membrane permeability to NH3
(Pm,NH3) was determined from pHi changes (reported
by AM-loaded cSNARF1) in response to rapid switching
between NH3-containing and NH3-free microstreams
HULIKOVA AND SWIETACH

using the double-barrel pipette (n.b.: both solutions
buffered with 20 mM HEPES at pH 7.4; the former
also contained poorly membrane permeant NH4).
Entry of NH3 raises pHi due to cytoplasmic NH3
protonation, and NH3 efflux evokes the opposite pHi
response (Fig 2Ai). The pHi response is not rate
limited by the rapid chemical reactions, but inade-
quately fast solution exchange would underestimate
Pm,NH3. The rate of solution switching was measured
in separate experiments by labeling one microstream
fluorescently with fluorescein (30 M). The time
constant of the appearance of fluorescence was 20 2
ms (Supplemental Fig. S1B), 2 orders of magnitude
faster than the NH3-evoked pHi changes, which con-
firms that NH3 permeation across membranes, but not
solution delivery, is rate limiting. Pm,NH3 was calculated
from the initial slope of pHi change and the buffering
capacity, determined from the overall pHi change at
steady state (see Supplemental Data). Diffusion in the
unstirred intracellular compartment may delay pHi
CO2 TRANSPORT IN TISSUE
Figure 1. Membrane water permeability. A)
Dissociated cells, AM-loaded with calcein. i)
Time course of fluorescence used as index of
volume-change in response to halving of extra-
cellular osmolarity (reduction of [NaCl] by 150
mM). ii) Water permeability (Pf) calculated
from initial rate of volume-change in HCT116
cells; Pf was not reduced by pretreatment with 2
mM p-chloromercuribenzoic acid (PCMB) for
15 min or superfusion with 1 mM Hg2 (n20,
20, 15). Experiments also performed on T47D
cells (n51, 20). n.b.: basal Pf of membranes
without water channels is 25 m/s. NHDF-Ad
fibroblasts as positive control for PCMB/Hg2-
sensitive, channel-facilitated Pf (n24, 14, 11).
*P 0.05. B) HCT116 spheroids. i) Rate of
swelling of calcein-loaded spheroids in re-
sponse to halving of extracellular osmolarity
was not reduced in presence of 1mM Hg2
(n10). ii) Initial slope. C) Western blot. i)
Aquaporin-1 expression in a panel of 6 spher-
oid-forming cell lines: highest expression in
NHDF-Ad fibroblasts. Experiments repeated af-
ter 48 h incubation with dimethyloxaloylglycine
(DMOG; 1 mM) to test for hypoxic response. ii)
NDHF-Ad fibroblasts positive for aquaporins 1
and 9 (proposed CO2 gas channels).
changes, but a diffusion-reaction model estimated
this to account for only 6% of the total resistance, the
remainder attributable to membrane permeation.
Pm,NH3 across HCT116, T47D, and NHDF-Ad mem-
branes (Fig. 2A) was only modest and comparable to
the permeability across lipid bilayers (31, 32). Further-
more, Pm,NH3 was pHi independent and not affected by
4,4=-diisothiocyano-2,2=-stilbene-disulfonic acid (DIDS)
at a dose that inhibits Rh protein (30 M) or 15 min
pretreatment with PCMB (2 mM), arguing that NH3
permeability is not channel facilitated.
Estimating the lower limit for membrane CO2
permeability in HCT116 membranes
The experimental approach for measuring Pm,NH3 in
single cells cannot normally evaluate Pm,CO2 because
the chemical reactions linking pHi with intracellular
CO2 are rate limiting. Rapid solution switching between
CO2/HCO3-free (20 mM HEPES-buffered) and 5%
2765
Figure 2. Membrane NH3 permeability, intra-
cellular CA activity, and cell-cell cytoplasmic
coupling. A) Dissociated, cSNARF1-loaded
HCT116 cells exposed for 30 s to 20 mM
NH3/NH4 (removal shown only). i) Rapid
solution change achieved by alternating flow
between 2 microstreams from a double-barrel
pipette (exchange-time tested by labeling one
microstream with 30 M fluorescein). ii) NH3
permeability calculated from pHi time course;
permeability unaffected by DIDS (30 M) or
DIDS plus PCMB (2 mM) (n28, 29, 31).
Experiments repeated on T47D (n30, 15) and
NHDF-Ad (n16, 15) cells. B) HCT116 cells. i)
Extracellular pCO2 raised to 5% and then
lowered to 0%. ii) Rate of pHi change is rate
limited by intracellular CA activity. DIDS (30
M) and PCMB (2 mM) did not block CA
activity (n310/datapoint). C) i) Monolayer
of calcein-loaded HCT116 cells; central cell
photobleached with 488 nm laser. ii) Fluores-
cence did not recover (i.e., absence of cell-cell
coupling; n30). Asterisk indicates point of
photobleaching. Cell diameter: 16.7 0.5 m.

CO2/22 mM HCO3-containing microstreams evoked
a pHi change in HCT116 cells (Fig 2Bi), which was
highly pH sensitive and reduced by the CA inhibitor
acetazolamide (ATZ; 100 M). Thus, the pHi response
provides a measure of intracellular CA activity. Figure
2Bii plots CA activity as the ratio of the CO2 hydration
rate normalized to measurements in ATZ. CA activity was
not abolished in the presence of DIDS (30 M) or
pretreatment with PCMB (2 mM). The fastest CO2 hydra-
tion rate measured was 5 s1, which is attainable only
with a Pm,CO2 10 m/s; otherwise, membrane perme-
ation would be rate limiting (i.e., Pm,CO2 3/r kf).
