A number of tightly controlled and complex processes are performed by heterogeneous cell populations of the kidney, including blood filtration by microvascular endothelial cells and podocytes and secretion and re-absorption by proximal tubular epithelial cells. Acute kidney injury (AKI) is a frequent clinical event associated with a rapid loss of kidney function, leading to high morbidity and mortality [1] in up to 22% of hospitalised patients [2]. However, it is estimated that one-fifth of AKI that occurs after hospital admission is predictable and avoidable [3]. In light of the economic burden of 0.4% to 0.6% of the total healthcare costs, corresponding to £400 m - £600 m, annually spent on treatment for acute kidney injury in the UK alone [4], a clear need for development of means to predict and/or early detect and prevent/treat AKI arises.

It has been by now widely accepted that AKI represents a continuum or spectrum of diseases that could be identified at an early stage, rather than the previous terminology of acute renal failure describing an “all or nothing” condition [5]. The recognition of renal impairment at an early stage would allow for an immediate course of action to alleviate symptoms and disrupt the process of functional decline [6], however this implies that this condition is comprehensively understood on a molecular level to allow for targeted intervention therapies.

The factors causing AKI and their interaction with each other are still incompletely understood, which is surprising in light of the vast amount of scientific reports in this area. However, a systematic and in-depth mining of the literature, as employed by a comprehensive Systems Biology approach, coupled with high-resolution mass spectrometry analysis of diseased tissue could pave the way forward. Systems Biology, or integrative biology, uses descriptive terms to understand complex interactions based on the analysis of network behaviour and dynamic aspects. Here, we demonstrate a comprehensive analysis of AKI induced in a mouse model system to investigate and identify key signalling cascades, inferred metabolic alterations and molecules modulated in this highly complex and clinically relevant disease. Nevertheless it should be noted that animal models of AKI might not reflect this clinically critical condition observed in man, where an unobserved laps of mere hours between onset and catastrophic irreparable kidney failure can mean life or death. An acceptable approximation is the model chosen here, where an single overdose of folic acid causes the same end-effect of kidney damage observed in clinically manifested AKI.

We first established a high-resolution proteomic map of protein expression profiles using 6 sham injected controls and 6 folic acid overdosed mice. Folic acid nephropathy is a well established murine model of AKI using a single intraperitoneal injection overdose of folic acid and harvesting of kidneys after 24 hours [8]. Kidney cortex proteins were extracted, trypsin-digested, and analysed by high-resolution liquid chromatography mass spectrometry (LC-MS/MS). After merging the individual independent mass-spectrometric analyses, 6564 non-redundant proteins were detected (Additional file 1: Table S1, and detailed information of data processing can be found in Additional file 2). 14 of these proteins were detected only in control samples, 1050 only in AKI, and 5500 in both sets of conditions. Statistical testing using Mann–Whitney showed that 2521 proteins were statistically significant (p-values <0.05), and when also taking fold-change calculations into considerations then the dataset indicated 1480 molecules to be significantly changed in AKI (p-values <0.05 and fold changes >2). 420 of these proteins were found in both control and AKI samples, 14 only in controls, and 1046 molecules only in AKI samples. After initial database matching and employing gene ontology clustering to assess data quality and overlapping characteristics, the data were mapped onto existing general pathway maps to identify affected cascades and modulatory events. More than 50 signalling cascades (see Additional file 2), mainly due to overlapping features in the existing maps, were initially identified to be of interest in further downstream analysis based on their potential functional relevance. In parallel, we assembled potential pathway maps derived from the pertinent AKI-focussed literature. Through a manual iterative feedback analysis and deep-mining of inferred and prior knowledge (Figure 1), extending the pathway analysis beyond the reported AKI-signalling cascades, and molecule-by-molecule pathway delineation and manual feature look-up, we assembled plausible signalling cascades which led to a combined pathway map (Additional file 3: Figure S1). The same approach was also used to map the metabolic pathways modulated in AKI, as well as probable gene activation cascades based on reported modulation of transcription factors such as NFκB, which were integrated into the AKI model.

