Calmodulin controls spatial and temporal specificity of calcium-induced calcium release Authors: Joanna Jędrzejewska-Szmek, Kim T. Blackwell, Ausra Saudargiene, Andrea E. Martin, Ausra Saudargiene, Andrea E. Martin, Ausra Saudargiene, Andrea E. Martin, Ausra Saudargiene, Andrea E. Martin Published: 2026 ABSTRACT Calcium dynamics controls learning and memory, and abnormal calcium dynamics have been implicated in neurodegenerative disorders, such as Alzheimer’s disease (AD). Calcium dynamics are influenced by calcium-induced calcium release (CICR), which is mediated by ryanodine receptors (RyR) located on endoplasmic reticulum (ER) membrane. Calmodulin, one of the most abundant proteins in the brain, inhibits RyR2, expressed in the dendrites of hippocampal CA1 neurons, with several reported consequences: relief of this inhibition is responsible for heart failure, and enhancing calmodulin to RyR binding alleviates cell loss and AD-like neuronal hyperexcitability. To investigate the role of calmodulin in aging and AD, we built a sophisticated reaction-diffusion model of a dendritic branch with ER. We showed that relieving calmodulin inhibition of RyR2 increased spatial and temporal spread of calcium transients in the dendrite. This effect was also visible in a model of old age, where disinhibition of half of the RyR2 population increased spatial spread of calcium transients by a factor of 2, and disinhibition of RyR2 combined with increased concentration of calcium buffering molecules increased duration of calcium transients. Lower activation of plasma membrane calcium ATPase (PMCA), which is also activated by calmodulin and inhibited by β-Amyloid oligomers, and not RyR2 disinhibition, led to an increase in resting intracellular calcium concentration as observed in AD. Overall, our research demonstrates that changes in calmodulin that are associated with AD and aging, by regulation of RyR2 (in old age) and PMCA (in AD), underlie changes in calcium dynamics that might have consequences for learning and memory. Introduction In many neurodegenerative disorders such as Alzheimer’s disease (AD), neural information processing is deficient, interfering with a patient’s daily life [1]. Neuronal information processing begins in the dendrites [2], which continuously receive information from other neurons in the form of synaptic activation. Dendritic calcium dynamics, one of the modes of information processing [3] in the nervous system, has been extensively studied, with a focus on activation of N-methyl-D-aspartate (NMDA) receptors [4] and voltage-gated calcium permeable channels (VGCC) [2]. However, relatively little is known about the role of calcium stores such as the endoplasmic reticulum (ER) in information processing in the dendrite [5]. It is crucial to investigate the role of the ER in calcium dynamics since its calcium-release channels have been implicated in development and alleviation of Alzheimer’s disease in murine models [6–8]. The ER is a large intracellular organelle with continuous membranes and can be viewed as a cable-within-the neuron [9]. Calcium is released from the ER through inositol triphosphate (IP3) receptors (IP3R) in response to IP3 and ryanodine receptors (RyR) in response to calcium elevation [5]. The latter process is known as calcium-induced calcium release (CICR) and serves as an amplifier of cytosolic calcium signals. There are 3 RyR isoforms, which are expressed in the brain [10], with RyR2 expressed in the dendritic shafts [11] and RyR3 expressed in the spines [12] of hippocampal CA1 pyramidal neurons. RyRs are ion channels that are modulated by many proteins such as calmodulin (CaM), calcium/calmodulin activated protein kinase II and protein kinase A. Calmodulin in the brain is mostly viewed as a fast acting calcium buffering molecule [13]. However, it is a vital part of the calcium signal transduction pathway, not only activating the calcium/calmodulin activated protein kinases, but also activating RyR1 and RyR3 isoforms and inhibiting RyR2. Moreover, due to its abundance, calmodulin might be constitutively bound to RyRs in the brain [14,15]. Yet, with oxidative stress and aging the affinity of calmodulin to RyR is lowered [16], possibly relieving the inhibition of RyR2 by calmodulin. Disinhibited RyR2s might contribute to aberrant calcium homeostasis [5] that is responsible for disruption of neuronal function in AD [7,17] and deficits in learning and memory accompanying old age. Indeed, enhancing calmodulin binding to RyR2 in mouse models of AD reversed cognitive dysfunction [6]. The role of CICR in information transfer along the dendrite is unclear. Information transfer from the synapse to the nucleus and back might involve electrotonic propagation involving the ER along the dendritic tree [9], but experiments performed in cultured hippocampal neurons show that ER membrane potentials do not spread beyond the site of activation [18]. Another possibility of biochemical signaling to the nucleus is induction of an intracellular calcium wave propagating down the dendritic tree. Excitation-translation coupling in neuronal cultures depends on activation of L-type VGCCs [19,20] and also recruits RyR2s [19,21]. CICR contributes to the propagation of dendritic transients [22] and RyR2 activation contributes to nuclear calcium release [21]. It has recently been suggested that RyR2s might support propagation of intracellular calcium waves evoked by back-propagating action potentials (bAP) in apical dendrites [23]. Yet, experimental evidence indicates that intracellular calcium waves in dendrites of pyramidal neurons in-vitro cannot be evoked by bAPs [24], and require activation of metabotropic glutamate receptors, which in turn leads to IP3 production and IP3-induced calcium release [25–27], which was investigated computationally in [28,29]. To test whether RyR2s alone can support calcium wave propagation, as suggested by [23], we built spatial stochastic reaction-diffusion models of hippocampal CA1 pyramidal thin, medium and thick dendrites. We show in-silico that RyR2s cannot support intracellular calcium wave propagation in a model of a healthy neuron (control conditions) nor when calmodulin inhibition of RyR2s is relieved. Furthermore, we show that disinhibition of RyR2 increases spatial spread of calcium transients (lowers spatial specificity) and increases duration of calcium transients (decreases temporal specificity) in the dendrite. Using a model of CA1 pyramidal dendrites in old age we show that spatial and temporal specificity of calcium elevations is diminished in old age due to oxidization of calmodulin and RyR2s, possibly contributing to neuronal hyperexcitability observed in aging and AD [30,31]. Finally, we show that calmodulin oxidation can be responsible for increased basal intracellular calcium concentration by lowering plasma membrane Ca2+ ATPase (PMCA) activity but not by RyR2-disinhibition. Materials and methods Computational methods To investigate the extent of calcium elevation in the dendritic branch following synaptic stimulation we developed a multi-compartment stochastic reaction-diffusion model of signaling pathways underlying calcium-induced calcium release. We adapted the calcium regulation pathways implemented in [32], and added a model of the ER containing: luminal calcium (CaER) and calreticulin (CRT) with sarco/endoplasmic reticulum calcium ATPase (SERCA) and a leak channel (LeakER). The ER contained ryanodine receptors, as explained below, and store operated calcium entry (SOCE) [33], with parameters fitted to measurements from [34]. The ER was treated as a separate compartment that is uniformly distributed in the dendrite, occupying 25% of its volume. Tortuosity of the ER was taken into account by reducing the diffusion rate of calcium molecules in the ER, compared to diffusion rate of the cytosolic calcium molecules. The signaling pathways model diagram is shown in Fig 1. Molecular pathways implemented in the model. Calcium flows into the cell through the calcium permeable channels in the cell membrane (denoted VGCC in the diagram and Leak in Table 1). Calcium is extruded into the extracellular space by the plasma membrane Ca2+-ATPase (PMCA) and sodium calcium exchanger (NCX) and into the endoplasmic reticulum by the Sarco/endoplasmic Reticulum Calcium ATPase (SERCA). In the cytosol calcium is buffered by calbindin, indiffusible buffer (Buffer) and calmodulin (CaM). Apo-calmodulin (unbound to calcium) is buffered by neurogranin (Ng). Calcium flows from the ER to the cytosol through open ryanodine type 2 receptors (RyR2), which are activated by calcium and inhibited by calmodulin. In the ER calcium is buffered by calreticulin (CRT) and binds stromal interaction molecules (STIM), which serve as luminal calcium sensors. When calcium in the endoplasmic reticulum is too low, STIM molecules form dimers and bind calcium release-activated calcium channel protein 1 (Orai1) opening them and allowing calcium from the extracellular space to flow into the cytosol, where it would be transferred by SERCA to the endoplasmic reticulum, thereby replenishing it. The ORA1-STIM channels constitute store-operated calcium entry (SOCE). In the model ER molecules were distributed randomly according to a mean concentration. SERCA and STIM2 concentrations were scaled by area/volume ratio of the dendrite. RyR2 copy numbers were kept the same for all dendritic diameter models, because experimental evidence suggests that their concentration depends linearly on the total length of the dendrite [35]. ER luminal species (CaER, CRT and Ca-bound calreticulin (CRT·CaER)) were proportional to ER volume with concentration set to achieve ER molecule copy numbers expected from published measurements of concentration [36]. In our simulations we used two models of ryanodine receptors, which were based on the Keizer-Smith model [37]: RyR2CaM – a model of ryanodine receptor constitutively bound to calmodulin (inhibited by calmodulin) and activated by calcium, fitted to data presented in Fig 2 of [38], RyR2CaM – a model of ryanodine receptor constitutively bound to calmodulin (inhibited by calmodulin) and activated by calcium, fitted to data presented in Fig 2 of [38], RyR2 no CaM – a model of ryanodine receptor not bound to calmodulin (disinhibited) as occurs when calmodulin is oxidized, which was fitted to data presented in Fig 2 of [39]. RyR2 no CaM – a model of ryanodine receptor not bound to calmodulin (disinhibited) as occurs when calmodulin is oxidized, which was fitted to data presented in Fig 2 of [39]. Morphology We implemented the signaling pathways model (Fig 1) in 51μm long dendrites with diameters of 1.2μm, 2.4μm and 6.0μm representing thin, medium and thick dendrites, respectively (Fig 2). Each morphology was divided into voxels. Submembrane voxels, which allow for placing of membrane-associated molecules, were cubes of 0.5μm×0.4μm×0.6μm (the same for all dendritic diameters). Cytosolic voxels were the same as submembrane voxels for the 1.2μm model, cubes of 0.5μm×0.46μm×0.6μm for the 2.4μm model and cubes of 0.5μm×0.4615μm×0.6μm for the 6.0μm model. Dendritic morphologies used in simulations. We used three dendritic morphologies: thin dendrite (diameter of 1.2 µm), medium dendrite (diameter of 2.4μm) and thick dendrite (diameter of 6.0μm). All dendrites were 51 µm long. Each morphology was subdivided into voxels, allowing two-dimensional diffusion. Submembrane voxels were cubes of 0.5μm×0.4μm×0.6μm (the same for all dendritic diameters). The submembrane voxels allow for placement of membrane associated molecules. Cytosolic voxels were the same as submembrane voxels for the 1.2μm model, cubes of 0.5μm×0.46μm×0.6μm for the 2.4μm model and cubes of 0.5μm×0.4615μm×0.6μm for the 6.0μm model. The dotted lines show some of the lengthwise and widthwise subdivisions into voxels. For simulating synaptic activation, calcium molecules were injected into submembrane voxels in the middle of the dendritic morphology (denoted synaptic stim on the diagram). For simulating b-AP like stimulus, calcium molecules were injected into the cytosol to an end dendritic voxel (point stimulus, denoted bAP stim on the diagram). Control model The control model comprised a model of cytosolic calcium dynamics (Table 1) with total concentrations shown in Table 2 and diffusible species in Table 3. This model was fitted with the model of the ER containing calcium extrusion and buffering (Table 4), opening and calcium release by the RyR2 inhibited by calmodulin (upper half of Table 5) and SOCE (Table 6) with total concentrations specified in Table 7. 