Confluent HCT116 monolayers are not coupled by
gap junctions
Many types of cells are coupled into a syncitium via
intercellular channels that include gap junctions. If
available, cell-to-cell transport of NH3 or CO2 could
take the form of NH4 or HCO3 diffusion and an
associated H flux. Without syncitial cytoplasm, fast
transport of gases between cells is restricted to the
permeation of the uncharged gas molecule across
membranes. Cell-to-cell diffusive coupling was investi-
gated in calcein-loaded confluent HCT116 monolayers
using fluorescence recovery after photobleaching
(FRAP). Fluorescence in a photobleached cell did not
recover (Fig. 2C), demonstrating the absence of cou-
pling. Similar observations were made in the presence
of the gap junctional blocker carbenoxolone (CBX;
100 M).
Cell-to-cell coupling was tested further by measuring
the diffusive dissipation of a pHi gradient generated
across HCT116 monolayer (Supplemental Fig. S2). The
monolayer was uniformly acid loaded by exposure to 20
mM NH3/NH4-containing solution for 5 min, fol-
lowed by Na-free HEPES-buffered solution to block
Na-dependent pHi recovery by Na/H exchange
(NHE). A double-barrel pipette, supplying Na-free
and Na-containing microstreams in parallel, was then
placed over the monolayer to activate NHE-dependent
pHi recovery in the Na-exposed region. If the mono-
layer cytoplasm were diffusively coupled, delayed pHi
recovery would be observed in cells under the Na-free
microstream. However, even after 10 min, there was no
evidence for diffusive dissipation of the pHi gradient,
confirming the absence of coupling.
Fluorescent dye diffusion in multicellular spheroids
Fluorescent dyes were used to probe the diffusive
properties of spheroids and provide information on
tortuosity that is required for deriving CO2 diffusivity
later. Highly lipophilic 7-amino-4-methylcoumarin
(AMC; partition coefficient 15), applied to spheroids
(radius 1176 m) from the bulk superfusate (100

2766 Vol. 28 July 2014 The FASEB Journal www.fasebj.org HULIKOVA AND SWIETACH
M), is expected to diffuse freely across spheroid
membranes, its transport restricted only by macromo-
lecular volume exclusion and chemical binding. At
37C, the apparent AMC diffusion coefficient in spher-
oid tissue (Dt,AMC), calculated from normalized fluores-
cence time courses at different spheroid depths (Fig.
3A), was 260 20 m2/s, that is, 30% of AMC
diffusivity in water estimated from the Stokes radius of
the molecule (Table 1). Cooling to 20C reduced
Dt,AMC by 38%, as expected of the temperature sensi-
tivity of diffusion (2%/C). Hg2 (1 mM), which
would target membrane proteins, did not affect Dt,AMC,
as expected for the diffusion of a highly lipophilic
compound that does not require facilitation by chan-
nels. In dissociated cells, steady-state intracellular AMC
fluorescence was 2.0-fold higher than the ambient
extracellular signal, indicative of extensive chemical
binding. This factor, in addition to tortuosity, reduces
DAMC below its level in pure water.
Diffusion in the spheroids continuous extracellular
CO2 TRANSPORT IN TISSUE
space was probed using membrane-impermeant fluo-
rescein-5-(and-6-)-sulfonic acid (FS; 30 M; ref. 4). The
apparent FS diffusion coefficient (Dt,FS) in HCT116
spheroids (radius 1416 m) was 236 14 m2/s
(Fig. 3B), giving an extracellular tortuosity factor (e)
of 0.38 (based on FS diffusivity in water at 621 m2/s,
calculated from Stokes radius; Table 1). Tortuosity
arises because of increased path length around impen-
etrable cells (polar-to-cartesian remapping demon-
strates that the spacing between fluorescent spaces was
13 m, i.e., the predicted cell diameter). Based on the
flame photometrically determined amount of K re-
leased on lysis of a known volume of spheroid mass
(washed in ice-cold K-free medium), the volume
fraction of the intracellular space (vi) was found to be
0.677 (see Supplemental Data). Thus, the extracellular
volume fraction, ve 1 vi 0.323, is in agreement
with tortuosity e. Dt,FS was not greatly different in
larger HCT116 spheroids (radius 34036 m; Fig
3Biii), suggesting that the degree of cell-cell packing
Figure 3. Diffusivity in spheroids. A) Diffusiv-
ity of membrane-permeant AMC. Scale bar
100 m. i) Fluorescence time course normal-
ized to end point (solid lines) at different
depths (5 of 9 layers shown), 37C. Best fit to
diffusion equation (dotted lines). ii) Delays at
50% rise in fluorescence (histograms) and
best fit (gray line) (n8). iii) Experiment
repeated in 1 mM Hg2 and at 20C (n8, 9).
*P 0.05. B) Diffusivity of membrane-imper-
meant fluorescein-5-(and-6-)-sulfonic acid
(FS). Scale bar 100 m. i) Fluorescence
time course normalized to end point (solid
lines) at different depths (4 of 9 layers shown),
37C. Best fit to diffusion equation (dotted
lines). ii) Delays at 50% rise in fluorescence
(histograms) and best fit (gray line) (n10).
iii) Measurements repeated on larger HCT116
spheroids, HCT116 spheroids hyperosmotically
shrunken with 300 mOsm mannitol, and on
NHDF-Ad spheroids. C) NH3 transport in
HCT116 spheroids. i) Top: intracellular pH
time courses (reported with cSNARF) under
high extracellular buffering (40 mM HEPES).
Bottom: spheroid extracellular pH time courses
(reported with pH-sensitivity of FS; 30 M)
measured in 40 and 1 mM [HEPES]. ii) Peak
[H]e evoked by NH3 influx with 40 mM (n5)
and 1 mM HEPES (n6). Best-fit intrinsic
extracellular buffering: 13.6 mM/pH. iii) Delay
from point of NH3 addition to the midpoint of
pHi change (estimated by back-extrapolating
pHi time course to the point of solution
change). Gray curve: best-fit diffusivity is 21% of
NH3 diffusivity in pure water (n8). Shaded
area: model predictions assuming unrestricted
diffusion (lower limit) to extracellular diffusion
only (upper limit).
