Periplasmic arabinogalactan glycoproteins act as a calcium capacitor that regulates plant growth and development


Author for correspondence:

Derek T. A. Lamport

Tel: +44 1273 486430



  • Arabinogalactan glycoproteins (AGPs) are implicated in virtually all aspects of plant growth and development, yet their precise role remains unknown. Classical AGPs cover the plasma membrane and are highly glycosylated by numerous acidic arabinogalactan polysaccharides O-linked to hydroxyproline. Their heterogeneity and complexity hindered a structural approach until the recent determination of a highly conserved repetitive consensus structure for a 15-sugar residue arabinogalactan subunit with paired glucuronic carboxyls.
  • Based on NMR data and molecular dynamics simulations, we identify these carboxyls as potential intramolecular Ca2+-binding sites.
  • Using rapid ultrafiltration assays and mass spectrometry, we verified that AGPs bind Ca2+ tightly (Kd ˜ 6.5 μM) and stoichiometrically at pH 5. Ca2+ binding is reversible in a pH-dependent manner.
  • As typical AGPs contain c. 30 Ca2+-binding subunits and are bulk components of the periplasm, they represent a Ca2+ capacitor discharged at low pH by stretch-activated plasma membrane H+-ATPases, hence a substantial source of cytosolic Ca2+. We propose that these Ca2+ waves prime the ‘calcium oscillator’, a signal generator essential to the global Ca2+ signalling pathway of green plants.


Eukaryotes exploit nonhomologous hydroxyproline-rich structural glycoproteins that have different functions reflecting the deep dichotomy between plants and animals. Collagen and elastin provide structural support and tensile properties of animal tissues. In plants, the extensin family (Lamport & Northcote, 1960) of uniquely O-Hyp glycosylated proteins (Lamport, 1967) consists of extensins (Lamport et al., 2011), proline-rich proteins (Fowler et al., 1999) and arabinogalactan proteins (AGP; Ellis et al., 2010). Extensins are self-assembling amphiphiles that help to form the new cell plate at cytokinesis (Cannon et al., 2008); crosslinked extensins also contribute to the tensile strength of the cell wall (Tire et al., 1994) and as a negative regulator of extension growth (Cleland & Karlsnes, 1967; Sadava & Chrispeels, 1973; Gille et al., 2009).

Classical AGPs comprise a medium-sized family of large (c. 120 kDa) glycoproteins containing c. 95% carbohydrate (Supporting Information, Table S1). Generally, each AGP consists of a small polypeptide (c. 200 residues) with numerous (12–24) O-Hyp-linked arabinogalactan polysaccharides (Hyp-AG; Zhao et al., 2002). AGPs cover the surface of the plasma membrane (Lamport et al., 2006) anchored via a C-terminal glycosylphosphatidylinositol (GPI) lipid (Oxley & Bacic, 1999; Sherrier et al., 1999; Svetek et al., 1999). Thus cleavage by phospholipase eventually releases freely soluble AGPs into the periplasm and then into the expanding cell wall matrix where their role as a pectic plasticizer remains a plausible conjecture (Lamport et al., 2006).

Arabinogalactan glycoproteins were originally discovered as polysaccharides (Aspinall et al., 1969) in suspension-cultured sycamore cells (Lamport, 1964). However, further purification showed that AGPs were glycoproteins ubiquitous in bryophytes and higher plants (Lamport, 1970; Jermyn & Yeow, 1975; Nothnagel, 1997; Seifert & Roberts, 2007; Ellis et al., 2010). Although implicated in numerous biological processes, the precise function of AGPs, ‘crucial to understanding the mysterious relationship between the cell wall, the plasma membrane, and the cytoplasm’ (Seifert & Roberts, 2007), remains unknown. Although frequently described as signalling molecules, receptor proteins for AGPs have not been found. Indeed, AGPs are relatively immobile and therefore unlikely to act as signalling molecules per se. However, we show here that they possess a pH-dependent Ca2+-binding glycomotif in abundance. As AGPs are also located at the plasma membrane and highly correlated with extension growth (Lee et al., 2005), we propose that AGPs are an essential component of the global Ca2+ signalling process in plants.