Key pathways which were previously reported to be main events in AKI induction, such as the Renin-Angiotensin Aldosterone System (RAAS) and involvement of the TNFα-signalling cascade, could be confirmed in our analysis to be up-regulated. Upstream activators, such as the kallikrein and cathepsin systems, were also induced, suggesting a specific activation of the aforementioned RAAS axis. Additional up-regulated modulatory events, which were per se not associated with the currently perceived molecular model of AKI are the glutamatergic signalling cascades and associated calcium-flux pathways, which have a major detrimental effect on both apoptosis and necrosis. A substantial amount of information is available about glutamate-dependent pathways and signalling events in a non-renal, specifically neurological context, and it was surprising to encounter a considerable level of glutamatergic pathway elements associated with renal dysfunction. The specific involvement under physiological conditions of ionotropic as well as metabotropic glutamate receptors in kidney is currently unknown, however a dysfunction, such as over-stimulation and –activation is expected to lead to the same effects observed in other systems, e.g. uncontrollable calcium-influx and ultimately cell death. This observation is further acerbated by an apparent simultaneous induction of the calcium-flux machinery, involving the calcium-import and –export channels, such as calcium pumps (SERCA and PMCA) as well as ryanodine receptors and calcium-sensitive modulators. A potential assembly of signalling events originating from the RAAS axis and involving the most prominent glutamate-sensitive calcium-channel NMDA receptor is depicted in Figure 2B. As shown, signalling from the renin-induced angiotensin receptor leads to a cascade of known signalling and induction events involving PLC2β, PKCα, Ras, RalA, p38kinase, MSK and activation of the transcription factor SP1. The latter promotes gene activation of the NMDA receptor GRIN1 which however is also dependent on SP3 inhibition. SP3 inhibition can be induced by oxidative stress, reflected in this scheme in the stimulation of NADPH oxidase NOX, predicted to result from the activation of the up-stream kinase PKCα. Oxidative stress also stimulates the death-associated protein kinase Dapk1, which can target the NMDA receptor and leads to a permanent-open state of the channel, allowing calcification of the intracellular environment. This in turn triggers the mitochondrial- as well as nuclear-based apoptotic and necrotic molecular machinery.

To enhance these hypotheses and predictions, we validated specific key elements in this delineated pathway by immunohistochemistry, as shown in Figure 2A. We also tested whether the potentially anti-apoptotic protein Hsp27 is regulated according to the MS-results. This protein is involved in cytoskeletal remodelling, which might also be an involvement of the NMDA receptor under physiological conditions, however based on our pathway model this would imply that apoptosis is to some extent inhibited, whereas necrosis is not. All of the proteins tested, namely PLC2β, Ras, MSK, Grin1, Dapk1 and Hsp27, could be confirmed to be up-regulated, as measured by the proteomic mass-spectrometry approach, thereby further emphasizing the importance of this downstream pathway in the development of the AKI phenotype. Unsurprisingly we could not detect any indication of the involvement of the insulin pathway, either up- or down-stream of the insulin receptor, but instead observed an up-regulation of known inhibitors of insulin-signalling such as Socs3 and Ptprf.

Additionally, our analysis of the glycolysis and glycogenolysis pathways, which are modulatory targets of insulin-signalling, shows a significant up-regulation of enzymes involved in glucose and fructose release, and no indication of an equally activated down-stream metabolism from Fructose-1,6-bisphosphate to pyruvate or lactate. This suggests an accumulation of Fructose-1,6-bisphosphate, which is a known allosteric inhibitor of Adss, an enzyme in the de novo pathway of AMP production. This inhibition would lead to an accumulation of IMP, and in conjunction with the observed up-regulation of 5′(3′)-deoxyribonucleotidase Nt5c, to inosine and ultimately hypoxanthine and xanthine build-up. Those last two stages are a characteristic hallmark of AKI, thereby explaining the previously described observations of an accumulation of the latter two chemical compounds on a molecular level (see Table 1).