10.1371/journal.pcbi.1013752.t001Table 1Calcium extrusion to the extracellular space and calcium buffering in the cytosol: reaction rates.Reaction k𝐟 (nM−1ms−1) k𝐫 (ms−1) kcat (ms−1) source PMCA+Ca2+⇌PMCA·Ca→PMCA+Caext2+ 0.5e-40.0070.0035[40] NCX+Ca2+⇌NCX·Ca→NCX+Caext2+ 1.68e-50.01120.0056[41,42] Leak+Caext2+⇌Leak·Caext→Leak+Ca2+ 1.5e-61.1e-31.1e-3adjusted to maintain Cai2+ concentration of 75 nM [32], represents VGCCs Calbindin+Ca2+⇌CalbindinCa 2.8e-50.0196[43] IndiffusibleBuffer+Ca2+⇌IndiffusibleBufferCa 0.004e-320.0e-2[44] 2Ca2++CaM⇌CaMCa2 6e-69.1e-3[45] 2Ca2++CaM⇌CaMCa2 0.1e-31000e-3[46] CaM+Ng⇌NgCaM 2.80E-53.60E-2[47] Calcium extrusion to the extracellular space and calcium buffering in the cytosol: reaction rates. 10.1371/journal.pcbi.1013752.t002Table 2Concentrations of molecular species governing cytosolic calcium dynamics (Table 1): calcium extrusion (PMCA and NCX) and calcium buffering (calmodulin, calbindin, indiffusible calcium buffer) molecules. Concentration units of cytosolic species ((Ca2+), Caext2+, calbindin, calmodulin, neurogranin, indiffusible calcium buffer, ER calcium and calreticulin) are nmolar, whereas densities of membrane species (Leak, PMCA, NCX) are expressed in pmol µm–2.Speciecontroldisinhibited RyR2old ageADsourcecalcium (Ca2+)757575[48,49]extracellular calcium (Caext2+) 2×106 2×106 2×106 2×106 chosen to match in-vitro experimentsLeak6 0006 0005 0006 000chosen to counterbalance PMCAPMCA5 6005 6005 6005 600[50]NCX3 0003 0003 0003 000adjusted for Vmax of NCX to match [51]Calbindin 150×103 150×103 300×103 150×103 to match 160 µM high affinity binding sites [52]Calmodulin 25.8×103 25.8×103 23.0×103 25.8×103 [14]Neurogranin (Ng) 20×103 20×103 40×103 20×103 [53]Indiffusible calcium buffer 2×106 2×106 4×106 2×106 [44]ER calcium (CaER2+) 320×103 320320320[54]calreticulin (CRT) 400×103 400×103 400×103 400×103 [36] Concentrations of molecular species governing cytosolic calcium dynamics (Table 1): calcium extrusion (PMCA and NCX) and calcium buffering (calmodulin, calbindin, indiffusible calcium buffer) molecules. Concentration units of cytosolic species ((Ca2+), Caext2+, calbindin, calmodulin, neurogranin, indiffusible calcium buffer, ER calcium and calreticulin) are nmolar, whereas densities of membrane species (Leak, PMCA, NCX) are expressed in pmol µm–2. 10.1371/journal.pcbi.1013752.t003Table 3Diffusion rates of diffusible species. Abbreviations same as in Tables 1,4,6,5.Namekdiff (µm2 s–1)sourceCa2+100[55] ext.Ca2+ 10[32] ERCa2+ 10Lower than cytosolic diffusion to account for complex structure of ER tubulesCaM4[56]Calbindin8.3[57] Diffusion rates of diffusible species. Abbreviations same as in Tables 1,4,6,5. 10.1371/journal.pcbi.1013752.t004Table 4Calcium extrusion to the ER and calcium buffering in the ER: reaction rates.Reaction k𝐟 (nM−1ms−1) k𝐫 (ms−1) kcat (ms−1) sourceSERCA+Ca2+⇌ SERCACa17147e-78426.3e-4[58] SERCACa+Ca2+⇌SERCACa2→SERCA+CaER2+ 17147e-78426.3e-450e-3[58,59] CaER2++LeakER⇌Ca2++LeakER 1.8e-71.8e-7adjusted to fit intracellular Cai2+ in Fig 1. panel A of [34] CRT+CaER2+⇌CRT·CaER 1e-50.1[60] Calcium extrusion to the ER and calcium buffering in the ER: reaction rates. 10.1371/journal.pcbi.1013752.t005Table 5Model of calmodulin-bound RyR2 (upper half) and disinhibited RyR2 (lower half). Both models were based on [37], with calcium binding constants fitted to data in Fig 2 [38] (+1 µM CaM) for the RyR2CaM model, and Fig 2 of [39] for the RyR2 model.Reaction k𝐟 (nM−4ms−1) k𝐫 (ms−1) source RyR2CaM+4Ca2+⇌RyR2CaMC1 10e-111[38] RyR2CaMC1+4Ca2+⇌RyR2CaMO1 5e-104.8[38] RyR2CaMO1+4Ca2+⇌RyR2CaMO2 5e-1213[38] RyR2CaMC2+4Ca2+⇌RyR2CaMO2 3.33333e-1666.667e-2[38] RyR2CaMC1+4Ca2+⇌RyR2CaMC2 1e-91.248e-2[38] RyR2CaMC2+4Ca2+⇌RyR2CaMC3 1.5e-153[38] RyR2CaMO1+CaER2+⇌RyR2CaMO1+Ca2+ 5e-35e-3[61] RyR2CaMO2+CaER2+⇌RyR2CaMO2+Ca2+ 5e-35e-3[61] RyR2+4Ca2+⇌RyR2C1 5e-115e-3[39] RyR2C1+4Ca2+⇌RyR2O1 5e-109.6[39] RyR2O1+4Ca2+⇌RyR2O2 5e-1113[39] RyR2C2+4Ca2+⇌RyR2O2 3.33333e-1566.667e-3[39] RyR2C1+4Ca2+⇌RyR2C2 5000e-151.235e-3[39] RyR2C2+4Ca2+⇌RyR2C3 1e-143[39] RyR2O1+CaER2+⇌RyR2O1+Ca2+ 5e-35e-3[61] RyR2O2+CaER2+⇌RyR2O2+Ca2+ 5e-35e-3[61] Model of calmodulin-bound RyR2 (upper half) and disinhibited RyR2 (lower half). Both models were based on [37], with calcium binding constants fitted to data in Fig 2 [38] (+1 µM CaM) for the RyR2CaM model, and Fig 2 of [39] for the RyR2 model. 10.1371/journal.pcbi.1013752.t006Table 6Store operated calcium entry model: reaction rates. ER calcium to STIM dimers binding rates were adjusted to fit intracellular Ca𝐢2+ in Fig 1. panel A of [34]. Calcium flux through the open Orai1 channels was adjusted to fit intracellular Ca𝐢2+ in Fig 1. panel A of [34] and the graded opening of SOCE by activation of different numbers of Orai1 subunits [62].Reaction k𝐟 (nMM−1ms−1) k𝐫 (ms−1) source STIM2+CaER2+⇌STIM2CaER 8e-72e-3adjusted to fit intracellular Cai2+ in Fig 1. panel A of [34] STIM2+STIM2⇌STIM2dimer 4.8e-711e-5[33] PMJ+STIM4⇌PMJ+PMJSTIM4 1.8e-043e-4[33] PMJSTIM4+Orai1⇌Orai1STIM4 1.5e-048e-3[33] Orai1STIM4+Orai1STIM4⇌Orai12STIM4 3.75e-53.75e-5[33,34] Orai2STIM4+Orai1STIM4⇌Orai3STIM4 9.375e-065e-4[33,34] Orai3STIM4+Orai1STIM4⇌2Orai2STIM4 9.375e-06[33,34] Caext+Orai1STIM4⇌Orai1STIM4+Ca2+ 1.5e-111.5e-11[34,62] Caext+Orai2STIM4⇌Orai2STIM4+Ca2+ 5e-85e-8[34,62] Caext+Orai3STIM4⇌Orai3STIM4+Ca2+ 5e-85e-8[34,62] Store operated calcium entry model: reaction rates. ER calcium to STIM dimers binding rates were adjusted to fit intracellular Ca𝐢2+ in Fig 1. panel A of [34]. Calcium flux through the open Orai1 channels was adjusted to fit intracellular Ca𝐢2+ in Fig 1. panel A of [34] and the graded opening of SOCE by activation of different numbers of Orai1 subunits [62]. 10.1371/journal.pcbi.1013752.t007Table 7Total concentration of molecular species localized in the ER membrane. We assumed that the ER comprises 25% of the cytosol with ryanodine receptors (RyR2 and RyR2CaM), SERCA, LeakER, and STIM dimers uniformly distributed in the cytosol with concentration units of nmolar, and Orai1 and plasma membrane-endoplasmic reticulum junctions (PMJ) located in dendritic membrane with units of pmol µm–2. *adjusted to match RyR2 density from [23]. **adjusted to fit intracellular Ca𝐢2+ in Fig 1. panel A of [34]. ***adjusted to match RyR2 copy numbers for the 1.2 model. ****scaled by area-to-volume ratio of the ER.Dendrite modelSpeciecontroldisinhibited RyR2old ageADsource 1.2μm RyR2CaM8088[23]* 1.2μm RyR20888[23]* 1.2μm LeakER 380380265380adjusted to counterbalance SERCA 1.2μm SERCA 2.8×103 2.8×103 2.8×103 2.8×103 [36,63] 1.2μm Orai12002000200[34]** 1.2μm PMJ6606[34]** 1.2μm STIM dimers STIM24000400400[34]** 2.4μm RyR2CaM6066*** 2.4μm RyR20666*** 2.4μm LeakER 200200175200adjusted to counterbalance SERCA 2.4μm SERCA 1.5×103 1.5×103 1.5×103 1.5×103 [36,63] 2.4μm Orai12002000200[34]** 2.4μm PMJ6606[34]** 2.4μm STIM dimers STIM22002000200[34]** and **** 6.0μm RyR2CaM2022*** 6.0μm RyR20222*** 6.0μm LeakER 76766676adjusted to counterbalance SERCA 6.0μm SERCA600600600600[36,63] 6.0μm Orai12002000200[34]** 6.0μm PMJ6606[34]** 6.0μm STIM dimers STIM28080080[34]** and **** Total concentration of molecular species localized in the ER membrane. We assumed that the ER comprises 25% of the cytosol with ryanodine receptors (RyR2 and RyR2CaM), SERCA, LeakER, and STIM dimers uniformly distributed in the cytosol with concentration units of nmolar, and Orai1 and plasma membrane-endoplasmic reticulum junctions (PMJ) located in dendritic membrane with units of pmol µm–2. *adjusted to match RyR2 density from [23]. **adjusted to fit intracellular Ca𝐢2+ in Fig 1. panel A of [34]. ***adjusted to match RyR2 copy numbers for the 1.2 model. ****scaled by area-to-volume ratio of the ER. Old age model Aging is related to changes in calcium homeostasis [64]. Calmodulin becomes oxidized [65], which reduces its binding to RyR2 [16]. Specifically, in the oxidized conditions with cytosolic calcium levels calmodulin affinity for RyR2 drops twenty-fold [66], resulting in 50% RyR2s bound to calmodulin and hence inhibited by calmodulin. PMCA activity is diminished [67–69] and in in-vitro aging cultures, store-operated calcium entry is down-regulated [70]. The buffering capacity of intracellular buffers of CA1 pyramidal neurons is increased [71]. Also the RyR2 expression is increased [72]. The T-type calcium channels are down-regulated in the hippocampus of aging mice and men [73], whereas the L-type channels are up-regulated [74]. It has also been reported that the resting intracellular calcium concentration in CA1 pyramidal models from aged rats is unchanged [48]. To take all these changes into account in the old age model we modified the control model in several ways. We lowered PMCA kcat by 20% (Table 8). We lowered PMCA kcat by 20% (Table 8). To emulate changes in calmodulin binding to RyR2 due to oxidation, we divided the population into halves: one bound to and, in consequence, inhibited by calmodulin (upper half of Table 5) and one not inhibited by calmodulin (lower half of Table 5), as indicated by Table 7. This resulted in the total RyR2 population favoring activation by calcium consistent with experimental data [75]. To emulate changes in calmodulin binding to RyR2 due to oxidation, we divided the population into halves: one bound to and, in consequence, inhibited by calmodulin (upper half of Table 5) and one not inhibited by calmodulin (lower half of Table 5), as indicated by Table 7. This resulted in the total RyR2 population favoring activation by calcium consistent with experimental data [75]. To account for increased calcium buffering in old age we doubled calbindin and indiffusible buffer concentrations (Table 2). To account for increased calcium buffering in old age we doubled calbindin and indiffusible buffer concentrations (Table 2). To account for an upregulation of L-type VGCCs we increased calcium inputs (a comparison of calcium inputs is shown Table 9). To account for an upregulation of L-type VGCCs we increased calcium inputs (a comparison of calcium inputs is shown Table 9). 10.1371/journal.pcbi.1013752.t008Table 8Calcium extrusion to the extracellular space reaction rates in the old age model.Reaction k𝐟 (nM−1ms−1) k𝐫 (ms−1) kcat (ms−1) sourcePMCA+Ca2+⇌ PMCA·Ca→PMCA+Caext0.5e-40.0070.0028[67–69] Calcium extrusion to the extracellular space reaction rates in the old age model. 