As the basic principle for mass transportation through diffusive means, Adolf Fick described that the diffusive flux (J) was related to the diffusion coefficient (D) and gradient of the substrate (in case of two phases, rewritten as the concentration difference between the two phases ∇ φ ) concentration: J = − D ∇ φ (1)

Later, Meyer and Overton established the rule of spontaneous permeation of solutes and solvents across membranes, stating that the flux of the substance across a membrane, J, was linearly dependent on the permeability of the membrane, P _m, with a concentration difference ∆ c s at two surfaces of the membrane, when the partition coefficient of the substance, K _p, was given.

Eq. 1 can be rewritten as following: J = − P m ∙ ∆ c s (2)

Where P m = K p ∙ D m ∕ d , D _m is the diffusion coefficient, d is the thickness of the membrane.

As indicated by the rule, the permeability of a given molecule is related to the partition coefficient K p . Therefore, gas molecules, such as CO₂ would have a membrane permeability as fast as permeating a water layer, with a K p ≥1. However, experimental data has shown contradictory results with extremely low gas permeability from certain biological membranes. As proposed by Pohl’s group, the existence of an unstirred layer, which dominates the resistance of CO₂ diffusion, and the variation in the thickness of this layer could explain for this discrepancy (Missner et al., 2008b). In 2011, a joint correspondence letter was published that summarized the main agreement and disagreement on channel protein-mediated CO₂ diffusion by the research groups Boron, Gros and Pohl (Boron et al., 2011). In summary, they agreed that channel-mediated CO₂ transport would require a high resistance of the non-channel part of the membrane to CO₂ diffusion and relatively low resistance to CO₂ from USLs. In 2015, further cross-talk was initiated to collect new comments or views on CO₂ transport mediated by channel proteins under physiological conditions by Cooper, Occhipinti, and Boron (Cooper et al., 2015). In this proposal, a new “access-solubility-diffusion-egress” model was proposed, where resistance of non-channel proteins, different headgroup of lipids, the role of cholesterol, as well as USLs, all accounted for the apparent CO₂ permeability of biological membranes. While the disagreement still remains, Pohl pointed out the concern of data generated by both stopped flow and mass spectrometry, due to the fast process of CO₂ diffusion in the range of milliseconds. Furthermore, it could be the limited availability of carbonic anhydrases (CAs), which led to extremely low CO₂ permeability to the apical membranes. Finally, new points were raised: 1) How to explain the role of sterols and high percentage of membrane proteins, on the diffusion of diffusion of CO₂ of biological membranes? 2) Mutation work that influences the function of certain aquaporin, resulting in the change of P M , C O 2 , could not be correctly mapped to the change in thickness of USLs. 3) The altered activity CA activity of certain cells was not correlated with CO₂ permeability. 4) The existence of USLs still challenged the proponents of CO₂ channels.

CO₂ and O₂ are gas molecules that play crucial roles in respiration by providing energy through oxidative phosphorylation reactions. Both gases need to be exchanged efficiently between the cellular organelles and the atmosphere, guided by their osmotic gradients. Unlike animals, plant cells or other photosynthesis microorganisms take up CO₂ as a substrate for photosynthesis, and the concentration gradient is less significant compared to animals (Uehlein et al., 2017). Therefore, a higher efficient diffusion of CO₂ from the atmosphere to the chloroplast stroma, where photosynthesis occurred, would be more beneficial for photosynthesis-active organisms (Kaldenhoff et al., 2014).

For quite a long time, the resistance of the mesophyll to CO₂ was overlooked for green-leaf plants. Instead, the regulation of stroma and CO₂ interconversion to bicarbonate and protons catalyzed by carbonic anhydrases (CA) was considered to be the limiting factor in CO₂ availability (Kaldenhoff, 2012). However, even with complete deletion of CA activity in the chloroplast stroma, photosynthesis decreased by only about 7% (Price et al., 1994; Kaldenhoff, 2012). Furthermore, the mesophyll CO₂ conductance varied rapidly in response to temperature, light, or water stress, instead of having a relatively constant value. This contradicted the pure physical model of mesophyll CO₂ diffusion. Together, these shreds of evidence pointed to the existence of other major factors that regulated CO₂ diffusion, such as aquaporin-mediated transportation.