AKI has a major impact on survival in hospitalised patients undergoing extensive surgery [9], and a large number of studies have been undertaken to understand the molecular causes leading to and being involved in this complex disease [10, 11]. Here, we employed a non-biased systematic approach using a combination of Systems Biology, proteomics and de-novo pathway mapping and immunohistochemical validation to elucidate major pathways and signalling cascades in order to contextualise previous findings and uncover novel potential targets for pharmacological intervention. The analysis focuses mainly on very significant changes observed in AKI covering all the known hallmarks using the folic acid-overdose animal model and can be regarded as a first attempt to describe this condition in a molecular mechanistic way.

An initial finding of this mapping effort was the apparent complexity of AKI and associated pathways, where a plethora of signalling cascades appears to be modulated simultaneously. This could be due to many kidney cell types being in an active state of inflammation signalling, apoptosis induction, stress, and other modulatory events, as also supported by the literature [12]. As many cascades can be activated in more than one way, and induction of AKI can occur via multiple stimulation or entry points [13], it appears plausible that these pathways are merging to similar down-stream targets leading to the observed deleterious events in kidney injury.

Our analysis revealed that a major pathway involved in AKI is the RAAS axis, which has been reported multiple times before [14, 15], further confirming the validity of the approach. Renin activation can occur in several ways, either by kallikrein, cathepsin or other stimuli as indicated in Additional file 3: Figure S1. Many of those activating proteins were markedly up-regulated in AKI.

A further main crucial initiation step leading towards AKI is based on up-stream activation via TNFα, where inhibition of this molecule prevents apoptotic cell death [16]. Mapping of modulated proteins found in this study clearly indicates the involvement of this cascade in AKI. TNF signals via Jnk to inhibit phosphoinositide 3-kinases (PI3Ks), which in turn leads to the release of cathepsin B-containing lysosomes and to renin activation [17], as well as activation of the pro-apoptotic protein p53 [18]. TNFα may also potentially contribute to RAAS activation as shown in Figure 3. Mounting evidence suggests that the RAAS plays a major role in kidney injury and inflammatory processes [19], whereas TNFα and TNF-like cytokines are involved in induction of cellular responses such as inflammation as well as induction and progression of apoptotic cell death [20].

Along these lines, the data obtained also suggest up-regulation of inflammatory pathways, indicated in our analysis to involve activation by the RAAS axis through the NFκB pathway, leading to downstream signalling by interleukins (see Figure 3). These results are supported by an observed role for TLRs in promoting inflammation and tissue injury in AKI, which synchronise their inflammatory signal activations via NFκB [21]. The activation of the RAAS axis also leads to the activation of a number of transcription factors [22], and our results indicate gene activation cascades involving among others NFκB, PPARγ, SMAD and HIF1α.

The significant changes in levels of e.g. collagens, Ena/VASP, Hsp27/Hspb1 and proteases such as MMP2 were interpreted by the bioinformatic tools to be correlated with cytoskeletal remodelling, extracellular matrix (ECM) degradation, and associated cell detachment, all linked to kidney injury [23]. The cytoskeletal reorganisation by Hsp27 has been shown to inhibit fibrogenesis in obstructive nephropathy [24].

The marked up-regulation of the molecular clusters associated with apoptotic pathways, including NADPH oxidases (NOX) and P450 monooxygenases, nitric oxide synthase Nos1, calpains Capn3 and 11, caspase Casp8, poly-[ADP-ribose]-polymerase Parp3, 6, and 12 and Serine-protein kinase ATM, indicates activated cell-death cascades linked to prolonged hypoxia. Vasoconstriction, one of the hallmarks of AKI, is directly responsible for hypoxia, which stimulates several proteins such as NOX and P450 monooxygenases and leads to ROS generation, including superoxide and hydrogen peroxide production by xanthine oxidase as a result of metabolic hypoxanthine accumulation [25]. This is acerbated by the notion that kidney injury can also induce the over-production of nitric oxide (NO), which, with superoxide, forms peroxynitrite, and causes cell damage via DNA fragmentation [26]. Over-production of ROS is linked to the induction of apoptosis and necrosis, and a number of pro-apoptotic and necrotic pathways have been shown to exist, some of which also require the activation of the mitochondrial calpain system through sustained Ca²⁺-influx [27]. The classical apoptotic pathway involves the release of mitochondrial proteins, leading ultimately to activation of caspases [26]. Regulated necrosis also contributes to cell loss [28] (Figure 3, Additional file 3: Figure S1 and Table 1).