10.1371/journal.pcbi.1013752.t009Table 9Calcium inputs for Figs 3, 5–6. Stimulation was applied at 3000ms to the submembrane voxels in the middle of the morphology.diamduration (ms)input control (particlesms−1)maxCa𝐢2+ in stimulated dendrite (µM)input old age (particlesms−1)maxCa𝐢2+ in stimulated dendrite µM) 1.2μm 406000.512000.6 1.2μm 4012001.014001.3 1.2μm 4024002.348003.3 1.2μm 4040004.980007.0 1.2μm 4048006.596009.0 2.4μm 4010000.420000.6 2.4μm 4020001.040001.3 2.4μm 4040002.380003.3 2.4μm 4060004.5120006.0 2.4μm 4080007.0240009.6 6.0μm 4020000.440000.5 6.0μm 4040001.080001.2 6.0μm 4080002.6160003.6 6.0μm 40120005.0240007.2 6.0μm 40160007.83200012.3 Calcium inputs for Figs 3, 5–6. Stimulation was applied at 3000ms to the submembrane voxels in the middle of the morphology. AD model Alzheimer’s disease is characterized by calcium dyshomeostasis [76] and elevated resting intracellular calcium concentration [77,78]. Animal models of AD exhibit increased RyR expression [79,80], high levels of oxidative stress [81], and reduced calbindin levels compared to old age [82]. High levels of oxidative stress likely result in high calmodulin oxidation resulting in lower PMCA activity and lower RyR2 inhibition. To investigate how PMCA and RyR2 influence resting intracellular calcium dynamics in early onset AD, we built an AD model of the dendrite (Table 2 and Table 7). This model exhibited lower PMCA activity (Table 8), increased RyR2 concentration (Table 7) with half the RyR2 population inhibited by calmodulin (upper half of Table 5) and half disinhibited (lower half of Table 5). In contrast to the old-age model, it did not have altered calcium buffering and it preserved calcium leak values of the control model. Stimulation protocols All model variants were simulated without any stimulation to reach steady state (usually for 1000s). Initial conditions of all the model species were recalculated based on the last 25s of those simulations and the models were rerun without any stimulation to verify the stability of the steady state. If the steady state was not reached the procedure was repeated. For simulations of calcium wave induction only models in steady state were used. We simulated two types of calcium influx to the dendrite: (Fig 2) synaptic-like activation as a 40 ms influx of calcium molecules into submembrane voxels in the middle of the dendrite (summary of the stimulation is presented in Table 9), synaptic-like activation as a 40 ms influx of calcium molecules into submembrane voxels in the middle of the dendrite (summary of the stimulation is presented in Table 9), bAP-like stimulus as a 3 ms influx of calcium molecules into the cytosol voxels at the end of the dendrite similarly as in [23] (summary of the stimulation is presented in Table 10). bAP-like stimulus as a 3 ms influx of calcium molecules into the cytosol voxels at the end of the dendrite similarly as in [23] (summary of the stimulation is presented in Table 10). 10.1371/journal.pcbi.1013752.t010Table 10Calcium inputs for Fig 4. Stimulation was applied at 5000ms. Stimulation was applied to the cytosol of the first dendritic segment.diamduration (ms)input (particlesms−1)maxCa𝐢2+ in stimulated dendrite µM) 1.2μm 360003.4 1.2μm 3120007.7 1.2μm 32400019.2 1.2μm 34000041.0 1.2μm 34800054.9 2.4μm 3100003.4 2.4μm 3200008.0 2.4μm 34000021.1 2.4μm 36000041.2 2.4μm 38000069.6 6.0μm 3200003.1 6.0μm 3400008.2 6.0μm 38000027.0 6.0μm 312000056.7 6.0μm 316000089.9 Calcium inputs for Fig 4. Stimulation was applied at 5000ms. Stimulation was applied to the cytosol of the first dendritic segment. We used 5 different calcium input values to thoroughly explore CICR in the dendrite. It is estimated that calcium concentration in the dendritic spine due to an excitatory postsynaptic potential (EPSP) can reach up to 12 µM and 1 µM due to a bAP [83]. Calcium levels in the thin apical dendrites and dendritic spines are highly correlated [84], hence we chose calcium inputs resulting in an elevation of calcium concentration in the stimulated dendrite up to 10 µM (Table 9). For simulation of the bAP we selected injections of similar numbers of calcium molecules as with the EPSP though in a much smaller time window, to fully explore CICR properties (Table 10). Calcium injections (Table 9 and 10) were calibrated to obtain similar calcium elevations in the dendrite for the 2.4μm and 6.0μm models. Models were run for 3s before synaptic-like activation (Table 9) and for 5s before the b-AP like stimulus (Table 10). In total models were run for 10s. We established that 10s simulation duration was sufficient to observe changes in calcium dynamics by running simulations for 20s first. For each condition (calcium injection, model variant) we ran 10 simulations with different seeds. Simulations of the impact of RyR2 variants and PMCA variants on resting intracellular calcium concentration were performed in the absence of any stimulation for 1000s. For each model we ran 4 simulations with different seeds. Simulation The model was implemented using the stochastic reaction–diffusion simulator NeuroRD [85] version 3.2.3, using the adaptive (asynchronous τ-leaping) numerical method with tolerance of 0.01. All data and code used for running experiments and plotting is available on a GitHub repository at https://github.com/asiaszmek/stochastic_ER. Analysis We calculated the mean intracellular calcium concentration every 0.5μm in the longitudinal direction (the voxel width in the longitudinal direction was 0.5μm) by averaging across the width. Spatial spread was calculated as the furthest location from the stimulated site, in which calcium amplitude following stimulation exceeds 250% of resting calcium concentration of baseline (prior to the stimulation). For synaptic activation-like stimulation, in which the middle of the dendrite was stimulated, we average the spatial spread in the left and right halves of the dendrite. Temporal decay constant was calculated by fitting an exponential function a·exp(−tb)+c to the 1sec long average calcium concentration with initial time (and a) corresponding to the time (and value) of maximum calcium concentration in the stimulated site. Time decay constant value was the b parameter obtained in the fit. Figs 3, 4, 5 and 6 show mean values of the spatial spread and temporal decay constant with standard error as errors bars obtained in 10 simulations with different seeds. The ctrl model does not support intracellular calcium wave propagation in a CA1 pyramidal neuron dendrite model. Spatial extent (a) and temporal decay constant (b) of calcium transients for the model with RyR2s inhibited by CaM (ctrl), the model with disinhibited RyR2s (RyR2 no CaM) and the model without RyR2s (no RyR2). Disinhibition of RyR2 lowers spatial and temporal specificity of calcium transients (p < 0.001 for every panel of (a) and (b)). The horizontal error bars do not appear because they are smaller than the data points on the plots. With disinhibition of RyR2 (RyR2 no Cam) the spatial spread of calcium transients exceeded 5μm. The spatial extent is the furthest location from stimulated site (middle of the dendrite) where calcium amplitude following stimulation exceeds 250% of resting calcium concentration. (b) CICR did not prolong calcium transients. Time decay constant was obtained by fitting a single exponential decay function to calcium concentration after the stimulation at the stimulated section of the dendrite. Spatial spread of calcium transients for a bAP-like stimulus. Only results for the thin dendrite are shown. A short calcium input (3ms) was applied directly to the cytosol at one side of the dendrite as in [23] (also shown in Fig 2). Disinhibition of RyR2 by calmodulin (empty circles) increases spatial extent, but does not allow for calcium wave propagation. Spatial extent was the furthest location from the stimulated site where calcium amplitude following stimulation exceeded 250% of resting calcium concentration. The horizontal error bars do not appear because they are smaller than the data points on the plots. Old age prolongs calcium transients in space and time. The difference in spatial spread and temporal decay constant for the ctrl model (RyR2 inhibited by calmodulin) and old age model is statistically significant (p < 0.001 for every panel of (a) and (b)). (a) Spatial extent of calcium elevation in control conditions (filled circles) and the old-age model (empty circles). Spatial extent was the furthest location from stimulated site where calcium amplitude following stimulation exceeded 250% of resting calcium concentration. (b) Duration of calcium elevation in control conditions (filled circles) and the old-age model (empty circles). Time decay constant was obtained by fitting a single exponential decay function to calcium concentration after the stimulation at the stimulated section of the dendrite. Comparison of spatial spread and temporal decay constant of calcium transients for models applying single changes from the old age model with the control model (solid circles) and the old age model (empty circles). Only results for the thin dendrite are shown. (a) Restoring RyR2 inhibition in the old age model (empty squares) lowered the spatial extent to levels similar to the ctrl model. Spatial extent was the furthest location from stimulated site where calcium amplitude following stimulation exceeded 250% of resting calcium concentration. (b) Either increasing Ca2+-buffering of the ctrl model (filled squares) or disinhibition of RyR2 (filled triangles) decreases temporal decay constant. Filled circles show the extent of temporal specificity in control conditions. Time decay constant was obtained by fitting a single exponential decay function to calcium concentration after the stimulation at the stimulated section of the dendrite. The kymographs (Fig 7) show temporal dependence of the intracellular calcium concentration (averaged across the width) for every 0.5μm in the longitudinal direction. The mean and standard deviation of the intracellular calcium concentration (Fig 8) were calculated across the entire morphology for the whole simulation of 1000s. Spontaneous generation of calcium events in the AD model are seen in kymographs of typical resting intracellular calcium concentration (Ca𝐢2+ in the AD model with (a) the 1.2μ𝐦 dendrite, (b) the 2.4μ𝐦, and (c) the 6.0μ𝐦 dendrite. (d) Expansion of (a) showing that calcium waves sometimes begin in the middle of the dendrite. Lowering activity of PMCA pumps is necessary for an elevation in resting calcium concentration, while disinhibition of RyR2s increases the amplitude of calcium fluctuations. Each panel shows mean and standard deviation of intracellular calcium concentration Cai2+ in the AD model (red box) and 7 models with varying PMCA activity and concentration of RyR2 inhibited by calmodulin (denoted as RyR2CaM) and disinhibted RyR2 (denoted as RyR2). The ctrl model is denoted with the blue box. RyR2 disinhibition alone did not change resting intracellular calcium concentration and can not explain the elevated resting intracellular calcium concentration observed in AD. Further increasing the concentration of disinhibited RyR2 (100% RyR2 concentration of the ctrl model) only increased resting intracellular calcium fluctuations. Mean and standard deviation of resting calcium concentration was calculated across all the time points and voxels in a 1000s long simulation. For Fig 9 we identified calcium events in the AD models as clusters, in space and time, of elevated intracellular calcium concentration, larger than 250% of the mean intracellular calcium concentration of the control model). Amplitude is the mean of maximum amplitude of the calcium events; peak frequency is the number of calcium events divided by the simulation duration (1000s); peak duration is the average duration of the calcium events; and spatial spread is the average of spatial spread of the calcium events. Fig 9 shows mean values of the spatial spread and temporal decay contant with standard error as errors bars obtained in 4 simulations with different seeds. Properties of calcium events observed in the AD model (red box) and the 4 models with diminished PMCA activity. Calcium events were identified as a calcium elevation exceeding 250% of mean resting intracellular calcium concentration of the control model. (a) Frequency of calcium events was inversely proportional to the dendritic diameter. Frequency was calculated as the average number of calcium events divided by the duration of the simulation. (b) Amplitude of calcium events was inversely proportional to the dendritic diameter. Mean amplitude was calculated as the average maximum amplitude of the calcium events. (c) Peak duration of the calcium events was proportional to the dendritic diameter. Mean peak duration was calculated as the average duration of the calcium events. (d) There is not a consistent effect of diameter on the peak spatial spread. Mean spread was calculated as the average spatial spread of the calcium events. To establish whether different model variants yield different spatial spread and temporal