The physiological influence of altered expression of potential permeable CO₂ aquaporins was recently systematically evaluated and reviewed by Evans‘ group (Groszmann et al., 2017). To understand the role of certain putative permeable aquaporins with CO₂, transgenic plants were generated and the impact on parameters relevant to photosynthesis was determined. In general, the change of mesophyII conductance was correlated with the tuned expression level of corresponding aquaporins. However, the mesophyII drawdown should be negatively correlated with the mesophyII conductance and the CO₂ assimilation rate, where causal links between AQP and mesophyII conductance can be established. To provide the general ranges of photosynthetic-related parameters under varied mesophyII conductance, they performed simulations by changing the mesophyII conductance, when either stomatal conductance or C _i was set to be constant. They gave an estimated range of mesophyII drawdown, CO ₂ assimilation rates, transpiration rate, and transpiration efficiency based on consistent literature data (Groszmann et al., 2017). Furthermore, mesophyII conductance is a combined feature that could be influenced by many factors other than membrane permeability, such as the chloroplast surface area, adjacent to the intercellular air space per unit of leaf area and cell wall thickness (Evans, 2021). Since 2014, more direct or indirect evidence accumulated, supporting the aquaporin-facilitated CO₂ transportation, i.e., AtPIP2;5, SlPIP1;2 (from tomato), OsPIP1;2 and OsPIP1;3 (from rice), HvPIP2;1, 2;2, 2;3, 2;5 (from Barley), ZmPIP1;5, 1;6 (from Maize), as well as SiPIP2;7 from C4 plant-foxtail millet (see Table 1).

Recent studies have shown that the influence of altered expression of potential CO₂ permeable aquaporins on the mesophyII conductance and photosynthesis rate should be calibrated by growth and environmental conditions, as well as the oligomeric/phosphorylation status of the corresponding aquaporins. Although there are accumulated evidences for aquaporin-mediated CO₂ transportation, there have also been studies that have shown that simple manipulation of these aquaporins did not lead to changes in mesophyII conductance or photosynthetic efficiency. In one study, the knockout of three aquaporin genes-AtPIP1;2, AtPIP1;3, AtPIP2;6 from Arabidopsis thaliana did not result in changes in mesophyII conductance nor photosynthetic efficiency. The authors discussed possible reasons for these results: i) functional redundancy within aquaporin families; ii) the possible change in hydraulic conductance together with the higher light intensities (200 μmol m⁻² s⁻¹) altered the photosynthetic capacity, which would be sufficient to remove the effect on both g _m and g _s; iii) altered the hydraulic conductance of mutant lines through functional stimulation by colocalization of PIP1s and PIP2s on the plasma membrane (Kromdijk et al., 2020). However, the hydraulic conductance of mutant lines was not measured in the above study, which left this question to be further investigated. In another case, the ectopic expression of either AtPIP1;2 or AtPIP1;4 in tobacco did not further increase mesophyII conductance nor the rate of assimilation of CO₂. Similarly, the authors pointed out the influence of plant growth and environmental conditions on the ability of certain CO₂ permeable aquaporins to alter mesophyII conductance, particularly, when a high basal g _m was observed in control wild-type control plants (Clarke et al., 2022). This effect was also observed from rice PIPs (Huang et al., 2021) and tomato SiPIP1;2 knockout mutants (Kelly et al., 2014), where g _m was affected only when grown in a CO₂ enriched environment. Other studies have pointed out that the oligomeric or phosphorylation state of overexpressed CO₂ permeable aquaporins can directly impact their function (Otto et al., 2010; Maurel et al., 2015; Groszmann et al., 2017). Additionally, aquaporins can act as signaling molecules, responding to different environmental stimuli and regulating stomatal dynamics in response to changes in ambient CO₂ concentration (Ding and Chaumont, 2020) or ABA-mediated biotic stress (Fang et al., 2019). Finally, one important aspect to consider is the relative humidity within the substomatal cavity, which was assumed to be saturated when calculating the intercellular CO₂ concentration determined by the gas exchange experiment (Cernusak et al., 2018). As recently investigated by Farquhar’s group, the relative humidity within the substomatal cavity could drop down to around 80%, with the saturation edge retreating to the mesophyII cell walls. Surprisingly, the mesophyII conductance to CO₂ remained less affected when alter the Δw (the difference between saturated humidity and the humidity in the air) if compared to uncorrected data, which might be controlled by the aquaporins within the mesophyII cell membranes (Wong et al., 2022). Although there are several aquaporins reported to function as both water and CO₂ channels, the detailed mechanism of such potential dual functions still needs to be investigated, which could be investigated with new methods such as in situ measurement of water potentials within leaves using the fluorescent powder-hydrogel nanoreporters (Jain et al., 2021), as well as cell specific overexpress experiment to avoid functional redundancy from endogenous aquaporins using plant leave single cell RNA sequence data base (Kim et al., 2021).