A major novel finding is the modulation of proteins involved in glutamatergic signalling. This is exemplified by the up-regulation of key components of the glutamate response system, such as the metabotropic glutamate receptor Grm1 (also known as mGluR1), and ionotropic glutamate receptors Grin1 (NMDA receptor NR1), Gria2 (AMPA receptor GluR2), and specific scaffolders of these receptors, as well as known down-stream signalling molecules (e.g. CamK2b, PP2B, and Nos1) (Additional file 3: Figure S1). Activation of glutamatergic signalling components is one possible route to activate Ca²⁺-flux pathways [29]. Podocytes and tubular cells express glutamate receptors [30] including functional Grin1 NMDA receptors [31], raising the possibility that neuron-like signalling contributes to glomerular function [32]. Glutamatergic signalling driven by podocytes contributes to the integrity of the glomerular filtration barrier and may regulate glomerular filtration [33]. Additionally, one of the metabotropic glutamate receptors, Grm1, is expressed in glomerular podocytes [34]. Although the stimulus for increased NMDA receptor expression in AKI has not been characterised, hypoxia is known to induce the transcription of glutamatergic signalling components such as the NMDA receptors in non-renal cells through inhibition of the transcription factor SP3. SP3 inhibits transcription by occupying the genomic translation initiation site, and activation of SP1 which in turn occupies the freed site and promotes the transcription of target genes [35].

Collectively, based on the presented data, we could hypothesize that excessive activation or sustained activation of NMDA receptors in tubular cells or podocytes promotes oxidative stress, mobilization of cation channels, disproportionate Ca²⁺ influx and overloading, excessive ROS generation, and apoptotic cell death [36] similar to what was shown to occur in the neurological environment [37]. Indeed tissue glutamate is high in AKI [38] and pharmacologically blocking this receptor with D-AP5, a synthetic glutamate analogue which binds to the glutamate-binding site of the NMDA receptor and suppresses pore opening, prevented tissue injury [39]. Shown in Additional file 3: Figure S1 are proteins and signalling events, assembled to depict probable cascades driving both, apoptosis and necrosis in kidney tissue after injury. The glutamate receptor Grin1/NMDA-R1, which is responsible for cellular Ca²⁺-overload leading to irreversible caspase activation and apoptosis, in conjunction with sustained ROS production, and ultimately energy-depletion and necrosis, may be a major contributor to these events.

Dapk1, a gene activation target of p53, can mediate the pro-apoptotic activity of TNFα by inhibition of NFκB signalling [40]. Additionally, the NMDA receptor is a known target of Dapk1, where Dapk1-mediated channel modulation results in a permanently open NMDA receptor, leading to cell death [41]. Remarkably, we also observed an orchestrated up-regulation of relevant scaffolders (Homer1/3, Dlg1, and Dlgap4) which link the glutamate receptors directly to downstream signalling cascades ranging from Ca²⁺-signalling to phospholipase and adenylate cyclase cascades and interlinking pathways. This further supports the importance of these receptors in AKI.

Other Ca²⁺ channels, which we observed to be up-regulated in AKI, and that could potentially contribute to Ca²⁺-overload are Ca²⁺-import channels such as voltage-sensitive calcium channels VSCC (isoforms Cacna1g/h and Cacnb1) and transient receptor potential channels TRP (isoforms Trpc4, Trpm1/3), intracellular release channels sarcoplasmic/endoplasmic reticulum calcium ATPases SRCA (isoforms Atp2a1/2) which are dependent on ATP hydrolysis, and others (Slc8a3, Ryr1). Simultaneously we also detected up-regulation of Ca²⁺-efflux channels plasma membrane calcium-transporting ATPases PMCA (isoforms Atp2b1/2/3), which are dependent on available ATP to function. The depletion of intracellular ATP pools could therefore lead to an asymmetric Ca²⁺-flux and exacerbate the intracellular calcification even further.