2767
TABLE 1. Model parameters for HCT116 spheroids

Symbol Definition Value Reference

vi Intracellular volume-fraction of spheroid-mass 0.677 Supplemental Data

ve Extracellular volume-fraction of spheroid-mass 0.323 vi ve 1
R Spheroid outer radius Variable Measured
r Cell radius 6.75 m Fig. 1A
Surface area-volume ratio 3/r From r
DAMC AMC diffusion coefficient in water (37C) 871 m2/s 41
DFS FS diffusion coefficient in water (37C) 621 m2/s 41
DH H diffusion coefficient in water (37C) 1.5 104 m2/s 42
DNH4 NH4 diffusion coefficient in water (37C) 2544 m2/s 42
DNH3 NH3 diffusion coefficient in water (37C) 2790 m2/s 42
DHCO3 HCO3 diffusion coefficient in water (37C) 1800 m2/s 1
DCO2 CO2 diffusion coefficient in water (37C) 2500 m2/s 1
DHepes Hepes diffusion coefficient in water (37C) 740 m2/s 41
e Extracellular tortuosity due to volume-exclusion by cells 0.381 Fig. 2A
Pm,NH3 NH3 membrane permeability (37C) 147 m/s Fig. 1C
KNH4 NH4 acid-dissociation constant (37C) 109.03 M 43
KCO2 CO2 acid-dissociation constant (37C) 106.1 M 43
KHepes Hepes acid-dissociation constant (37C) 107.5 M 33
kp Fast protonation constant (37C) 5 1010 s1 33
kh Spontaneous CO2 hydration constant 0.19 s1 Supplemental Fig. S4A
i Intracellular intrinsic buffering capacity Variable Measured
e Extracellular intrinsic buffering capacity 13.6 mM/pH Fig. 2Ciii
cai Intracellular CA activity Variable Measured
pHi0 Resting pHi Variable Measured

is constant. Dt,FS was smaller (17111 m2/s) in
NHDF-Ad spheroids (ve0.33), indicating that extracel-
lular diffusive pathways are cell type dependent (Fig
3Biii). Hyperosmotic solutions (addition of 300 mOsm
mannitol) reduced HCT116 spheroid radius by 45% to
98 2 m within 5 min. This treatment reduced
tortuosity e to 0.185, implying a greater path length
around shrunken cells (Fig 3Biii).
NH3 transport in HCT116 spheroids is rate limited by
membrane permeation
Further characterization of diffusion in HCT116 spher-
oids was performed by measuring the pHi response to
NH3/NH4-containing solution (at pH 7.4, 37C). Fig-
ure 3Ci shows intracellular alkalinization in spheroids
(radius 142.67.9 m) reported with cSNARF1 at
different depths. The response arises from NH3 entry
and protonation inside cells, and it provides a readout
of the diffusive spread of NH3, but additional informa-
tion is required to obtain an accurate estimate of tissue
NH3 diffusivity (Dt,NH3; Supplemental Data). Intracel-
lular H-buffering capacity (i) was measured from the
pHi change, back extrapolated to the point of NH3/
NH4 addition. Extracellular H-buffering capacity
(e) was measured in separate experiments using the
pHe reporter FS (30 M), which measures the ampli-
tude of pHe-transients (pHe) associated with extracel-
lular NH4 deprotonation triggered by NH3 entry into
cells. Experiments were performed using solutions con-
taining 40 mM and 1 mM HEPES buffer (Fig 3Cii), so
that e could be determined from the relationship
between [HEPES] and pHe amplitude; best-fit e was
13.6 mM/pH.
Dt,NH3 was estimated by best-fitting experimental pHi
time courses with a diffusion-reaction model parame-
terized for volume (vi, ve), tortuosity (e), buffering
capacity (i, e), and NH3/NH4 reaction kinetics
(33). NH4 and H ions were assumed to be mem-
brane impermeant but were able to facilitate extracel-
lular NH3 diffusion. Figure 3Ciii quantifies time delays
from the start of solution change to the midpoint of
pHi change at different spheroid depths. The shaded
area shows the range of model predictions, from the
shortest delays with unrestricted diffusion, to the lon-
gest delays with extracellular diffusion only. The best-fit
Dt,NH3 was 0.21 0.03 DNH3, that is, 5-fold lower than
NH3 diffusivity in water (DNH32.79103 m2/s; Table
1). At least part of this restriction must arise from
rate-limiting membrane permeation, as expected from
measurements of Pm,NH3 in dissociated HCT116 cells
(Fig 2Ai).
Measuring CO2 diffusivity in HCT116, T47D, and
NHDF-Ad spheroids
NH3 permeability across HCT116 membranes (Fig. 2A)
is sufficiently low to restrict NH3 diffusivity in spheroids
(Fig. 3A). To determine whether membrane permea-
bility is also rate limiting for CO2 transport, CO2
diffusivity (Dt,CO2) was measured in spheroids. Dt,CO2
was computed from spatiotemporal pHi dynamics
evoked by raising bulk superfusate pCO2 to 5% for 4
min, and then lowering back to 0% at 37C (Fig. 4A).
Solutions bubbled with CO2 also included 22 mM
HCO3 to keep pH at 7.4 (n.b.: CO2-free solutions were
buffered by 20 mM HEPES). At the end of each
experiment, the response to NH3/NH4 addition was

2768 Vol. 28 July 2014 The FASEB Journal www.fasebj.org HULIKOVA AND SWIETACH

recorded to provide a reference value for membrane-
limited diffusivity and to determine the rate of solution
change (NH3-evoked alkalinization at the spheroid
periphery is essentially instantaneous).
At the spheroid periphery, the slope of the pHi
response to changes in pCO2 depends on intracellular
buffering capacity and CA activity. Intracellular buffer-
ing capacity was measured from the amplitude of pHi
change on CO2 addition/removal, back extrapolated to
the point of solution change. Intracellular CA activity
was then derived from the rate of pHi change. Expo-
sure of cells to CO2/HCO3 may drive HCO3 influx
in parallel to CO2 influx, resulting in an ablated pHi
change. The magnitude of this flux was investigated by
switching to Cl-free medium to drive net HCO3
influx in exchange for Cl. The pHi change associated
with HCO3 influx was an order of magnitude slower
CO2 TRANSPORT IN TISSUE
Figure 4. CO2 diffusivity in spheroids. A) i) pHi
time course (at 3 spheroid depths as indicated
in icon), offset to starting pHi, in response to
raising CO2 tension to 5%, then back to 0% and
finally to 20 mM NH3/NH4; 30 M AMB
included throughout to inhibit exofacial car-
bonic anhydrase activity (and thereby minimize
degree of enzyme-facilitated CO2 diffusion).