To conclude, the impact of changes in CO₂ permeable aquaporins on mesophyll conductance and photosynthesis rate should be considered, with respect to growth and environmental conditions, in particular the relative humidity within the substomatal cavity, as well as the oligomeric and phosphorylation status of the corresponding aquaporins.

Since the discovery of the CO₂ channel protein: aquaporin-1 from humans and NtAQP1 from tobacco, many aquaporins from different organisms were reported to mediate CO₂ transport, covering many members from mammals, plants, microalgae, and fish (see Table 1). The family of aquaporins has a relatively conserved structure, with six membrane-spanning helices, two reentrant short helices with NPA motifs, and flexible N-/C-termini heading towards the cytosol. The six bundle-like membrane-spanning alpha helices were tightly arranged in a circle, constituting the solute conduction pore/channel. Although aquaporins function as the water channel in monomers, they often form a quaternary tetramer assembly in native membranes and even large orthogonal arrays in the case of AQP4 (Ho et al., 2009). Until now, the physiological relevance of such a tetrameric assembly is not completely clear; however, a few cases showed that the central pore formed by the aquaporin tetramer was likely to be the CO₂ channel. Early work based on X. laevis oocytes with low intrinsic CO₂ permeabilities provided experimental evidence that AQP1 acts as a permeable CO₂ channel (Nakhoul et al., 1998). Later, a molecular simulation based on the high-resolution structure of AQP1 gave the atomic level of details that the central pore of the AQP1 tetramer could mediate fast CO₂ diffusion in low intrinsic CO₂-permeable membranes (Hub and de Groot, 2006). This hypothesis was further demonstrated by the yeast protoplast system to determine the altered permeability of CO₂ when the assembly of the artificial tetramer with a fixed ratio of NtPIP1;2 and NtPIP2;1, connected by a short linker (Otto et al., 2010). The results demonstrated that the homo-tetrameric assembly of CO₂ permeable NtPIP1;2 was necessary for CO₂ channel activity. However, such a relationship between the oligomeric state and CO₂ permeability was not further investigated in other model plants, except for tobacco.