We also extracted supportive information from the literature, where pharmacological or other intervention studies reported modulatory effects on kidney injury. The results are summarised in Additional file 4: Table S2, and also indicated in Figure 3. Inhibition of molecules downstream of TNFα and involved in ROS production and hyperglycaemia, as well as specific activation of NFκB cascades, all had an improving or attenuated outcome on AKI prevention. Of particular interest is the inhibition of TNFα signalling by using TNFα antibodies which resulted in prevention of AKI [16]. Inhibition of glutamate signalling via the NMDA receptor, by using the channel blocker D-AP5, has been shown to significantly reduce ischemia/reperfusion injury (I/RI) -induced glomerular and tubular dysfunction [39], and NMDA-induced injury, which leads to calcification of the intracellular space, was shown to be reduced by using calcium antagonists nimodipine and dantrolene in an animal model of retinal injury [42]. Nimodipine targets and blocks L-type calcium channels, whereas dantrolene inhibits ryanodine receptors. Nimodipine was shown to alleviate tubular cell necrosis in transplantation-induced AKI [43]. Dantrolene in particular gained interest as it was shown to be beneficial in protecting against I/RI in animal models of heart-, brain- and potentially liver-induced I/RI [44]. However, dantrolene was shown to be ineffective in protecting renal function [45]. The same study investigated two other calcium-channel inhibitors, namely nicardipine, which blocks L-type calcium channels such as CACNB1, and TMB-8, blocking inositol 1,4,5-trisphosphate receptors. They could demonstrate that both compounds applied in pre-conditioning prior to I/RI had a positive effect both in vitro on tubular cells by decreasing the apoptotic effect, and in vivo in a rat renal model in kidney functional integrity. The results of those pharmacological studies listed in Additional file 4: Table S2 suggest that AKI in principle can be prevented or its effect diminished, and a combination of various drugs, targeting specific AKI-induced pathways and molecules, could be a way forward in disease prevention or therapeutic intervention.

A potential synopsis of pathways and signalling events evoked during AKI leading to Ca²⁺-overload and apoptosis as well as necrosis is shown in Figure 4, and a focus centred around the NMDA receptor is shown in Figure 2B. The relevant endogenous NMDA receptor ligand in AKI has not been characterised, however kidney glutamate levels are increased in AKI [38]. To further substantiate this cascade, we validated the up-regulation of key molecules involved in this axis as by immunohistochemistry shown in Figure 2A. The signalling cascade downstream of angiotensin-receptor 1 activates PLC2β which signals to PKCα. Both these proteins were up-regulated in AKI based on the mass-spectrometric analysis and confirmed for PLC2β by immunohistochemistry. PKCα activation leads to ROS production via NOX [46], and induces gene activation of the NMDA receptor, which we found to be up-regulated by both techniques used. This gene induction step also requires activation of MSK, which we also found to be up-regulated. Additionally, we could show that Hsp27 is up-regulated, indicative of cytoskeletal reorganisation, and we could demonstrate that Dapk1 is also up-regulated. Hsp27 up-regulation may protect from angiotensin II- and high-glucose-induced apoptosis [47]. All these events are potentially linked to both apoptosis and necrosis, and through this up-regulated pathway could be demonstrating a probable route of tubular epithelial fate in AKI.

The tissue proteomics approach identified here may eventually be optimized for its use in diagnostic/staging clinical pathology and the molecular fingerprint identified may be used to explore different molecular types of clinical AKI that may require different therapeutic approaches.

In summary we assembled a global pathway model for AKI based on proteomic changes, and further support the concept that the approach of in-depth Systems Biology coupled with high-resolution proteomic analysis represents a valuable approach to elucidate complex pathways and systems, especially in multifaceted diseases such as AKI. This combined methodology can be applied to search for druggable targets and identify key nodes where molecular/pharmacological interventions are most likely to have an impact on the system as a whole, as well as predictions for disease biomarkers. One such target could be the glutamatergic system, however the NMDA receptor itself might not be an ideal lead due to its involvement in neurotransmission in other areas like cardiac function, but a renally localised combinatorial pharmacological intervention, blocking both glutamatergic signalling and Ca²⁺-flux could be of beneficial effect.