Materials and Methods

Molecular dynamics simulations

Three starting structures for Interferon Hyp-polysaccharide-1 (IFNHP1) were obtained from the molecular ensemble derived using NMR NOESY data (Tan et al., 2010). Each of these structures was placed in a truncated octahedral box of c. 5000 TIP3P water molecules, and ions were added to neutralize the system: one Ca2+ or two Na+ (Case et al., 2008). Parameters for sugars were taken from the GLYCAM_06g parameter set (Kirschner et al., 2008), and those for water and ions were taken from standard Amber libraries (Case et al., 2008). After standard equilibration procedures (Tan et al., 2010), each system was simulated unrestrained for 500 ns at constant pressure (101.3 kPa) and temperature (27° C) with coupling constants of 5 ps, using periodic boundary conditions and the Particle–Mesh Ewald method to calculate electrostatic interactions.

Preparation of AGPs

We dissolved gum arabic (Ratansey Damji & Co, New Bombay, India) directly into distilled water (10 mg ml−1) with no further treatment unless so stated. Closely related AGPs were prepared from either culture filtrates (final yield c. 5 mg 100 ml−1) or clarified tissue homogenates, by precipitation with the Yariv reagent (1,3,5-tri-(galactosyloxyphenylazo)-2,4,6-trihydroxybenzene) at a high ionic strength as previously described (Lamport et al., 2006). Here we substituted 2% CaCl2 for NaCl for some preparations (Table 1). This precipitated calcium pectate before addition of the Yariv reagent but yielded soluble AGPs fully loaded with Ca2+ (Table 1). Larch arabinogalactan and high-methoxyl pectin were obtained from Sigma-Aldrich.

Table 1. Uronic acid and Ca2+ content of arabinogalactan glycoproteins (AGPs) and pectin
SpeciesGlcA (μg mg−1)Ca2+ (μg mg−1)GlcA : Ca2+ (molar ratio)
  1. AGP, arabinogalactan glycoprotein; GalA, galacturonic acid; GlcA, glucuronic acid.

  2. a

    Average of nine samples; gum arabic was dissolved directly in distilled water without further precipitation by the Yariv reagent.

  3. b

    Gum arabic sample #8 initially had the highest GlcA : Ca2+ molar ratio of 3.5; it was then precipitated by the Yariv reagent in the presence of 2% CaCl2, resulting in a decreased GlcA : Ca2+ molar ratio, close to the theoretical value of 2.

  4. c

    Sample pectin contained 475 μg mg−1 methyl-esterified GalA. Note that pectin initially secreted into the wall (Albersheim et al., 2011) is close to 100% esterified in contrast to this 63% esterified apple pectin.

Samples prepared with no Ca2+ added
Gum arabica12710.12.8
Arabidopsis culture AGP855.33.6
Broccoli florets AGP818.72.1
Daucus carota culture AGP978.52.6
Nicotiana tabacum BY2 culture AGP17011.03.5
Solanum lycopersicum culture AGP15311.03.2
Samples prepared in the presence of Ca2+
Gum arabicb18015.82.6
N. tabacum BY2 culture AGP114132.0
N. tabacum SPa culture AGP869.42.1
S. lycopersicum culture AGP86102.0
Apple pectinc285 GalA19.73.3
Larch arabinogalactan00N/A

Ca2+ release as a function of pH

Ca2+-preloaded gum arabic and pectin solutions of 10 mg ml−1 were titrated with HCl from pH 5.0–1.5, and the released Ca2+ was separated by rapid ultrafiltration of 500 μl aliquots taken at 0.25 pH intervals. Preloaded gum arabic was prepared by precipitation by the Yariv reagent in the presence of CaCl2 as described earlier; preloading of pectin involved adding increasing amounts of Ca2+ until the binding capacity was exceeded at pH 5.0.

Calcium binding assay

A Ca2+-free gum arabic solution was prepared by titration to pH 2.0 with HCl, followed by ultrafiltration, washing twice with water at pH 2.0, and final adjustment to pH 5.0 with KOH. We used a constant 1 mg ml−1 gum arabic/AGP solution, and 25 mM Ca2+ was added to obtain a total Ca2+ concentration ranging from 0 to 0.8 mM Ca2+ in 2 ml samples. After 15 min equilibration time, the samples were ultrafiltered in Amicon (Millipore Corporation, MA, USA) ultrafiltration devices (30 kDa cutoff) for 35 min at 5000 g, followed by inductively coupled plasma mass spectrometry analyses of the ultrafiltrates. Preliminary assays were performed by spectroscopy (612 nm) of the ultrafiltrates (100 μl) added to 1 ml methylthymol blue reagent (containing 8-hydroxyquinoline; Gindler & King, 1972).