decay constant of calcium transients we performed linear regression analysis of spatial spread/temporal decay values versus log2max(Cai2+), which linearized the results. For all analyzed model variants (ctrl, no RyR2, RyR2 no CaM, old age, old age with RyR2CaM, ctrl with 200% Ca2+-buffers) linear regression fits were statistically significant with p < 0.001. To establish if model variants differ we performed ANCOVA, with model variant as co-variate, for every model panel in Figs 3,5 and 6. For statistical analysis we used statsmodels Python package [86]. Results RyR2s do not support intracellular calcium wave propagation To test in-silico whether RyR2 alone can support intracellular calcium wave propagation caused by synaptic activation, we stimulated with calcium input models of a 51μm long dendrite with three diameters: 1.2μm (thin), 2.4μm (medium) and 6.0μm (thick). We modeled RyR2 constitutively bound to calmodulin (denoted as ctrl model), because the concentration of free calmodulin in the brain exceeds 10μm [14,15] and calmodulin affinity for the inhibitory binding site of RyR2 is in the range of 820 nM for apo-calmodulin [38] to 4 nM [66] for calcium-bound calmodulin. The binding of calmodulin to RyR2 is in fact very slow with dissociation rate τ12 reaching up to minutes [66] suggesting that most of RyR2s are bound to calmodulin in control conditions even during a calcium transient. To demonstrate the role of calmodulin inhibition, we also simulated the model without any RyR2. We found that the ctrl model cannot support intracellular calcium wave propagation, and the absence of calmodulin-bound RyR2 (RyR2CaM) had no effect on the spatial extent of calcium transients in thin and medium dendrites (Fig 3a, Table 11). Spatial spread of synaptically evoked calcium transients depended on the stimulation intensity, which was consistent with experimental observations [24] (Table 11). The increase in spatial extent with stimulation intensity was lowest in thick dendrites, probably due to the larger impact of buffering [87]. Without RyR2 or with calmodulin inhibited RyR2, the spatial spread of calcium transients did not exceed 5μm from the stimulation location, which is consistent with values reported in [87] (2.5μm measured around the source as full width around half maximum). Stimulation intensity also prolonged calcium transients, again with no difference between excluding RyR2 and calmodulin inhibited RyR2. 10.1371/journal.pcbi.1013752.t011Table 11ANCOVA statistical significance results. Independent variables were condition and log2 max(Ca𝐢2+).Figvariabledend diamF, pT, P for log2max(Ca𝐢2+)T, P for condition3aspatial extent1.2364, < 1e-517.9, < 1e-5RyR2 no CaM: 21.1, < 1e-5; no RyR2: -1.83, 0.0693aspatial extent2.4122, < 1e-516.0, < 1e-5RyR2 no CaM: 7.4, < 1e-5; no RyR2: -1.48, 0.1423aspatial extent6.0148, < 1e-517.7, < 1e-5RyR2 no CaM: 9.6, < 1e-5; no RyR2: -0.49, 0.8673btime decay1.263, < 1e-511.2, < 1e-5RyR2 no CaM: 8.7, < 1e-5; no RyR2: 0.1, 0.9343btime decay2.448, < 1e-55, < 1e-5RyR2 no CaM: 9.0, < 1e-5; no RyR2: -0.34 0.7343btime decay6.066, < 1e-511.9, < 1e-5RyR2 no CaM: 5.8, < 1e-5; no RyR2: -1.1, 0.2675aspatial extent1.2434, < 1e-523.8, < 1e-513.7, < 1e-55aspatial extent2.4344, < 1e-522.2, < 1e-511.1, < 1e-55aspatial extent6.0368, < 1e-525.6, < 1e-55.7, < 1e-55btime decay1.268, < 1e-57.6, < 1e-57.6, < 1e-55btime decay2.478, < 1e-57.6, < 1e-58.9, < 1e-55btime decay6.0185, < 1e-512.4, < 1e-513.0, < 1e-56aspatial extent1.2470, < 1e-531.5, < 1e-5aging: 15.7, < 1e-5; aging with ctrl RyR2: 2.4, 0.0156btime decay1.274, < 1e-59.6, < 1e-5aging: 10.1, < 1e-5; ctrl+200% buffers: 3.1, 0.002; RyR2 no CaM: 4.3, < 1e-5 ANCOVA statistical significance results. Independent variables were condition and log2 max(Ca𝐢2+). Calmodulin inhibition of RyR2 preserves spatial specificity of calcium transients To investigate the role of calmodulin inhibition of RyR2 in shaping calcium dynamics, we substituted the model of RyR2CaM with a model of disinhibited RyR2 developed by [37]. We immediately observed an increase in calcium leak from the ER, which resulted in lowering of luminal calcium concentration by 10% in medium and thick dendrites. Resting intracellular calcium concentration levels stayed constant and excess intracellular calcium was extruded into the extracellular space by the PMCA and NCX. Disinhibition of RyR2 increased the spatial spread (Fig 3) by a factor of 2 and greatly increase the temporal decay constant of the calcium elevation (Fig 3). Both effects were statistically significant (p < 0.001 for both spatial spread and temporal decay constant for every dendritic diameter, Table 11). However, the increase in spatial spread did not allow for calcium wave propagation either for the synaptic-like (Fig 3) or short bAP-like stimulation (Figs 4 and S1). For the disinhibited model both spatial and temporal specificity of calcium transients were lowest in thin dendrites and highest in thick dendrites. Calmodulin controls spatial and temporal specificity of calcium transients in old age Old age is accompanied by larger [88] and longer [89] calcium transients with higher spatial spread [90], which lead to higher calcium accumulation in the cytosol [91], which is likely responsible for difficulties in learning and memory. Old age is also characterized by higher calmodulin oxidation [65], which in turn causes lower activity of PMCA pumps [67–69] and reduces calmodulin binding to RyR2 [16,64,92] resulting in RyR2 favoring activation by calcium [75]. RyR2 expression is also increased [72,93]. Moreover, the buffering capacity of intracellular calcium buffers of CA1 pyramidal neurons is increased [71]. Finally in old age store-operated calcium entry [70] and the T-type VGCC [73] are down-regulated. To investigate how these changes interact to produce deficits in calcium dynamics we created an old age model implementing those changes. First we simulated synaptic activation as a 40ms calcium input into submembrane voxels for the model of old age (Fig 5). We saw that increased calmodulin oxidation (resulting in disinhibition of RyR2 and lower activity of PMCA pumps) in the old age model was responsible for decreased spatial and temporal specificity of calcium transients. The simulations showed that thick dendrites were the most robust to calmodulin oxidation (Fig 5a), although for all dendrite models, the increase of spatial spread in the old age compared to the ctrl model was statistically significant (Table 11, p < 0.001). All dendrite models exhibited prolongation of the calcium elevation (Fig 5b) (Table 11, p < 0.001) which is consistent with results from the previous section, although the duration of calcium transients was two times lower than for disinhibition of RyR2 alone. Next we investigated which pathways downstream of calmodulin activation are responsible for the increased spatial extent (Fig 6a) and longer duration (Fig 6b) of calcium transients in old age. We investigated two separate aspects of the old age changes in calmodulin activity with the following parameter variations: old age with ctrl RyR2: 80% of PMCA activity, 200% calcium buffering molecules concentration, 200% RyR2CaM, old age with ctrl RyR2: 80% of PMCA activity, 200% calcium buffering molecules concentration, 200% RyR2CaM, ctrl with 200% of Ca2+-buffers: ctrl model with 200% calcium buffering molecules concentration, ctrl with 200% of Ca2+-buffers: ctrl model with 200% calcium buffering molecules concentration, and repeated in-silico experiments. Disinhibition of RyR2s was responsible for diminished spatial specificity of calcium transients. Figs 6a and S2 show that the old age model with ctrl RyR2, i.e., inhibited by calmodulin (old age with ctrl RyR2, empty squares) exhibited similar spatial spread to the control model (filled circles) (Table 11, p = 0.002). Figs 6b and S3 indicates that both RyR2 disinhibition (RyR2 no CaM, filled triangles) and higher calcium buffering (ctrl+200% Ca2+-buffers, filled squares) each can increase duration times of calcium transients (Table 11, p = 0.002), although this increase is lower than the one observed with the old age model. This is consistent with results shown in the previous section, which revealed that disinhibition of RyR2 significantly impacted temporal properties of calcium transients, and indicates that both these mechanisms contribute to diminished temporal specificity of calcium transients in the old age model. Disinhibition of RyR2 alone cannot account for increased resting intracellular calcium concentration in Alzheimer’s disease Calcium dyshomeostasis [76] is one of the hallmarks of Alzheimer’s disease [76]. One of the ways this dyshomeostasis is manifested is an elevation in resting cytosolic calcium concentration, which is observed in pyramidal neurons of mice models of AD [77,78]. This elevation may be caused by RyR2 recruitment [94], as indicated by an increase in CICR and higher RyR expression [79,80]. Decreased inhibition of RyR2 by calmodulin can also cause increased RyR2 recruitment. Limiting RyR2 open time either by increasing its binding to calmodulin [6] or by a point mutation [7] alleviates cell loss and AD-like neuronal hyperexcitability, respectively. Alternatively, increased resting intracellular calcium concentration may be caused by a dysfunction of PMCA pumps [95], which are inhibited by tau proteins [96,97] and β-amyloid oligomers [98], both of which are accumulated during AD progression. To investigate possible causes of elevated resting intracellular calcium concentration in AD we tested the impact of these implicated mechanisms in-silico using an early onset AD model, which comprised the ctrl model (Table 2) with lower PMCA activity (Table 8), increased RyR2 concentration (Table 7) with half the RyR2 population inhibited by calmodulin and half disinhibited. It is important to note that the old age model and the AD model represent two different conditions. The old age model differs from the AD model in (a) the lack of SOCE and (b) higher buffering molecules, including neurogranin, and (c) unchanged calcium leak from the extracellular space, which represents resting calcium influx to the dendrite. A typical kymograph of the AD model resting intracellular calcium shows spontaneous generation of spreading calcium elevations (Fig 7a) for the 1.2μm model, and calcium elevations that traveled throughout the whole dendrite (calcium waves) for the 2.4μm (Fig 7b) and 6.0μm (Fig 7c) models. Amplitudes of these elevations exceeded 250% of the mean resting intracellular concentration in the control model (73 nM). These spontaneous events were not observed for the old age model, suggesting that in old age increased buffering helps preserve calcium homeostasis. We varied the parameters to investigate which mechanisms are responsible for generation of the spontaneous calcium events: 100% PMCA kcat 0% RyR2CaM 0% RyR2 – normal PMCA activity and no RyR2 receptors, 100% PMCA kcat 0% RyR2CaM 0% RyR2 – normal PMCA activity and no RyR2 receptors, PMCA kcat 0% RyR2CaM 100% RyR2 – normal PMCA activity and all RyR2 receptors disinhibited, PMCA kcat 0% RyR2CaM 100% RyR2 – normal PMCA activity and all RyR2 receptors disinhibited, 80% PMCA kcat 100% RyR2CaM 0% RyR2 – lower PMCA activity and all RyR2 receptors inhibited by calmodulin (AD without inhibited RyR2), 80% PMCA kcat 100% RyR2CaM 0% RyR2 – lower PMCA activity and all RyR2 receptors inhibited by calmodulin (AD without inhibited RyR2), 80% PMCA kcat 0% RyR2CaM 100% RyR2 – lower PMCA activity and all RyR2 receptors disinhibited(AD without disinhibited RyR2), 80% PMCA kcat 0% RyR2CaM 100% RyR2 – lower PMCA activity and all RyR2 receptors disinhibited(AD without disinhibited RyR2), 80% PMCA kcat 50% RyR2CaM 50% RyR2 – lower PMCA activity and half RyR2 receptor inhibited by calmodulin and half disinhibited (AD model with lower levels of both RyR2), 80% PMCA kcat 50% RyR2CaM 50% RyR2 – lower PMCA activity and half RyR2 receptor inhibited by calmodulin and half disinhibited (AD model