Dotted line: best-fit model simulation. ii) pHi
time courses normalized to the amplitude of
pHi change, estimated by back-extrapolating
time courses to the point of solution-switch. iii)
Time delays at different spheroid depths mea-
sured from the point of solution switch to the
midpoint pHi (n15). Shaded area: predicted
delays assuming unrestricted diffusion (lower
limit) to extracellular diffusion only (upper
limit). Gray line: best fit diffusivity is 76% of
CO2 diffusivity in water. B) Experiment in the
presence of 30 M DIDS on spheroids pre-
treated with 2 mM PCMB (n12). i) pHi time
course in response to changes in CO2 tension.
ii) pHi time courses normalized to the ampli-
tude of pHi change. iii) Time delays at different
spheroid depths.
than the pHi change associated with CO2/HCO3
exposure, suggesting that CO2 traffic exceeds HCO3
transport considerably (Supplemental Fig. S4B). Intra-
cellular acidification evoked by CO2 addition may also
be ablated by NHE activity, but the NHE blocker
5-(N,N-dimethyl)amiloride (DMA; 30 M) did not alter
the initial rate of pHi change (Supplemental Fig. S4C).
Delays in the pHi response recorded at increasing
spheroid depths are due to CO2 diffusion in gaseous
form plus facilitated diffusion (HCO3H) in the
extracellular space (n.b.: absence of cytoplasmic cou-
pling excludes facilitated diffusion between cells). To
improve the power to resolve Dt,CO2, time delays were
exacerbated by eliminating CA-facilitated CO2 diffu-
sion in the extracellular space by treatment with the
membrane-impermeant CA inhibitor 4-aminomethyl
benzenesulfonamide (AMB; 30 M; Supplemental Fig.
2769
S3). Intracellular CA activity (cai) can also influence
time delays because the chemical conversion to (and
from) HCO3 in electrically uncoupled cells is equiva-
lent to a CO2 buffering effect. To account for this, pHi
time courses were best fitted to Dt,CO2 and cai simulta-
neously. Parameter sensitivity analysis demonstrates
that 2-way fitting produces a unique best fit because the
pHi slope at the surface of the spheroid is strongly
dependent on cai and not Dt,CO2. Figure 4Aiii shows the
time delays from the start of solution change to the
midpoint of pHi change at different spheroid depths
(radius 100.45.3 m). The shaded area depicts the
range of simulated delays, from shortest with unre-
stricted diffusion (Dt,CO2DCO22.5103 m2/s) to
longest with diffusion in extracellular spaces only.
The experimental data are in agreement with fast
CO2 diffusion (only 244% slower than in water),
compared with the more restricted NH3 diffusivity
(794% slower than in water; Fig. 5A). Comparable
Dt,CO2 measurements [0.770.08DCO2; Fig. 5A and
Supplemental Fig. S5] were made in spheroids of
T47D cells that have only basal water and NH3
permeability (radius 19915 m; e0.54). Earlier
(Fig. 1), NHDF-Ad cells were shown to have high
water permeability due to aquaporins (Fig. 1). To test
whether these channels can also facilitate CO2 transport,
Dt,CO2 was measured in NHDF-Ad spheroids (radius 913
m; e0.33) in the presence of 100 M carbenoxolone
to block gap junctions. Despite high water permeability,
Dt,CO2 was not increased [0.810.12DCO2; Fig. 5A]. In
summary, CO2 diffusivity in spheroids of 3 different
cell-types was fast (77% of its value in pure water).
CO2 diffusivity in HCT116 spheroids is reduced by
protein crowding but not membrane permeation
To test whether blockers of gas channels affect Dt,CO2,
experiments were repeated on HCT116 spheroids of 2
different sizes (radius: 122.33.7 and 189.32.9 m)
Figure 5. Summary of diffusivity measurements.
A) CO2 and NH3 diffusion coefficients in
HCT116, T47D, and NHDF-Ad spheroids (rel-
ative to diffusivity in pure water) estimated by
best fit (n15, 10, 15, 12, 15, 4, 4, 8, 8). B)
Intracellular CA activity by best fit. C) Relation-
ship between gas diffusivity and protein density
(normosmotic and hyperosmotic conditions).
D) Predicted relationship between CO2 perme-
ability (Pm,CO2) and diffusivity in spheroids
(Dt,CO2), normalized to diffusivity in pure water
(DCO2). Relationship calibrated against experi-
mental data for NH3 transport (indicated by
star; see Supplemental Data).
2770 Vol. 28 July 2014 The FASEB Journal www.fasebj.org
pretreated with PCMB (2 mM for 15 min) and then
superfused with solutions containing DIDS (30 M, a
dose that would block Rh protein gas channels; ref. 25
and Fig. 4B). These drugs did not affect Dt,CO2 or cai
(Fig. 5A, B), arguing that CO2 transport is not channel
facilitated.
Volume exclusion by the presence of protein has a
well-documented effect on decreasing gas diffusivity
(2). The concentration of protein in HCT116 spheroid
lysates was 75 g/L spheroid, which would reduce overall
diffusivity by 26% (refs. 1, 2; see Supplemental Data).
Thus, volume exclusion by proteins accounts fully for
the decrease in Dt,CO2, relative to diffusivity in water,
but further restrictions must apply for NH3. Increasing
protein density by hyperosmotic treatment with 300
mOsm mannitol decreased both Dt,CO2 and Dt,NH3, as
expected for greater volume exclusion (Fig. 5A). The
relationship between diffusivity and relative protein
density shows that Dt,CO2 extrapolated to zero protein
would equal the diffusivity in pure water, whereas
Dt,NH3 would remain greatly restricted (Fig. 5C). A
plausible explanation for this difference is that mem-
brane permeation is rate limiting for the transport of
NH3, but not CO2. This condition is met if Pm,CO2 is
104 m/s (Fig. 5D; derivation in Supplemental Data).
HCT116 apical membranes have high CO2
permeability
Studies of normal colonocytes have shown that apical
membrane CO2 permeability (Pam,CO2) is significantly
lower than basolateral permeability (15). To test
whether HCT116 apical membranes impose a resis-
tance to CO2 transport, one half of a monolayer was
exposed to elevated (20%) CO2, and the pHi response
beyond this region was measured to determine the
extent to which CO2 spreads diffusively between cells.
Low Pam,CO2 would favor a greater spread of CO2
because of smaller apical leakage. Sharp extracellular
HULIKOVA AND SWIETACH
CO2 partitioning was produced by a double-barrel
pipette releasing 2 parallel microstreams bubbled with
20 or 5% CO2 (both containing 22 mM HCO3). The
degree of extracellular CO2 partitioning was visualized
in separate experiments by labeling 1 microstream
fluorescently with 30 M fluorescein (Fig. 6A). Figure
6A shows that the steady-state spatial profile of pHi was
dissipated more than extracellular CO2 partitioning.