Uronic acid assay

We assayed 200 μl of gum arabic/AGPs aliquots for uronic acid via the m-hydroxy diphenyl assay as described earlier (Blumenkrantz & Asboe-Hansen, 1973).

Results and Discussion

Hyp-arabinogalactan polysaccharides isolated from AGPs (Lamport, 1977; Pope, 1977; Zhao et al., 2002; Xu et al., 2008) are built of small repetitive subunits. Each subunit (c. 15-sugar residues) consists of a trigalactosyl backbone motif that contains two bifurcated 6-sugar residue sidechains (Fig. 1). As each sidechain contains one glucuronic acid residue (GlcA), a single 15-residue subunit contains two glucuronic acids (Tan et al., 2004). Molecular dynamics simulations performed, in this laboratory, on the NMR ensemble of IFNHP1 (Tan et al., 2010) in the presence of Na+ showed a high degree of conformational flexibility. However, simulations with Ca2+ showed that a pair of sidechain GlcA carboxyls binds Ca2+ efficiently, and this intramolecular binding stabilizes the sidechain conformation (Fig. 1, Movie S1). Thus each Hyp-AG subunit comprises a putative Ca2+-binding subunit with a potential role in Ca2+ signalling at the surface of the plasma membrane.

Figure 1.

A conserved O-Hyp-linked arabinogalactan polysaccharide (Hyp-AG) glycomotif binds Ca2+. (a) Schematic representation of the primary structure of the consensus subunit. (b) Three-dimensional molecular model of the Hyp-AG Interferon Hyp-polysaccharide-1 (IFNHP1) with Ca2+ ion (green) bound by two glucuronic acid (GlcA) sidechains (red); the galactan backbone is shown in dark blue and sidechains in light blue. Molecular dynamics simulation of Ca2+ binding by IFNHP1 Hyp-AG. Time series along 500 ns molecular dynamics simulations show the distance between GlcA and cations Na+ (black) and Ca2+ (red) (c), and the distance between the two GlcA residues in the presence of Na+ and Ca2+ (d). Note that Ca2+ binds to GlcA within 100 ns and locks the sidechain conformation, in contrast with Na+, which does not bind to the GlcA carboxylates.

Four major questions arise. Does the experiment verify the predicted Ca2+ binding by AGPs? Is periplasmic AGP–Ca2+ sufficient to meet cellular demand? Precisely how can Ca2+ be released to meet that demand? How can Ca2+ release be related to rapid responses, particularly those involving tip growth of pollen tubes (Roy et al., 1999) and root hairs (Monshausen et al., 2007)? We answer these questions in the following sections.

AGPs bind Ca2+ stoichiometrically

Classical AGPs bind Ca2+ ranging from 0.5 to 1.6% w/w and have a GlcA content of 8–18% w/w (Table 1). This includes gum arabic with sidechains described 25 years ago (Defaye & Wong, 1986) and therefore used as an abundantly available model AGP (Akiyama et al., 1984; Ellis et al., 2010). Thus AGPs have GlcA : Ca2+ molar ratios of between 2 and 3.8, the lower value being consistent with the intramolecular binding predicted by molecular simulations. Significantly, AGPs purified by precipitation with the β-glycosyl Yariv reagent at high ionic strength CaCl2, rather than the usual NaCl, gave a binding ratio of c. 2 : 1, corresponding to stoichiometric Ca2+ binding by paired GlcAs. Thus, a single AGP molecule of c. 120 kDa (Zhao et al., 2002) that binds c. 1% w/w Ca2+ has c. 30 Ca2+-binding subunits, which constitutes a Ca2+-binding capacity around fivefold greater than that of classical Ca2+-binding proteins such as calmodulin. Note that high-methoxyl pectin (galacturonic acid) and larch arabinogalactan (no uronic acids) were included as positive and negative controls, respectively (Table 1). We evaluated the equilibrium Ca2+-binding constant of gum arabic by titration of Ca2+ into Ca2+-depleted gum arabic (Fig. 2a). This confirmed that c. 75% of GlcA forms Ca2+-binding sites with an associated Kd of 6.5 μM (Fig. 2b).