with lower levels of both RyR2), 80% PMCA kcat 0% RyR2CaM 200% RyR2 – lower PMCA activity, doubled RyR2 concentration and all RyR2 receptors disinhibited, 80% PMCA kcat 0% RyR2CaM 200% RyR2 – lower PMCA activity, doubled RyR2 concentration and all RyR2 receptors disinhibited, and simulated them without any stimulation for 1000s to evaluate their resting intracellular calcium concentration. Fig 8 shows that RyR2 disinhibition alone did not increase resting intracellular calcium concentration. However, decreasing PMCA activity by 20% increased resting intracellular calcium concentration by 25% compared to the ctrl model. Interestingly, models with 80% of PMCA activity with increasing RyR2 concentration and increasing disinhibition of RyR2 did not show further increase of resting intracellular calcium concentration. Instead fluctuations of resting intracellular calcium concentration (represented by increased standard deviation of resting intracellular calcium concentration, shown in Fig 8) were observed. This shows that relieving calmodulin inhibition of RyR2s alone was not responsible for increased resting calcium concentration, but, instead, increased calcium fluctuations. A closer look at cases summarized in Fig 8 revealed that emerging calcium events, some of which were calcium waves, were observed for models with diminished PMCA activity (denoted as 80% PMCA kcat), such as the model with 80% of PMCA activity and 50% of RyR2 disinhibition and 50% of RyR2CaM (this behavior was observed for the three dendrite models). Fig 9 summarizes the properties of the calcium events observed in all 5 models with diminished PMCA activity. The frequency of calcium events depended on the dendritic diameter (Fig 9a). With 50% of RyR2 disinhibited, calcium events were more frequent (Fig 9a), but also more localized (Fig 9a) and Fig 9c). The amplitude of calcium events depended only on the dendritic diameter (Fig 9b). Calcium events in thick dendrites had the longest duration (Fig 9b), but the largest spatial spread of calcium waves was observed in medium dendrites (Fig 9d). Store operated calcium entry does not affect CICR spatial specificity in control conditions Finally we investigated the role of store replenishment in spatial specificity of CICR. The endoplasmic reticulum membrane contains STIM proteins, which serve as ER calcium sensors. When ER calcium is depleted, STIM molecules bind and activate Orai1 channels located in the plasma-membrane. Calcium flows into the cell through those channels and gets transported by SERCA pumps into the ER. Fig 10 shows that store operated calcium entry (SOCE) does not increase the spatial extent of calcium transients due to the much slower time scale of SOCE activation than that of calcium dynamics. Although, experimental evidence [99] shows that STIM-Orai1 signaling in the spine is necessary for maintaining long-term structural plasticity, in this work we utilized stimulation resembling single solitary synaptic activation, which barely affects luminal calcium levels. Store operated calcium entry (SOCE) did not influence spatial specificity of calcium elevations. Spatial extent was the furthest location from stimulated site where calcium amplitude following stimulation exceeded 250% of resting intracellular calcium concentration. Robustness of the model We analyzed the change in spatial spread of calcium elevations in response to small changes in PMCA activity and concentration, and the luminal calcium concentration (shown in Fig 11). As expected due to the large impact of calcium buffering, thick dendrites were the most robust to parameter variation. Small increases in ER calcium concentration did not affect the spatial extent of calcium elevations for all three model dendrites. PMCA concentration and activity changes impacted thin and medium dendrites the most. Thick dendrites are most robust to parameter variation. Panels illustrate the change in spatial extent for the range of calcium influx values. Change in PMCA activity produced the highest variation in spatial extent, especially for thin and medium dendrites. Horizontal lines represent mean and standard deviation of percentage error for each paradigm. Open symbols represent individual values of percentage error. Discussion In this work we investigated how calcium homeostasis is controlled by calmodulin in control conditions, aging and Alzheimer’s disease. We concentrated on calcium-induced calcium release, in which calcium flows out of the ER into the cytosol through RyR2, and calcium extrusion into the extracellular space by PMCA. We showed that RyR2s alone are unlikely to support intracellular calcium wave propagation both in control conditions (when they are constitutively bound to and in consequence inhibited by calmodulin) and in old age (when both RyR2s and calmodulin become oxidized, which in turn disinhibits some of the RyR2s). Calmodulin inhibition of RyR2s played a crucial role in controlling spatial and temporal specificity of calcium transients. Finally, we investigated possible sources of increased resting intracellular calcium concentration in Alzheimer’s disease, showing that disinhibition of RyR2s could not account for this phenomenon, but, instead, diminished activity of PMCA increased intracellular resting calcium. The influence of calmodulin and RyR2 on calcium dynamics is critical because learning, memory and many other processes in the brain are controlled by calcium dynamics. Dendritic calcium plays a crucial role in behavioral time scale plasticity, a very powerful form of synaptic plasticity requiring a prolonged dendritic calcium elevation, which underlies tuning of spatial features in dendrites and is capable of producing place cells [100]. Interestingly, blocking mitochondrial calcium buffering, and hence increasing calcium release from the ER, led to higher stability of place cells [101]. Dendritic calcium is strongly correlated with spine calcium [84] with RyR3s (activated by calmodulin) on the spine endoplasmic reticulum (spine apparatus) transducing spine calcium signals to the dendrite [102,103]. Calmodulin acting on RyRs differentially regulates the calcium transients in the dendritic branch: leading to an amplification of calcium transients from individual spines and restricting their spread in the dendrite leading to larger spatial and temporal specificity of calcium signaling in the dendrite. Thus, RyRs can influence learning and memory through their contribution to dendritic and spine calcium transients. Results showing the lack of intracellular calcium wave induction in the control model are in agreement with experimental evidence, which showed that IP3 production is necessary [24–27], and in conflict with other modeling work [23], which did not incorporate calmodulin inhibiting RyR2. More recent work in hippocampal cultures shows that RyR2 signaling contributes to nuclear calcium release [21], although this process required depolarization of membrane potential [19,20], which was not included in the model. Moreover, contrary to [23] in Fig 3 we showed that, even if RyR2s were disinhibited by calmodulin, calcium inputs corresponding to excitatory postsynaptic potentials (EPSP) and bAPs (Figs 4 and S1) could not elicit intracellular calcium waves. The lack of calcium waves may be advantageous since limiting intracellular calcium release supports the establishment of dendritic feature selectivity [101]. Although, we used similar RyR2 densities (4 µm–2) in our model, there were many methodological differences from [23]: instead of using a deterministic model we used a stochastic approach for simulation of reaction-diffusion systems, which is better suited to dynamical systems with low copy numbers [104] and in such systems deterministic models might yield different predictions than stochastic models [105]. One of the differences is that deterministic rate laws depend on concentrations whereas stochastic rate laws depend on the number of possible combinations of reacting species. With high copy numbers stochastic and deterministic models will yield similar results, whereas with low copy numbers the reactions might differ substantially, e.g., contributing to higher noise sensitivity of deterministic models [105]. The stochastic approach we used is a spatial extension of Gillespie’s tau leap algorithm, which dynamically selects between simulating an exact event or a leap event both for reactions and diffusion between voxels [85]. This algorithm provides the computational efficiency needed for simulating this mathematically stiff system. Finally, we used a stochastic model of RyR2 activation [37] tuned to match mean open and close time of the RyR2 [39] instead of using the Keizer-Levine model of RyR2 activation [106], which exhibits very long RyR2 open times. It is possible that the induction of intracellular calcium waves in [23] resulted from using the Keizer-Levine RyR2 model. In fact, when we performed stochastic simulations with the model of resting dynamics, we observed spontaneous, oscillatory generation of intracellular calcium waves in thin dendrite. We did not compare predictions of our model with [28] and [29], because both models implemented IP3-dependent calcium wave generation. We modeled calmodulin constitutively bound to RyR2s in control conditions. Calmodulin concentration in the hippocampus is high (20 µM) [14]) and calmodulin affinity for the inhibitory binding site of RyR2 in the range of 820 nM for apo-calmodulin [38] to 4 nM [66], with dissociation rate τ12 reaching up to minutes [66]. This suggests that most of RyR2s are bound to calmodulin in control conditions even during a calcium transient. In fact, we did a limited number of simulations with dynamic binding of calmodulin to RyR2 during a transient calcium elevation and observed that new calmodulin did not bind to the RyR2 subunit due to the slow reaction rates. In oxidized conditions and with cytosolic calcium levels (100 nM), calmodulin affinity for RyR2 drops twenty-fold [66] with 50% RyR2s bound to calmodulin. Therefore we simulated calmodulin constitutively bound to RyR2 in control conditions and 50% of RyR2 constitutively bound to calmodulin in the model of aging. This resulted in total RyR2 population favoring activation by calcium for the model of aging consistent with experimental data [75]. The endoplasmic reticulum in dendrites forms a dense network of interconnected cysternae with an occasional single large cysternae close to dendritic membrane with mitochondria in the middle of the dendrite [35,107]. Modeling such morphology with fidelity would require a particle-based approach, which is computationally prohibitive in the scale of model used here. Instead, in our mesoscopic model of calcium dynamics in a 51μm long dendrite, the ER was treated as a separate compartment that is uniformly distributed in the dendrite, occupying 25% of its volume. Tortuosity of the ER was taken into account by reducing the diffusion rate of calcium molecules in the ER, compared to diffusion rate of the cytosolic calcium molecules. The randomly placed ryanodine receptors allow for a release of ER calcium molecules into the cytosol, with flow of calcium down the concentration gradient through open RyR receptors. We showed that both spatial [90] and temporal specificity of calcium transients decreased due to the higher oxidation of calmodulin in aging, which likely contributes to deficits in learning and memory accompanying old age. In control conditions the extent of calcium elevation in thin dendrites reached 5μm, which was similar to the average inter-spine distance after contextual fear conditioning task observed in apical dendrites of layer V motor neurons [108], indicating that this distance might indeed be governed by calcium release from the ER. When calmodulin inhibition of RyR2 was relieved (e.g., by higher oxidation of calmodulin), the spatial extent of synaptically evoked calcium transients was increased by roughly a factor of 2 and duration was increased for