The lateral spread of [H]i was quantified in terms of a
spatial constant , which is related to cytoplasmic
Figure 6. A) Rapid CO2 transport across monolayers. i)
Extracellular CO2 gradient produced by parallel microstream
flow of 20 and 5% CO2 solution. Extracellular CO2 partition-
ing visualized by labeling one microstream with fluorescein in
separate experiments. Resulting pHi gradient measured at
steady state (n5). Experiments repeated with 100 M ATZ
to inhibit CA activity (n5). ii) Lateral dissipation of pHi
gradient quantified in terms of spatial constant (see text).
B) Transmission of acid along HCT116 monolayer from a
local H-load produced cumulatively by repeated (1 Hz)
UV-photolysis of caged H-compound 2-nitrobenzaldehyde
(NBA) permeated passively into cells from superfusate. i) pHi
time courses in uncaging region (R1) and three serially
adjoining regions (R2R4; n20). ii) Lateral [H] profile
measured after 4 min of uncaging. Best fit: exponential decay
constant of 18 m. iii) Length constant measured under the
indicated experimental conditions (n10 15).
CO2 TRANSPORT IN TISSUE
diffusivity (Dc,CO2), monolayer height (hmono) and
eff
effective apical membrane permeability (Pam,CO2 ) by
Eq. 2:
2 hmono Dc,CO2 Pam,CO2
eff
(2)
eff
Pam,CO2 combines Pam,CO2 with diffusion across the
monolayer (a mean distance of hmono/2) and the
overlying unstirred layer (height of husl):
eff
(Pam,CO2 )1 (Pm,CO2
eff
)1 hmono (2 Dc,CO2)
hus1 DCO2 (3)
HCT116 hmono, measured by z scanning, was 18.3
3.0 m, which corresponds to 1.5-cell layers, as indi-
cated by nuclear staining using Hoescht-33342. Based
on fluid mechanics (34), husl was estimated to be 10.7
m (see Supplemental Data). Dc,CO2 was estimated
from CO2 diffusivity in spheroids (Dt,CO2) weighted by
the intracellular volume fraction (vi0.677). Thus,
Dc,CO2 was approximated as (0.76 ve)/vi DCO2
0.64 DCO2. In HCT116 monolayers, was 17.7 0.7
m. As expected from a pHi gradient at steady state,
inhibition of CA activity with ATZ did not affect (Fig
6Aii). Similar measurements were obtained with a
different colon cell line, HT29. According to Eq. 2,
eff
Pam,CO2 in HCT116 monolayers is equal to 112 m/s.
This value can be attributed entirely to diffusive resis-
tances, which add up to an apparent permeability of
[hmono/(2 Dc,CO2) husl/DCO2]1 104 m/s.
Equation 3 therefore predicts an unresolvably high
Pam,CO2.
Rapid CO2 diffusion can transmit acid between
uncoupled cells
Rapid diffusive spread of CO2 across multicellular
tissue is a means for transmitting H ions in the
absence of cytoplasmic continuity. This was tested in
HCT116 monolayers by measuring the dissipation of a
local intracellular H load. Monolayers were super-
fused with the membrane-permeant caged-H com-
pound 2-nitrobenzaldehyde (NBA; 1 mM), which re-
leases H ions on exposure to UV light (35). A local
intracellular H load was produced by restricting UV
exposure to a 30 m wide strip of the monolayer (Fig.
6B; n.b.: rapid superfusion washes away any extracellu-
lar H-load). Repeating UV flashing over 4 min gener-
ates an acid-load, some of which may dissipate into
neighboring cells. pHi was measured in the H-uncag-
ing region (R1) and in eight 15-m-wide adjoining
regions (R2R9). Under superfusion with 5% CO2/22
mM HCO3-buffered solution and in the presence of
DMA (30 M) and DIDS (150 M) to block acid
extrusion by NHE and HCO3 transport, respectively,
substantial acid loading was measured in regions adja-
cent to the uncaging site (Fig 6Bi). Since the cell
diameter in monolayers is only 16.7 0.5 m (Fig. 2C),
the appearance of acid in R3 must arise from cell-to-cell
transmission. The lateral [H] profile at 4 min of H
uncaging was fitted with an exponential length con-
stant of 18 m. This was not affected by CBX and only
2771
modestly reduced by ATZ (100 M). In the absence of
CO2/HCO3, the length constant was reduced to 13
m (i.e., H spread limited to R1 and R2 only). Lactic
acid, produced endogenously by HCT116 cells, could
conceivably carry acid between cells via H-monocar-
boxylate transport (MCT), but the length constant was
unaffected by the MCT inhibitor -cyano-4-hydroxycin-
namate (5 mM) or by raising intracellular lactate by
superfusion with 10 mM lactate. In summary, rapid
diffusive spread of CO2, unhindered by membrane
permeation, can carry a substantial H-equivalent flux
across tissue in the absence of a cytoplasmic syncitium.
DISCUSSION
Gas channels cannot facilitate the transport of gases
that cross the lipid matrix rapidly
Membrane permeability to small molecules, such as
water, NH3, and CO2, is a function of the properties of
the lipid matrix and embedded proteins (11, 13, 16,
21). The lipid matrix has only low water permeability
(25 m/s; refs. 28, 29); therefore, aquaporin water
channels have a physiologically important role in con-
ducting water across membranes. The lipid bilayer is
more permeable to NH3 than to water (31, 32), but
since the oil:water partition coefficient of NH3 is con-
siderably smaller than 1, bilayers are expected to be
substantial barriers to NH3 transport (11, 32). Conse-
quently, gas channels may increase membrane NH3
permeability, as shown by others (13, 20, 36).
CO2 transport is of major importance to all cells
because of the magnitude of CO2 production and the
need for constant venting. As the oil:water partition
coefficient of CO2 is 1, permeation across the lipid
bilayer is expected to be rapid (6, 7, 11), to the extent
that the aqueous compartments on either side of the
membrane become the principal resistances to flux.