Figure 2.

Titration of Ca2+-depleted model arabinogalactan glycoprotein (AGP; gum arabic) with Ca2+ at pH 5. (a) Saturation binding data (black squares) with theoretical curve (red line). (b) Scatchard analysis of data with best-fit line resulting in a Kd of 6.5 μM. YGlcA is the fraction of glucuronic acid (GlcA) bound. pH-dependent cation release from Ca2+-saturated gum arabic: (c) release of Ca2+ (red) and Mg2+ (black); (d) release of Ca2+ from gum arabic (red) compared with pectin (black).

A periplasmic AGP–Ca2+ capacitor meets cellular demand

Capacitors store charge, and periplasmic AGPs store Ca2+. Therefore, as a potential source that can be discharged and recharged rapidly and repeatedly, AGPs comprise a Ca2+ capacitor. To test this hypothesis, as a first step, we calculated the amount of bound periplasmic Ca2+ from AGP membrane coverage of a spherical cell and light microscopy data, using the equation W = inline image, where W is the weight of AGP required to form a GPI-anchored monomolecular layer at the surface of the plasma membrane in 1 g FW tobacco BY–2 cells, r is the cell radius (c. 35 μm) corrected for cell asymmetry, n is the cell number g–1 FW (c. 5 × 106), w is the weight of a single AGP molecule (c. 120 kDa), and A is the area covered by a single AGP molecule (c. 180 nm2; Zhao et al., 2002).

Tobacco BY-2 cell cultures typify rapidly growing cells and yield a theoretical W of c. 85 μg AGP  g–1 cells (FW), while membrane-bound AGP determined experimentally (Lamport et al., 2006) is c. 200 μg AGP g–1 cells. Thus AGPs fully cover the plasma membrane; the surplus AGPs may be the result of surface irregularity of the plasma membrane or multiple AGP layers.

Previous qualitative cytochemical observations based on immunogold (Freshour et al., 1996) and green fluorescent protein (Zhao et al., 2002) labelling show AGPs concentrated and quasi-uniformly distributed at the plasma membrane. Furthermore, Ca2+ localizes as a fine layer along the plasma membrane rather than the wall (Slocum & Roux, 1982; Cramer et al., 1985). Simply connecting the above shows that AGPs and Ca2+ colocalize at the periplasmic surface of the plasma membrane. These calculations show that AGPs cover the membrane, and thus AGPs potentially comprise a quantitatively significant source of Ca2+: a nominal concentration of 50% (w/v) periplasmic AGP contains c. 10% (w/w) glucuronic acid and c. 1% (w/w) Ca2+. This yields a local concentration of bound Ca2+ of c. 125 mM, several orders of magnitude greater than basal cytosolic Ca2+ values of 0.1–0.3 μM (Holdaway-Clarke & Hepler, 2003). Indeed, during rapid tip growth, there is a reported Ca2+ flux of 20 pmol cm−2 s−1 (Holdaway-Clarke & Hepler, 2003). Thus, an AGP–Ca2+ layer of 5 nm × 1 cm2 with a volume of 500 pl contains 62 pmol of Ca2+ that would sustain a 3 s Ca2+ pulse of 20 pmol cm−2 s−1. An AGP–Ca2+ periplasmic interface that alternately discharges Ca2+ into the cytosol and is subsequently recharged can therefore act as a physical Ca2+ capacitor surrounding the cell. A feasible mechanism for the generation of a pH-dependent Ca2+ current is discussed in the following.