thin and medium dendrites. The decrease in spatial specificity was caused by the oxidized calmodulin not inhibiting RyR2, but the decrease in temporal specificity was caused both by RyR2 disinhibition and an increase in cytosolic calcium buffering. Interestingly, increased calcium cytosolic buffering acted like a double edged sword: it counteracted some of the calcium dyshomeostasis accompanying old age and prevented spontaneous generation of calcium events and waves keeping the resting intracellular concentration unchanged [109], but also decreased temporal specificity of calcium transients. Increasing oxidation levels leads to both lower PMCA activity [110] and RyR2 disinhibition [5,16]. In our model of early onset AD, with unchanged resting calcium influx to the dendrite lower PMCA activity alone resulted in the elevation of intracellular calcium concentration. Coupling the lower PMCA activity with disinhibited RyR2 caused the emergence of spontaneous calcium events, with frequency dependent on dendritic diameter and disinhibited RyR2 concentration. In rat hippocampal CA1 pyramidal neurons [111–113] and CD1 mice hippocampal interneurons [114] spontaneous fast calcium events (calcium sparks) depended on RyR activation, resting intracellular calcium concentration and L-type VGCCs. The events observed in the model did not resemble that experimental phenomenon, instead their frequency was much lower (less than 0.02 event/sec compared to 1 event/sec [111,112])) and their duration much longer (tens of second compared to 400ms [112]). These results suggest that calcium events were driven by the degree of fluctuations caused by the increased activity of disinhibited RyR2s, which had a higher affinity for calcium ions, and elevated resting intracellular calcium concentration caused by lower activity of PMCA pumps. A crucial question is whether disinhibition of RyR2s can be responsible for the elevated resting intracellular calcium concentration observed in Alzheimer’s disease [6,7]. It is important to note that the old age model and the AD model represent two different conditions. The old age model differs from the AD model in (a) the lack of SOCE, (b) higher buffering molecules, including neurogranin, and (c) unchanged calcium leak from the extracellular space, which represents resting calcium influx to the dendrite. We showed that disinhibition of RyR2s, which was characterized by higher opening probability with longer open times [38,39], caused higher calcium leak from the ER. This did not elevate resting intracellular calcium, because excess calcium was extruded outside the cell by the calcium extrusion pumps. We also investigated an alternative hypothesis of dysfunction of PMCA pumps [95] by lowering their activity, which resulted in an elevation of resting intracellular calcium concentration by 25% compared to the ctrl model. This increase was lower than the observed doubling of resting intracellular calcium concentration [49,77,78]. One limitation of the model is that it did not include β-amyloid. Calcium promotes β-amyloid oligomer formation [115], which can promote oxidative stress in neurons [116,117]. Incubation of primary neuronal cultures with β-amyloid peptides leads to a significant increase in resting intracellular calcium and cell apoptosis [118,119], which was prevented by blocking RyRs. These results suggested that lowering of PMCA activity in the vicinity of β-amyloid plaques [98] might be responsible for elevated calcium [78]. It is possible that in early asymptomatic stages of AD soluble β-amyloid oligomers promote calcium dyshomeostasis via both inhibition of PMCA pumps and disinhibition of RyR2s, which in turn promotes β-amyloid plaque formation. Future models can evaluate whether a positive feedback loop between β-amyloid oligomer formation and calcium elevation can completely explain the experimentally observed doubling of resting calcium concentration. Conclusion Calmodulin, the jack of many trades, controls calcium dynamics and homeostasis not only by acting as a fast calcium buffer: spatial and temporal specificity of calcium transients in dendrites of CA1 pyramidal neurons is regulated by calmodulin inhibition of RyR2 and basal intracellular calcium concentration by calmodulin activation of PMCA pumps. It is rather difficult to pinpoint one molecule responsible for calcium dyshomeostasis. Instead, it is an ensemble of factors and molecular species interacting with each other with an emerging role for calmodulin, which regulates both calcium extrusion and calcium-induced calcium release. Supporting information Spatial spread of calcium transients for a bAP-like stimulus. A short calcium input (3 ms) was applied directly to the cytosol at one side of the dendrite as in [23] (also shown in Fig 2). Disinhibition of RyR2 by calmodulin (empty circles) increases spatial extent, but does not allow for calcium wave propagation. Spatial extent was the furthest location from the stimulated site where calcium amplitude following stimulation exceeded 250% of resting calcium concentration. The horizontal error bars do not appear because they are smaller than the data points on the plots. (EPS) Comparison of spatial spread of calcium transients for models applying single changes from the old age model with the control model (solid circles) and the old age model (empty circles). Restoring RyR2 inhibition in the old age model (empty squares) lowered the spatial extent to levels similar to the ctrl model. Spatial extent was the furthest location from stimulated site where calcium amplitude following stimulation exceeded 250% of resting calcium concentration. (EPS) Comparison temporal decay of calcium transients for models applying single changes from the old age model with the control model (solid circles) and the old age model (empty circles). Either increasing Ca2+-buffering of the ctrl model (filled squares) or disinhibition of RyR2 (filled triangles) increases temporal decay. Filled circles show the extent of temporal specificity in control conditions. Time decay constant was obtained by fitting a single exponential decay function to calcium concentration after the stimulation at the stimulated section of the dendrite. (EPS) 26 Mar 2026 Calmodulin controls spatial and temporal specificity of calcium-induced calcium release PLOS Computational Biology Dear Dr. Joanna Jędrzejewska-Szmek, Thank you for submitting your manuscript to PLOS Computational Biology. After careful consideration, we feel that it has merit but does not fully meet PLOS Computational Biology's publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by 25 April. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at ploscompbiol@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pcompbiol/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. Please include the following items when submitting your revised manuscript: * A letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. This file does not need to include responses to formatting updates and technical items listed in the 'Journal Requirements' section below. * A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. * An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If you would like to make changes to your financial disclosure, competing interests statement, or data availability statement, please make these updates within the submission form at the time of resubmission. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter We look forward to receiving your revised manuscript. Kind regards, Ausra Saudargiene, Ph.D. Academic Editor PLOS Computational Biology Andrea E. Martin Section Editor PLOS Computational Biology Journal Requirements: 1) Please ensure that the CRediT author contributions listed for every co-author are completed accurately and in full. At this stage, the following Authors/Authors require contributions: Joanna Jędrzejewska-Szmek, and Kim T Blackwell. Please ensure that the full contributions of each author are acknowledged in the "Add/Edit/Remove Authors" section of our submission form. The list of CRediT author contributions may be found here: https://journals.plos.org/ploscompbiol/s/authorship#loc-author-contributions 2) We ask that a manuscript source file is provided at Revision. Please upload your manuscript file as a .doc, .docx, .rtf or .tex. If you are providing a .tex file, please upload it under the item type u2018LaTeX Source Fileu2019 and leave your .pdf version as the item type u2018Manuscriptu2019. Reviewers' comments: Reviewer's Responses to Questions Comments to the Authors: Please note here if the review is uploaded as an attachment. Reviewer #1: The investigation of CICR in control and experimental conditions -aging, AD- is relevant and original. The model explores different scenarios and investigates the role ER Ca2+ release, RyR2, PMCA and CaM in this process. However, the authors need to address some issues. Justify why a stochastic model is really needed. Compare with a deterministic solution. The work mention the role of Ca2+ in learning and memory, but the model does not reproduce learning an memory phenomena. How to truly link the model results with learning and memory? Check throughout for typos and English errors. To mention some: ‘disinhibted’, pg. 9. ‘AD we tested …’, pg 12. Check labels and titles in all figure: Peak Ca or peak Ca^2+? Ca or Ca^2+? Make sure fonts are consistent in the figures. Fig. 15: Paradigm is upside-down. Fig 2: Use the proper superscript for 2+ in Ca2+ There is an overlap in the header of the tables 2, 7 and 8 between kf and kr. Justify the inconsistent units between tables for kf. Table 4: kf(nM/ms) Table 2 kf(nM^-1 ms^-1). Table 6 kf(nM^4/ms). How are the different units handled in the program code? Reviewer #2: Major Comments: 1) The text of the paper, including but not limited to the Discussion, needs to be revised to eliminate any implication that the authors are simulating any changes or processes specific to Alzheimer’s disease. Because, in fact they are not, all the listed changes in model parameters – altered RyR2-Cam binding, decreased PMCA pump activity, downregulation of SOCE and of T-type Ca channels – are generic to old age, not specific to AD. Indeed, some of the tables of model parameter values have columns listing old age values, but not columns listing different AD values. If the authors really want to claim they are simulating an AD condition, they need to extend the model and present additional simulations, by adding the effects of elevated beta-amyloid that they note in the Discussion. 2) The authors say their model is multi-compartment, but do not give or sufficiently justify a description of these compartments. How is their submembrane compartment different from the rest of the dendrite, why does it have the given width? Are there simple exchange terms from this compartment to the rest? Figure 3 has a very odd depiction of the dendrite as a gold cylinder hovering between two planar plasma membranes. I think the dendrite is instead supposed to be a box as in Fig. 2, so fix Fig. 3. Also, from the author’s preceding paper ref. 32, I guess the dendrite is divided into voxels. What is their size and shape? If NeuroRD does it automatically, still give an approximate description. Is diffusion modeled using the diffusion partial differential equation everywhere, or is it modeled by simple exchange terms at the boundaries of voxels with voxel interiors assumed homogenous, or is there both diffusion and exchange terms? 