However, the findings that apical membranes of gastric
and colon epithelia (14, 15) are substantially tighter to
CO2 compared with red cell membranes (18) and that
aquaporins and Rh-proteins can increase membrane
CO2 permeability (Pm,CO2; refs. 13, 17, 18) have been
interpreted to indicate that the lipid bilayer is a barrier
to CO2 transport. CO2 passage through gas channels
would explain how Pm,CO2 can be regulated physiolog-
ically, but this requires background (gas channel-inde-
pendent) Pm,CO2 to be 23 orders of magnitude lower
than the prediction based on partition coefficients (6,
7) and measurements on lipid bilayers (10, 11). The
present study tested this experimentally on native hu-
man cell membranes by measuring diffusivity of CO2
and NH3 in tissue growths, in which the series arrange-
ment of multiple membranes, repeated over long dis-
tances, would reveal rate-limiting permeability barriers
(Eq. 1).
2772 Vol. 28 July 2014 The FASEB Journal www.fasebj.org
Unfacilitated NH3 membrane permeability is rate
limiting to NH3 transport globally in tissue
NH3 permeability (Pm,NH3) in HCT116, T47D, and
NHDF-Ad membranes was comparable to the basal
permeability across lipid bilayers (31, 32) and was
unaffected by proposed blockers of gas channels (Fig.
2A). These findings argue for the absence of functional
NH3 gas channels, despite the presence of aquaporins 1
and 9 and high water permeability in NHDF-Ad fibro-
blasts. The marginally lower water (Fig. 1A) and NH3
(Fig. 2A) permeability in T47D membranes, compared
with HCT116, may reflect differences in lipid chemis-
try. NH3 permeability measurements were performed
on individual cells under superfusion, using a rapid
solution-switching device (26) to minimize unstirred
layer artifacts that can arise in cell suspensions (21, 34).
Relative to diffusion across typical cytoplasmic dimen-
sions (radius r510 m), membrane NH3 permeation
is expected to be rate limiting for NH3 transport
globally (since Pm,NH3rDNH3). This was confirmed
by studying the spread of NH3 across spheroid tissue
growths using spatiotemporal pHi data as the readout.
NH3 diffusivity in HCT116 spheroids (Dt,NH3) was re-
duced by 79% relative to its value in pure water (Fig.
5A). Volume exclusion by proteins would reduce Dt,NH3
by only 26%; therefore, measured Dt,NH3 is consistent
with additional barriers to transport, such as the surface
membrane and, possibly, membranous organelles. The
y axis intercept of the relationship between Dt,NH3 and
protein density (Fig. 5C) indicates that these additional
barriers reduce NH3 diffusivity by 2/3. Equation 1,
which relates independently measured Dt,NH3 (in spher-
oids) and Pm,NH3 (in dissociated cells), estimates that
permeability barriers (pairs of membranes) are 10
m apart. This spacing is consistent with the diameter
of cells and argues that the major membrane barrier to
NH3 transport is the continuous surface membrane,
rather than membranous organelles (imaged in Sup-
plemental Fig S6).
Membrane CO2 permeability is rapid and not rate
limiting for CO2 transport in tissue
The rapid solution-switching approach applied to iso-
lated cells cannot determine Pm,CO2 accurately because
the pHi response evoked by changing extracellular
pCO2 is rate limited by intracellular CA activity, rather
than permeation across membranes (Fig. 2B). Instead,
Pm,CO2 was investigated by measuring CO2 diffusivity
(Dt,CO2) across spheroids. This approach was confirmed
to demonstrate rate-limiting NH3 permeability in
HCT116, T47D, and NHDF spheroids (Fig. 5A). Spa-
tiotemporal pHi data were used to probe the spread of
CO2 across spheroids in response to a change in
superfusate pCO2. Although H-yielding CO2 hydra-
tion is rate limited by intracellular CA activity locally,
time delays in pHi changes at different spheroid depths
report the spread of CO2. These delays, readily resolv-
able in large spheroids, can ascertain whether CO2
HULIKOVA AND SWIETACH
diffuses freely (delays of a few seconds) or is restricted
by membrane permeation (delays of tens of seconds).
Dt,CO2 in HCT116 and T47D spheroids was found to be
only 24% lower than diffusivity in water (Fig. 5A),
despite the functional absence of aquaporin water
channels, i.e., gas channel candidates (Fig. 1). Dt,CO2 in
HCT116 spheroids was unaffected by the presence of
blockers of proposed gas channels (PCMB and DIDS
targeting aquaporins and Rh proteins, respectively).
Comparable Dt,CO2 measurements were made in
NHDF-Ad fibroblasts, which have high, aquaporin-facil-
itated water permeability mediated by at least 2 iso-
forms (AQP1 and AQP9) proposed to conduct CO2
(Fig 1Cii). These lines of evidence argue that gas
channels do not facilitate CO2 transport in tissue. The
24% reduction in Dt,CO2 was not due to substantially
differing permeability properties of the apical and
basolateral membranes (Fig. 6A). Instead, the modest
decrease in CO2 diffusivity could be fully accounted for
by volume exclusion due to proteins. Osmotic shrink-
age, which increases protein crowding but does not
change membrane width or area (i.e., Pm,CO2), reduced
Dt,CO2 (Fig. 5C), indicating that diffusion across aque-
ous compartments is the rate-limiting process. More-
over, Dt,CO2 extrapolated to zero protein (Fig. 5C) was
equal to free CO2 diffusivity in pure water, further
arguing that surface and organellar membranes cannot
be barriers to CO2 transport. Considering series resis-
tors (Eq. 1), this condition would be met when Pm,CO2
exceeds the ratio of cytoplasmic CO2 diffusivity and cell
radius by at least an order of magnitude, i.e., 104
m/s (Fig. 5D; see Supplemental Data), which is in
good agreement with theoretical predictions (6, 11)
and earlier experimental measurements (10, 11).
A number of notable studies performed on guinea-
pig colonocytes (15), MDCK cells (16), tsA201 cells
(16), and aquaporin-negative red cells (17, 18) derived
a much reduced Pm,CO2 (15150 m/s). Over this
range, Dt,CO2 in spheroids would have been substan-
tially lower than measured values herein. In all 3 cell
types studied (HCT116, T47D, and NHDF-Ad), Dt,CO2
approached the diffusive limit for protein-containing
aqueous solutions. Thus, the present method of using
tissue growths to study CO2 diffusivity has resolved a
higher Pm,CO2. It is conceivable that in specific cases of
low CO2 permeability, as suggested for gastrointestinal
apical membranes (14, 15) or gas channel-negative red
cells (18), the atypical abundance of integral proteins
or cholesterol may reduce Pm,CO2 to a fraction of that
measured herein. This would, however, necessitate a
tightening of CO2 permeability by several orders of
magnitude. For example, by largely excluding the lipid
bilayer from membranes, the necessary conditions can
be met for gas channels to raise Pm,CO2 meaningfully.