Low pH releases Ca2+ from the periplasmic AGP capacitor

Rapid ultrafiltration of a model AGP shows pH-dependent AGP–Ca2+ binding/unbinding and, as expected, a much lower Mg2+ content (Fig. 2c). During tip growth, stretch-activated H+-ATPases of the plasma membrane (Pertl et al., 2010) generate pulses of H+ (Monshausen et al., 2007). The H+ flux measured variably as 0.4–490 pmol cm−2 s−1 (Holdaway-Clarke & Hepler, 2003) would correspond to a pH of c. 0–3 in the small periplasmic volume of the growing region. Thus the periplasmic pH there is significantly lower than that of the wall, typically at pH ˜ 5. The reported high-Cl efflux at the pollen tube tip (Zonia et al., 2002) may be necessary to maintain the low periplasmic pH. Release of bound Ca2+ from strategically located periplasmic AGPs is therefore the most likely immediate source of cytosolic Ca2+ rather than the highly methylesterified pectin of the more distant cell wall (Fig. 2d). Presumably, the opening of stretch-activated Ca2+ channels (Pickard & Fujiki, 2005) synchronized with stretch-activated H+-ATPase enables the entry of periplasmic Ca2+ into the cytosol. The transiently low periplasmic pH releases Ca2+; subsequently the pH is restored by rapid H+ diffusion to the larger buffered wall domain.

Ca2+ binding by AGPs rationalizes their glucuronic (GlcA) rather than their galacturonic (GalA) acid content. First, GlcA carboxyls are free to ionize unlike the highly methylesterified (Albersheim et al., 2011) GalA of pectin. Secondly, inversion of the C4 hydroxyl results in a significant difference between the reported pKa values of the two uronic acids: 2.8 for GlcA (Wang et al., 1991) and 3.6 for GalA (Scanlan et al., 2010). Therefore, AGPs have a higher affinity for Ca2+ than pectin (Fig. 2d). The molar ratio of uronic acid : Ca2+ (Table 1) and the Ca2+ binding as a function of pH also show that AGPs bind Ca2+ more efficiently than pectin. Thus a discharged AGP–Ca2+ capacitor would be recharged by Ca2+ recycled from the cytosol (Fig. 3a) and possibly from the wall matrix. Thirdly, intramolecular Ca2+ binding ensures that AGPs remain as a sol in contrast to demethylesterified pectic gels crosslinked by intermolecular Ca2+ bridges.

Figure 3.

The calcium oscillator. (a) Ca2+ release from periplasmic arabinogalactan glycoproteins (AGPs). The oscillator generates pulses of Ca2+ (green dots), whose influx coordinates exocytosis and rapid tip growth. This involves a pulse of H+ (black dots) releasing Ca2+ from periplasmic AGPs (red beads) via stretch-activated Ca2+ channels into the cytosol, thence sequestration by exocytotic Golgi vesicles containing AGPs (red dots). Diffusion of the initial H+ pulse to the wall domain restores the periplasmic pH. Ca2+ is recycled by fusion of vesicles with the plasma membrane. Ca2+ bound to periplasmic AGPs is now ready for the next oscillation. W, wall; P, periplasm; PM, plasma membrane; G, Golgi; GV, Golgi exocytotic vesicles. (b) Ca2+ current as a molecular clock: a series RLC circuit with membrane-bound AGPs as the capacitor, C; Ca2+ sequestration as the inductance, L; vesicle exocytosis as the resistor, R, which limits the recycling rate. Hypothetically, C and L largely determine the oscillator frequency and amplitude of the Ca2+ current, ICa.

A calcium oscillator triggers rapid growth responses

Plant growth oscillates (Darwin, 1880). Such oscillations, particularly apparent in pollen tubes (Holdaway-Clarke & Hepler, 2003) and root hairs (Monshausen et al., 2007), are also accompanied by Ca2+ influx, which appears as tip-focused (inverse cone) cytosolic Ca2+ waves with a periodicity of c. two cycles min–1. This brief Ca2+ signal activates numerous enzymes, thus coordinating Golgi secretion, vesicle transport and exocytosis to the growing tip (Battey et al., 1999; Roy et al., 1999). The cytosolic Ca2+ is then rapidly sequestered, recycled and exported (Herrmann & Felle, 1995), most likely by the same (c. 60 nm) post-Golgi vesicles that transport AGPs to the cell surface (Fig. 3a), possibly enhanced by Golgi Ca2+-ATPases (Ordenes et al., 2002; Bonza & De Michelis, 2011). This could solve the problem of low Ca2+ diffusivity in the cytosol (Trewavas, 2000). Thus Ca2+ release and recycling viewed as a ‘calcium oscillator’ quite possibly shape the Ca2+ signal (Hetherington & Brownlee, 2004) and can be modelled as a simple electronic circuit diagram showing the Ca2+ current, ICa (Fig. 3b). In this model, H+ secretion releases Ca2+ from an extracellular store and thus differs significantly from the animal model of ‘capacitative calcium entry’ (Berridge, 1997), that is, Ca2+-induced Ca2+-release (CICR), where a small influx of extracellular Ca2+ induces a large release from an intracellular store.