3) The paper implies, at first reading, that the ER is modeled as distinct internal compartments with their own surfaces and internal volume. Figure 3, right panel, illustrates the model as if this was the case. But for this model variant (I comment below on the left panel) It looks like the authors distributed the variables RyR2, RyR2Ca, and CaER uniformly throughout (according to lines 74-76) a quarter of the dendrite, with this quarter consisting of one or more compartments with RyR2 and RyR2Ca volume-distributed. The authors did not, it appears, construct separate ER compartments with their own surface putting RyR2 on the surface of these compartments. And, given that the authors do not mention separate CaER variables for separate compartments, I think they only assume one internal ER compartment, which is contrary to what the figure shows. It is essential that the article text, and the figure in particular the location of RyR2, SERCA.., be clarified, and extended as necessary, so that no misleading material is present in this regard. It would also be good to give some detail about how existing data-based ER reconstructions compare with this model. 4) The ER model in the left panel of Fig. 3, in which the RyR2s are all located “at the dendritic membrane”, by which is meant either the lower or upper face of the model’s submembrane compartment, seems unjustified. It looks like here the authors put a planar box, right below the submembrane compartment, taking up a quarter of the dendrite volume, with the RyRs and SERCA all on that upper face. In reality, even if an ER cisternae lies against the membrane, I don’t think the RyRs and SERCAs would all be on that face, they should be on other faces of the ER compartment as well. Is there relevant data, has anyone showed by immunostaining or another method how RyR2 or SERCA is distributed within the dendrite? Anyway, the authors state later that simulations with this model variant are not significantly different from simulations with the model variant in the right panel. Thus, I think that model variant should be removed. 5) The authors need to state if, and for how long, their model with basal parameters and starting from initial concentrations was equilibrated, prior to giving a stimulus. 6) References, or other justification for estimates, need to be provided for the parameter values in Tables 1, 3, and 5. For non-ER parameters some of this may already be in the authors’ preceding paper, Ref. 32. If so make it clear to the reader, describing for which parameters references or justification is found in Ref. 32. But for parameters new to this model it needs to be done here. 7) It is stated that the simulations are performed stochastically, with use of the Gillespie algorithm. Is this true even for the most abundant molecules and fastest events, like calcium binding and unbinding from buffers? What about calcium diffusion. It appears the authors likely use a hybrid simulation scheme where (some or all) reactions are Gillespie and diffusion (at least) is not? 8) Tables 6 and 7 repeatedly use the phrase “adjusted to effect”. Explain what this means. For example it could mean data shown in a particular figure of the referenced publication. 9) I see one place, Figure 7, where the authors give a calcium stimulus amplitude, but it is in molecules injected, which is hard to relate to physiological stimulus amplitudes as commonly reported. Please compare the amplitudes to physiological stimulus amplitudes in some manner. What is the duration? On line 110 the authors say the authors increased calcium stimulus to take into account L-type channel upregulation. How much? But in fact, in the Discussion it says L type VGCCs were “not included in the model”? What about the T-type channel downregulation the authors say also occurs in aging? 10) During each simulation in the old age condition, when RyRs are only 50% saturated with CaM, this level of saturation is not a dynamic variable. In actuality, though, this level should vary because the binding of calcium to CaM is dynamic, calcium is varying, and apo-CaM and Ca-bound CaM should have different RyR2 affinities. If the authors don’t want to add simulations, they need to acknowledge in the Discussion that this is a model limitation, or else justify why it can be neglected. I guess if half of the RyR2s are actually normal and half can’t bind CaM at all, then it could be neglected, but this should be better justified. 11) For Figure 1, the legend is not adequate. I’m not sure this figure shows all the pathways, given that this model builds on that presented in Ref. 32. In the previous Ref. 32 the authors have another similar figure of signaling pathways. Tell the reader; if they combine that figure with this Fig. 1, do they see all the pathways? 12) Lines 275-277, there is a sentence, “This result is consistent with experiments showing that in hippocampal CA1 neurons IP3-sensitive calcium store refilling required ORAI1 channels, whereas ryanodine-sensitive calcium store refilling involved VGCCs”. This sentence is not justified, because none of the simulations in this paper were designed to specifically examine depletion of calcium stores or mechanisms for their refilling. Either remove this sentence or add additional, appropriate simulations carried out for a longer time during which ORA1I and VGCCs are separately inhibited. 13) Lines 79-82 as written imply two different models were used to describe the RyR2. This needs to be better explained. How is the model for the oxidized state (the Keizer Smith model) different from the model for the non-oxidized state? Can this difference be justified? Minor Comments: 1) In Fig. 11b, it looks like the calcium waves always start at one end of the dendrite. Why don’t some initiate in the middle, like the calcium sparks do? 2) Tables should be numbered in the order in which they are introduced in the manuscript, and should be briefly described when first introduced (this is not done for Table 1). 3) The legend of Fig. 2 is not adequate. Give dimensions, including the submembrane compartment. The legend of Fig. 3 needs to be revised because the model is not actually implementing “parallel cisternae” as compartments with distinct surfaces. 4) There are a number of minor grammar errors throughout the manuscript, go over it with either a human or AI editor. 5) Line 145, Fig. 4 should be Fig. 3. 6) The acronyms EPJ and SOCE need to be defined at first use. 7) Is “spatial extent” calculated for Fig. 5 the same as for Fig. 6? 8) The formatting of the References has numerous inconsistencies. 9) Lines 138-140, you say 4 um is a theoretical limit for calcium transient spread. I don’t see how this can be a theoretical limit, it must be parameter value dependent. Reviewer #3: This manuscript describes the use of a computational model to explore the contribution of calcium-induced calcium release (CICR) from endoplasmic reticulum (ER) to calcium transients and resting state in neuronal dendrites. Calcium dynamics are crucial to neuronal function and synaptic plasticity, so understanding the many mechanisms that determine such dynamics is highly important. This current work provides a useful increment to our knowledge, detailing the amplitude, spatial extent and time course of dendritic calcium transients due to pseudo-synaptic stimulation, and the contribution of CICR to this. The effects of changes in CICR during aging and Alzheimer’s disease are explored. The presentation of the work, however, could do with some improvement, as detailed below. Introduction Point the reader to Fig. 1 when ER first introduced eg around line 14. Methods Line 111: Details of the calcium stimuli used are needed (amplitudes, time course). Are these stimuli reasonable approximations to what is expected from synaptic activation on a spine head? Does figure 2 indicate that the stimuli are from a point source at the location indicated? Is the stimulus used for Fig. 6 different? Give details. Line 116: Much more detail about simulations and their analysis is needed. How many simulations are run to collect the values shown in the figures? Where are measurements made? What do the error bars show, and why are there not horizontal error bars showing variations in peak amplitude? Statistical analysis showing significance of differences between the different modelled situations is required to support the qualitative description of the results given. Why is spatial extent measured as distance at which amplitude drops below 250% of resting level? Results Some suggestions about improving the figures are given below. As noted above, since all of these results come from stochastic simulations, the statistical significance of differences between simulated situations can and should be calculated. Line 144: define explicitly what you mean by “spatial and temporal specificity” – presumably the spread of amplitude and decay time constant values? Line 167: more discussion is needed of what Fig. 6 is showing that is different from Figs. 4 and 5; what is different about these simulations and why do you only refer to these results re lack of calcium wave propagation (since figs. 4 and 5 indicated this too)? Line 168: what is Figure 7 for? Surely the same conclusions are already drawn before and this is just a different way of plotting the same data? Line 200: indicate to the reader that only results for the thin dendrite are shown in fig. 9 Line 248: change text to “for the situations summarized in Fig. 10”; how do the calcium dynamics illustrated in Fig. 11 (sparks and waves) impact on the data presented in Fig. 10: Surely the time varying amplitudes must affect the average calcium concentrations presented in fig. 10 (how were these measured? At a point in time and space, or across all time points?)? Do these dynamics cause the large increases in variance seen in Fig. 10 (ie is this what you mean in line 244?)? I am confused as to the difference between the results of Fig. 12 and those of Fig. 13a: this needs clarifying in the text. Line 260-262: what exactly do you mean by “widest” and “longest” calcium waves? Ie what are the widths and lengths that you are measuring? Line 283: and now, what do you mean by “CICR distance”? Explain this, or change back to previously used terminology eg spatial extent. Line 287: why is this unexpected? Discussion After the opening sentence, discuss more the evidence for the role(s) dendritic calcium might play in learning and memory, citing suitable references (such as [4] and [87]; any others?) Include brief discussion of RyRs in spines (RyR3 activated by calmodulin, rather than inhibited) and the role CICR might play in spine heads, as opposed to in dendritic shafts. Then start a new paragraph with “In this work…” Line 317 onwards: start new paragraph at “Although”; add “methodological differences from [23]”; add more text here describing where low copy numbers occur in your simulations that mean the use of a stochastic model is important and indicate what results may depend on this ie expand on your comment that predictions may depend on this. Line 379: clarify if you are referring to experimental data or to your simulations when you refer to emergence of spontaneous calcium events in this sentence. Figures In general, figure captions should specify what is being shown in the figure, rather than give a description of the results, which is covered in the main text (see my comment on fig. 4, but it applies to most of the figure captions). Figure 1: caption needs more detail about what is shown in the figure; the figure itself needs to be clearer as to the relationship between ER and the dendrite, and which is the dendritic (cellular) membrane and which the ER membrane. Figure 2: caption and/or figure could have more detail on subcompartment divisions (sizes). Figure 3: could possibly be merged with Fig. 2 by including one example on the medium dendrite and the other on the thick dendrite. Figure 4: here, and other figures, caption should describe what is plotted (and clarifying how calcium stimulus amplitudes relate to the data points shown), and not repeat the analysis of the results which is already given in the main text. Figure 5: this should be merged with Figure 4, as Fig. 4 only contains one situation that is not plotted on Fig. 5. Figure 6: caption needs further details of stimulation protocol, to distinguish results from Figs. 4 and 5 Figure 7: surely this figure could simply be removed; if there is a genuine need to retain it then more information must