An alternative explanation for low apparent permeabil-
ity in special cases of membranes is diffusive resistance
from large unstirred layers (11), independently of lipid
matrix properties. However, such a diffusive resistance
in series with the membrane cannot be overcome by gas
channels.
CO2 TRANSPORT IN TISSUE
CO2 venting from tissue can involve the interstitial
and transcellular routes. The relative contributions of
these to overall CO2 transport will depend on the
compartment volumes and the degree to which CO2
flux is carried in the form of HCO3 and H ions, i.e.,
facilitated diffusion. In tissues with a large intracellular
volume fraction, syncitial cytoplasm, and high intracel-
lular CA activity, a large fraction of CO2 traffic will be
transcellular. In contrast, the interstitial route will
dominate in tissues with large extracellular spaces,
absence of cytoplasmic continuity between cells and
high exofacial CA activity. It is noteworthy that the
latter is characteristic of many tumors (3739), where
CO2 venting is critical because of high metabolic rates
and long diffusion distances due to inadequate blood
perfusion (40).
CONCLUSIONS
The permeability sequence (waterNH3CO2) mea-
sured in native, human-derived membranes is in agree-
ment with Overton rule predictions based on oil:water
partition coefficients. Measurements in 3 different
types of tissue growths argue that natural membranes
are not rate-limiting barriers to CO2 transport. Instead,
CO2 venting from tissue is rate limited by diffusion in
aqueous compartments, and this process cannot be
facilitated by the insertion of gas channels into mem-
branes. The findings of this study are important for
identifying rate-limiting steps for the fundamentally
important process of CO2 venting.
This study was supported by the Association for Interna-
tional Cancer Research and the Royal Society. The authors
thank Professor Richard D. Vaughan-Jones for comments on
the manuscript, and mentoring, Professor Kenneth W.
Spitzer (University of Utah, Salt Lake City, UT, USA) for
training in the dual microperfusion technique, Dr. Daniele
Dini (Imperial College London, London, UK) for discussions
on fluid mechanics, and Dr. Steven Niederer (Kings College
London, London, UK) for advice on diffusion modeling.
REFERENCES
1. Geers, C., and Gros, G. (2000) Carbon dioxide transport and
carbonic anhydrase in blood and muscle. Physiol. Rev. 80,
681715
2. Gros, G., and Moll, W. (1971) The diffusion of carbon dioxide
in erythrocytes and hemoglobin solutions. Pflgers Arch. 324,
249 266
3. Nitsche, J. M. (1999) Cellular microtransport processes: inter-
cellular, intracellular, and aggregate behavior. Annu. Rev.
Biomed. Eng. 1, 463503
4. Swietach, P., Patiar, S., Supuran, C. T., Harris, A. L., and
Vaughan-Jones, R. D. (2009) The role of carbonic anhydrase 9
in regulating extracellular and intracellular pH in three-dimen-
sional tumor cell growths. J. Biol. Chem. 284, 20299 20310
5. Swietach, P., Wigfield, S., Cobden, P., Supuran, C. T., Harris,
A. L., and Vaughan-Jones, R. D. (2008) Tumor-associated car-
bonic anhydrase 9 spatially coordinates intracellular pH in
three-dimensional multicellular growths. J. Biol. Chem. 283,
2047320483
2773
 6. Missner, A., and Pohl, P. (2009) 110 years of the Meyer-Overton
rule: predicting membrane permeability of gases and other
small compounds. Chemphyschem 10, 14051414
7. Verkman, A. S. (2002) Does aquaporin-1 pass gas? An oppos-
ing view. J. Physiol. 542, 31
8. Simon, S. A., and Gutknecht, J. (1980) Solubility of carbon
dioxide in lipid bilayer membranes and organic solvents.
Biochim. Biophys. Acta 596, 352358
9. Mathai, J. C., Missner, A., Kugler, P., Saparov, S. M., Zeidel,
M. L., Lee, J. K., and Pohl, P. (2009) No facilitator required for
membrane transport of hydrogen sulfide. Proc. Natl. Acad. Sci.
U.S.A. 106, 1663316638
10. Gutknecht, J., Bisson, M. A., and Tosteson, F. C. (1977) Diffu-
sion of carbon dioxide through lipid bilayer membranes: effects
of carbonic anhydrase, bicarbonate, and unstirred layers. J. Gen.
Physiol. 69, 779 794
11. Missner, A., Kugler, P., Saparov, S. M., Sommer, K., Mathai, J. C.,
Zeidel, M. L., and Pohl, P. (2008) Carbon dioxide transport
through membranes. J. Biol. Chem. 283, 25340 25347
12. Forster, R. E., Gros, G., Lin, L., Ono, Y., and Wunder, M. (1998)
The effect of 4,4=-diisothiocyanato-stilbene-2,2=-disulfonate on
CO2 permeability of the red blood cell membrane. Proc. Natl.
Acad. Sci. U.S.A. 95, 1581515820
13. Boron, W. F. (2010) Sharpey-Schafer lecture: gas channels. Exp.
Physiol. 95, 11071130
14. Waisbren, S. J., Geibel, J. P., Modlin, I. M., and Boron, W. F.
(1994) Unusual permeability properties of gastric gland cells.
Nature 368, 332335
15. Endeward, V., and Gros, G. (2005) Low carbon dioxide perme-
ability of the apical epithelial membrane of guinea-pig colon. J.
Physiol. 567, 253265
16 Itel, F., Al-Samir, S., Oberg, F., Chami, M., Kumar, M., Supuran,
C. T., Deen, P. M., Meier, W., Hedfalk, K., Gros, G., and
Endeward, V. (2012) CO2 permeability of cell membranes is
regulated by membrane cholesterol and protein gas channels.