In 2021, Tyerman et al. gave a systematic review on multifunctional aquaporins, describing the dynamic regulation of the central pore with the high-resolution crystal structures of AQP1 and SoPIP2;1 (Tyerman et al., 2021). Furthermore, the MOLEonline MOLEonline channel radii analysis (Berka et al., 2012) showed a diameter of 3.6 Å at the Leu200 constriction residue, modelled by the closed water channel conformation of SoPIP2;1 (PDB: 1Z98) (Tornroth-Horsefield et al., 2006). According to the analysis, the constriction side would allow the CO₂ to pass when considering a kinetic diameter of 3.3 Å for CO₂. Additionally, Tyerman et al. (2021) proposed that both post-translational modification and protein-protein interactions could contribute to dynamic regulation of central pore permeability via local conformational changes, allowing a wide range of molecules, including K⁺, Na⁺, as well as CO₂ passing through the central pore. In early studies, aquaporins from the PIP1 family were found to be permeable to CO₂. While, later on, members from the PIP2 family were also reported to function as CO₂ channels. Despite the relative conservation of transmembrane helices in all PIP aquaporins, it was difficult to identify the crucial residues creating the selective filter of the central pore, which was not surprising given the very variable pore environment generated by tetrameric assembly. As seen in Figure1, the major part of the central pore lining area is composed of transmembrane helices 2 and 5 and loop D, which were dynamically influenced by other neighboring motifs as well. However, due to the lack of high-resolution structure of PIP1 aquaporin, it remains unresolved how the sequence difference between PIP1s and PIP2s could contribute to CO₂ permeability, especially the long N-terminal flexible loop that only exists in PIP1s. Although it is difficult to obtain structural details for the flexible loop region, a detailed biochemical assay might answer this question, such as domain switch or truncated variants in the case of an N-terminal loop. As indicated in Figure 1., conserved residues Leu in helix 5 and Ile at the end of loop E were reported to be essential to allow the passage of a CO₂ molecule based on either simulation or biochemical assays (Mori et al., 2014; Tyerman et al., 2021). However, other residues alone the channel might also be the restriction site, depending on the arrangement of the helixes structures that form the central channel.

The Singer-Nicolson fluid-mosaic model was widely recognized as the fundamental model for the structure and molecular dynamics of the plasma membrane (Singer and Nicolson, 1972). Many basic properties of biological membranes were characterized on the basis of this two-dimensional fluid model. Among the basic properties, the permeability was also intensively investigated using such a lipid bilayer model both theoretically and experimentally. However, other factors, such as sterols or integrated membrane proteins were not considered, which could influence the overall permeability (Suzuki and Kusumi, 2023). Therefore, lack of such factors could be the potential source for the inconsistency of measured membrane permeability. This inconsistency became non-trivial when determining CO₂ permeability. Due to the higher lipophilic properties of CO₂, the phospholipid-formed lipid bilayer exhibits very low resistance to CO₂, while the plasma membrane of X. laevis oocytes, Madin-Darby canine kidney (MDCK) cells, the transformed human embryonal kidney SV40 cell line (tsA201), as well as the apical membrane of the gastric glands, showed extremely low CO₂ permeability (Endeward et al., 2006a; Endeward et al., 2006b; Endeward et al., 2008; Itel et al., 2012). In a recent review, Gros et al. proposed that the cholesterol content in the majority biological membranes dominates its CO₂ permeability, regulating the CO₂ permeability by at least 2 orders of magnitude with a cholesterol content between 0%–70% (Arias-Hidalgo et al., 2018). However, an exception of normal native human red cells showed aquaporin-dependent CO₂ permeability instead of cholesterol content, indicating the existence of unidentified factors (Endeward et al., 2008). Kaldenhoff’s group suggested a possibility, pointing out the role of non-channel proteins on the CO₂ permeability of the phospholipid bilayer (Kai and Kaldenhoff, 2014). Finally, the existence of lipid rafts, which are rich in both sterols and proteins, could further contribute to the overall permeability [see review by Kai Simons and Elina Ikonen (Simons and Ikonen, 1997)].

One possible reason for the inconsistent permeability of CO₂ reported in many previous reports could be the limitations of different techniques in determining permeability of CO₂, due to the high permeability of the phospholipid bilayer (Endeward et al., 2014). Both stop flow-based and mass spectrometry-based methods were questioned for their inability to quantify dynamic fast CO₂ across the membrane (Boron et al., 2011; Hannesschlaeger et al., 2019). On the other hand, the scanning pH electrode could provide an alternative that was not limited by the fast dynamics of CO₂ passing through the black lipid membrane. However, the formation of a black lipid membrane with the solvent-containing method was challenged by the presence of organic solvent n-decan, as well as whether aquaporins still survive as a functional form during the formation of the corresponding black lipid membrane (Hannesschlaeger et al., 2019). Therefore, new techniques that can determine the fast transportation of CO₂ across the membrane and avoid the influence of solvents may be necessary to improve the accuracy of the CO₂ permeability measurements.