We surmise that oscillations in tip growth rate depend on the rheological properties of the wall itself (Chebli & Geitmann, 2012): accretion of new material increases wall thickness, which, in turn, slows the rate of turgor-driven tip extrusion (McKenna et al., 2009). Subsequently, as the wall thins, the rate of tip extension increases again. This reactivates stretch-sensitive H+-ATPase (Pertl et al., 2010), which lowers pH and dissociates Ca2+ from AGP at the outer surface of the plasma membrane with concomitant influx of the released Ca2+ through specific Ca2+ channels (Pickard & Fujiki, 2005) into the cytosol. Other factors, notably borate, pectin methylesterase (PME) and reactive oxygen species (ROS), may fine-tune this simple scenario. A borate diester crosslinks the pectin polysaccharide RG-II (O'Neill et al., 2004). PME facilitates Ca2+ crosslinking of de-esterified pectin in the pollen tube subapical region (Bosch et al., 2005) with concomitant wall hardening. ROS, generated by plasma membrane NADPH oxidase, may also harden the subapical wall by peroxidative crosslink formation (Held et al., 2004). ROS may also regulate Ca2+ channels by their direct activation or indirectly by the regulation of extracellular pH (Monshausen et al., 2007).

Arabinogalactan glycoproteins are generally considered as ‘signalling molecules’, but surprisingly they have never been implicated in Ca2+ signalling. Although ‘these mysterious molecules have yet to yield up all their secrets’ (Albersheim et al., 2011), the global calcium oscillator explains how periplasmic AGPs can be involved in so many biological processes (Seifert & Roberts, 2007) – not as signalling molecules per se but as a Ca2+ source directly involved in Ca2+-regulated tip growth (Herrmann & Felle, 1995). However, AGPs are ubiquitous and therefore essential to numerous other processes that depend on Ca2+ signalling, notably ‘the power of movement in plants’ described by Darwin (1880). A global role for Ca2+ signalling is more recent (Kudla et al., 2010). It began with enzyme secretion, but now connects many disparate processes (Dodd et al., 2010), including stomatal closure, legume nodulation, defence and response to elicitors, thigmotropism, gravitropism, phototropism, fertilization including self-incompatibility, circadian rhythms and osmoregulation by aquaporins. Indeed, one can reinterpret the ‘acid growth’ hypothesis (Rayle & Cleland, 1992) of cell extension as auxin-activated H+-ATPase (Serrano, 1989) with concomitant influx of Ca2+ (Scherer, 2011) from periplasmic AGPs, rather than the decrease of cell wall pH.

Based on structural analyses of Hyp-AGs ranging from Leguminosae to Solanaceae, classical AGPs are adapted for their role as a periplasmic Ca2+ capacitor by the abundance of their Ca2+-binding subunit. However, subtle sidechain modifications of this canonical subunit include 4-O-methyl glucuronic acid, the number of arabinose residues and the presence or absence of rhamnose, as well as size and spacing of Hyp-AGs along the extended polypeptide. This suggests fine-tuning of ion-binding affinity tailored to individual tissues and environmental factors. Indeed, under saline conditions, Na+ competes with membrane-bound Ca2+, thereby decreasing growth (Cramer et al., 1985), an effect mitigated by increasing Ca2+ concentrations (LaHaye & Epstein, 1969).

Finally, the occurrence of Ca2+ oscillations, variously described (Dodd et al., 2010) as influx, sparks, spikes or waves, in many different cell types suggests a general role for the calcium oscillator as a molecular clock that coordinates intracellular events, an hypothesis with wide ramifications. AGPs may be peripheral molecules but along with their close relatives, the extensins (Lamport et al., 2011), they play a central role in the regulation of plant growth and development (Velasquez et al., 2011; Wolf et al., 2012).


We thank Professors Marcia Kieliszewski (Ohio University, USA) and Barbara Pickard (Washington University in St Louis, USA) for helpful discussions, Marion Nemitz for exploratory Ca2+-binding experiments, and Chris Dadswell (University of Sussex, UK) for technical assistance with mass spectrometry.