be given in the caption about what is plotted here. Figure 8: these results could also possibly be merged with those of figs. 4 and 5 to create a single figure. Figure 9: caption should indicate only thin dendrites shown. Minor corrections There are a number of typographical errors. The Discussion, in particular, needs careful proofreading. Here are some errors that were noted: Line 69: brackets around Table 4 Line 74: refer to fig. 3 here Line 94: define SOCE here (if not done earlier) and check that all other acronyms are defined at first usage Line 132: “can not” to “cannot” Line 161: refer to Fig. 5 here (near “observed”) Line 223: here and throughout, “dyshomeostasis” to “dishomeostasis” Line 232: here and throughout, “dysfunction” to “disfunction” Line 255: (Fig. 12) Line 264: some wording has gone missing here Line 293: delete “the” after “investigated” Line 300: “becomes” to “become” and “disinhibit” to “disinhibits” Line 318: “instead using” to “instead of using” Line 326: “with model” to “with the model” Line 331: um to uM Line 340: full stop at end of sentence Line 372: “lead” to “leads” Line 395: add “of calcium transients” after “specificity” ********** Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: Yes Reviewer #2: Yes Reviewer #3: Yes ********** PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: No Reviewer #3: No [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] Figure resubmission: After uploading your figures to PLOS’s NAAS tool - https://ngplosjournals.pagemajik.ai/artanalysis, NAAS will process the files provided and display the results in the "Uploaded Files" section of the page as the processing is complete. If the uploaded figures meet our requirements (or NAAS is able to fix the files to meet our requirements), the figure will be marked as "fixed" above. If NAAS is unable to fix the files, a red "failed" label will appear above. When NAAS has confirmed that the figure files meet our requirements, please download the file via the download option, and include these NAAS processed figure files when submitting your revised manuscript. Reproducibility: ?> 19 May 2026 Submitted filename: Revision letter (1).pdf 7 Jul 2026 PCOMPBIOL-D-25-02396R1 Calmodulin controls spatial and temporal specificity of calcium-induced calcium release PLOS Computational Biology Dear Dr. Joanna Jędrzejewska-Szmek, Thank you for submitting your manuscript to PLOS Computational Biology. After careful consideration, we feel that it has merit but does not fully meet PLOS Computational Biology's publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process. Please submit your revised manuscript by 7 August 2026. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at ploscompbiol@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pcompbiol/ and select the 'Submissions Needing Revision' folder to locate your manuscript file. Please include the following items when submitting your revised manuscript: * A letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. This file does not need to include responses to formatting updates and technical items listed in the 'Journal Requirements' section below. * A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. * An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'. If you would like to make changes to your financial disclosure, competing interests statement, or data availability statement, please make these updates within the submission form at the time of resubmission. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter. As the corresponding author, your ORCID iD is verified in the submission system and will appear in the published article. PLOS supports the use of ORCID, and we encourage all coauthors to register for an ORCID iD and use it as well. Please encourage your coauthors to verify their ORCID iD within the submission system before final acceptance, as unverified ORCID iDs will not appear in the published article. Only the individual author can complete the verification step; PLOS staff cannot verify ORCID iDs on behalf of authors. We look forward to receiving your revised manuscript. Kind regards, Ausra Saudargiene, Ph.D. Academic Editor PLOS Computational Biology Andrea E. Martin Section Editor PLOS Computational Biology Additional Editor Comments (if provided): Journal Requirements: If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise. 1)We have noticed that you have uploaded Supporting Information files, but you have not included a list of legends. Please add a full list of legends for your Supporting Information files after the references list Reviewers' comments: Reviewer's Responses to Questions Comments to the Authors: Please note here if the review is uploaded as an attachment. Reviewer #1: The manuscript has improved significantly. However, the statement that these changes “underlie changes in calcium dynamics that cause deficits in learning and memory” appears stronger than what the model directly supports. Since the model does not explicitly simulate learning or memory mechanisms, the authors should either soften the wording to indicate that the results may be associated with such deficits, or provide an additional analysis linking the modeled calcium signals to a downstream model of synaptic plasticity or another mechanism more directly tied to learning and memory, such as dynamic changes in the number of AMPA receptors in the postsynaptic membrane. Although there is experimental evidence that RyRs can influence learning and memory through their contribution to dendritic and spine calcium transients, this influence on learning and memory was not explicitly incorporated into the current model and should be made clear in the text. Reviewer #2: Major Comment: 1) The paper has been greatly improved. However, one major concern needs to still be addressed. In Tables 2 and 8, comparing the Control to Old Age condition, two offsetting parameter changes are made to control plasma membrane calcium fluxes. The Leak into the cell is reduced for old age (corresponding to decreased L-VGCC channels), and the PMCA kf is reduced. These changes offset such that resting cytoplasmic calcium stays about the same. That’s OK. The concern comes when going from the Old Age to the Alzheimer’s (AD) conditions. Then, the PMCA kf stays reduced, but careful inspection of the table shows that the authors put the Leak parameter back to the control value for AD, rather than keeping it at the reduced Old Age value. This increase in Leak is actually why resting calcium increases for AD compared to old age, and why spontaneous calcium transients occur, in the AD model! (this is evident from thermodynamic grounds, increased resting calcium has to be mediated by plasma membrane parameter changes) The resting calcium increase in AD is not, as text elsewhere implies, because of any specific decrease of PMCA activity in the AD condition relative to the Old Age condition. The lack of discussion of this increase in Leak is misleading to the reader, because most readers would assume that a simulation of AD, a disease that typically occurs in old age, would also include the L-VGCC reduction that is typical of old age. It’s hard to see this issue when it is just buried in the Methods. I didn’t find this important issue the first time I read the paper. It needs to be adequately described in the text of Results and/or Discussion. The specific Discussion statement on line 495, that "lower PMCA activity alone" increased resting calcium, is misleading and needs to be removed. One way to handle this issue could be, to suggest that the model is most appropriate for simulating early-onset AD, which may occur before the age-related decrease in L-VGCCs. In contrast, simulation of calcium dynamics specific to late-onset AD may in addition require, as you actually already describe to some extent towards the end of the Discussion, inclusion of beta-amyloid effects on plasma membrane calcium flux. Also check the text wording elsewhere in the paper to make sure it does not imply that the model implements a PMCA decrease that is specific to AD (e.g. last two lines of Abstract, Discussion as noted above). Minor Comments: 1) Abstract line 7, “relieving of” should be “relief of” 2) Line 57 (Introduction), and line 280 (Results). An “increased” or “higher” temporal decay means, to most readers, an increased rate of decay and therefore a decreased duration of calcium transients. This is the opposite of what you want to say here. In these conditions the temporal specificity is reduced and the durations of transients are increased. Thus the decay rate, or just the decay, is decreased or lowered. Reviewer #3: I thank the authors for the work they have done to satisfy all my comments and suggestions on the first draft. ********** Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: None Reviewer #2: Yes Reviewer #3: Yes ********** PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: No Reviewer #3: No [NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.] Figure resubmission: Reproducibility: To enhance the reproducibility of your results, we recommend that authors of applicable studies deposit laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols 22 Jul 2026 Submitted filename: Revision letter 2.pdf 20 Aug 2026 Dear Dr Joanna Jędrzejewska-Szmek, We are pleased to inform you that your manuscript 'Calmodulin controls spatial and temporal specificity of calcium-induced calcium release' has been provisionally accepted for publication in PLOS Computational Biology. Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests. Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated. IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript. Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS. Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Computational Biology. Best regards, Ausra Saudargiene, Ph.D. Academic Editor PLOS Computational Biology Andrea E. Martin Section Editor PLOS Computational Biology *********************************************************** Reviewer's Responses to Questions Comments to the Authors: Please note here if the review is uploaded as an attachment. Reviewer #1: Thank you for the work that improved the manuscript. Minor: Please, replace the typo 'dishomeostasis' by 'dyshomeostasis' throughout the manuscript. Reviewer #2: I am satisifed with these revisions, and believe the paper now makes a significant contribution to understanding mechanisms related to Alzheimer's disease pathology ********** Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available? The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified. Reviewer #1: None Reviewer #2: Yes ********** PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files. If you choose “no”, your identity will remain anonymous but your review may still be made public. Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy. Reviewer #1: No Reviewer #2: Yes: Paul Smolen PCOMPBIOL-D-25-02396R2 Calmodulin controls spatial and temporal specificity of calcium-induced calcium release Dear Dr Jędrzejewska-Szmek, I am pleased to inform you that your manuscript has been formally accepted for publication in PLOS Computational Biology. Your manuscript is now with our production department and you will be notified of the publication date in due course. The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript. Soon after your final files are uploaded, unless you have opted out, the early version of your manuscript will be published online. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers. For Research, Software, and Methods articles, you will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing. Thank you again for supporting PLOS Computational Biology and open-access publishing. We are looking forward to publishing your work! With kind regards, Sharmila Kamatchi PLOS Computational Biology | Carlyle House, Carlyle Road, Cambridge CB4 3DN | United Kingdom ploscompbiol@plos.org | Phone +44 (0) 1223-442824 | ploscompbiol.org | @PLOSCompBiol