FASEB J. 26, 51825191
17. Endeward, V., Cartron, J. P., Ripoche, P., and Gros, G. (2008)
RhAG protein of the Rhesus complex is a CO2 channel in the
human red cell membrane. FASEB J. 22, 64 73
18. Endeward, V., Musa-Aziz, R., Cooper, G. J., Chen, L. M.,
Pelletier, M. F., Virkki, L. V., Supuran, C. T., King, L. S., Boron,
W. F., and Gros, G. (2006) Evidence that aquaporin 1 is a major
pathway for CO2 transport across the human erythrocyte mem-
brane. FASEB J. 20, 1974 1981
19. Nakhoul, N. L., Davis, B. A., Romero, M. F., and Boron, W. F.
(1998) Effect of expressing the water channel aquaporin-1 on
the CO2 permeability of Xenopus oocytes. Am. J. Physiol. Cell
Physiol. 274, C543548
20. Musa-Aziz, R., Chen, L. M., Pelletier, M. F., and Boron, W. F.
(2009) Relative CO2/NH3 selectivities of AQP1, AQP4, AQP5,
AmtB, and RhAG. Proc. Natl. Acad. Sci. U.S.A. 106, 5406 5411
21. Boron, W. F., Endeward, V., Gros, G., Musa-Aziz, R., and Pohl, P.
(2011) Intrinsic CO2 permeability of cell membranes and
potential biological relevance of CO2 channels. Chemphyschem
12, 10171019
22. Missner, A., Kugler, P., Antonenko, Y. N., and Pohl, P. (2008)
Passive transport across bilayer lipid membranes: overton con-
tinues to rule. Proc. Natl. Acad. Sci. U.S.A. 105, E123; author reply
E124
23. Somersalo, E., Occhipinti, R., Boron, W. F., and Calvetti, D.
(2012) A reaction-diffusion model of CO2 influx into an oocyte.
J. Theor. Biol. 309, 185203
24. Swietach, P., Vaughan-Jones, R. D., and Harris, A. L. (2007)
Regulation of tumor pH and the role of carbonic anhydrase 9.
Cancer Metastasis Rev. 26, 299 310
25. Hulikova, A., Vaughan-Jones, R. D., and Swietach, P. (2011)
Dual role of CO2/HCO3(-) buffer in the regulation of intracel-
2774 Vol. 28 July 2014 The FASEB Journal www.fasebj.org
lular pH of three-dimensional tumor growths. J. Biol. Chem. 286,
1381513826
26. Spitzer, K. W., Ershler, P. R., Skolnick, R. L., and Vaughan-
Jones, R. D. (2000) Generation of intracellular pH gradients in
single cardiac myocytes with a microperfusion system. Am. J.
Physiol. Heart Circ. Physiol. 278, H1371H1382
27. Geyer, R. R., Musa-Aziz, R., Qin, X., and Boron, W. F. (2013)
Relative CO(2)/NH(3) selectivities of mammalian aqua-
porins 0 9. Am. J. Physiol. Cell Physiol. 304, C985C994
28. Mathai, J. C., Mori, S., Smith, B. L., Preston, G. M., Mohandas,
N., Collins, M., van Zijl, P. C., Zeidel, M. L., and Agre, P. (1996)
Functional analysis of aquaporin-1 deficient red cells. The
Colton-null phenotype. J. Biol. Chem. 271, 1309 1313
29. Preston, G. M., Carroll, T. P., Guggino, W. B., and Agre, P.
(1992) Appearance of water channels in Xenopus oocytes ex-
pressing red cell CHIP28 protein. Science 256, 385387
30. Geyer, R. R., Parker, M. D., Toye, A. M., Boron, W. F., and
Musa-Aziz, R. (2013) Relative CO(2)/NH(3) permeabilities of
human RhAG, RhBG and RhCG. J. Membr. Biol. 246, 915926
31. Antonenko, Y. N., Pohl, P., and Denisov, G. A. (1997) Perme-
ation of ammonia across bilayer lipid membranes studied by
ammonium ion selective microelectrodes. Biophys. J. 72, 2187
2195
32. Walter, A., and Gutknecht, J. (1986) Permeability of small
nonelectrolytes through lipid bilayer membranes. J. Membr. Biol.
90, 207217
33. Eigen, M. (1964) Proton transfer acid-base catalysis enzymatic
hydrolysis .i. elementary processes. Angew. Chem. Int. Ed. Engl. 3,
119
34. Pedley, T. J. (1983) Calculation of unstirred layer thickness in
membrane transport experiments: a survey. Q. Rev. Biophys. 16,
115150
35. Swietach, P., Spitzer, K. W., and Vaughan-Jones, R. D. (2007)
pH-Dependence of extrinsic and intrinsic H()-ion mobility in
the rat ventricular myocyte, investigated using flash photolysis of
a caged-H() compound. Biophys. J. 92, 641653
36. Holm, L. M., Jahn, T. P., Moller, A. L., Schjoerring, J. K., Ferri,
D., Klaerke, D. A., and Zeuthen, T. (2005) NH3 and NH4
permeability in aquaporin-expressing Xenopus oocytes. Pflgers
Arch. 450, 415428
37. Clauss, M. A., and Jain, R. K. (1990) Interstitial transport of
rabbit and sheep antibodies in normal and neoplastic tissues.
Cancer Res. 50, 34873492
38. Loewenstein, W. R., and Kanno, Y. (1966) Intercellular commu-
nication and the control of tissue growth: lack of communica-
tion between cancer cells. Nature 209, 1248 1249
39. Swietach, P., Hulikova, A., Vaughan-Jones, R. D., and Harris,
A. L. (2010) New insights into the physiological role of carbonic
anhydrase IX in tumour pH regulation. Oncogene 29, 6509 6521
40. Vaupel, P., Kallinowski, F., and Okunieff, P. (1989) Blood flow,
oxygen and nutrient supply, and metabolic microenvironment
of human tumors: a review. Cancer Res. 49, 6449 6465
41. Tucker, W. A., and Nelken, L. H. (1982) Diffusion coefficients
in air and water. In Handbook of Chemical Property Estimation
Methods (Lymna, W. J., Reehl, W. F., and Rosenblatt, D. H., eds)
pp.17-117-50, American Chemical Society, Washington, DC,
USA
42. Vanysek, P. (1999) Ionic conductivity and diffusion at infi-
nite dilution. In CRC Handbook of Chemistry and Physics, Thermo-
chemistry, Electrochemistry and Kinetics (Linde, D. R., ed) pp.
9395, CRC Press, London
43. Leem, C. H., Lagadic-Gossmann, D., and Vaughan-Jones, R. D.
(1999) Characterization of intracellular pH regulation in the
guinea-pig ventricular myocyte. J. Physiol. 517, 159 180
Received for publication September 25, 2013.
Accepted for publication March 10, 2014.
HULIKOVA AND SWIETACH