Microglial﻿ signalling pathway deficits associated with the patient derived R47H TREM2 variants linked to AD indicate inability to activate inflammasome

The R47H variant of the microglial membrane receptor TREM2 is linked to increased risk of late onset Alzheimer’s disease. Human induced pluripotent stem cell derived microglia (iPS-Mg) from patient iPSC lines expressing the AD-linked R47Hhet TREM2 variant, common variant (Cv) or an R47Hhom CRISPR edited line and its isogeneic control, demonstrated that R47H-expressing iPS-Mg expressed a deficit in signal transduction in response to the TREM2 endogenous ligand phosphatidylserine with reduced pSYK-pERK1/2 signalling and a reduced NLRP3 inflammasome response, (including ASC speck formation, Caspase-1 activation and IL-1beta secretion). Apoptotic cell phagocytosis and soluble TREM2 shedding were unaltered, suggesting a disjoint between these pathways and the signalling cascades downstream of TREM2 in R47H-expressing iPS-Mg, whilst metabolic deficits in glycolytic capacity and maximum respiration were reversed when R47H expressing iPS-Mg were exposed to PS+ expressing cells. These findings suggest that R47H-expressing microglia are unable to respond fully to cell damage signals such as phosphatidylserine, which may contribute to the progression of neurodegeneration in late-onset AD.


Results
R47H het patient iPS-Mg exhibit loss of signalling capacity downstream of TREM2 activation. TREM2 signalling was activated by antibody crosslinking in myeloid precursor cells and induced downstream phosphorylation events. Thus, in TREM2 common variant (Cv, control) iPS-Mg lines, activation of TREM2 by 2 min of antibody crosslinking with F(ab') 2 fragments resulted in increased pSYK signalling compared with IgG2B isotype control. In contrast, activation of pSYK was not observed in R47H het iPS-Mg ( Fig. 1a,b). Additionally both pERK1/2 and pAKT signalling were activated downstream of TREM2 in Cv control iPS-Mg, (Fig. 1a,b) and again there was a deficit in the induction of these pathways in R47H het iPS-Mg, (Fig. 1a,b).

R47H het and R47H hom iPS-Mg show no deficit in phagocytosis of PS-expressing apoptotic cells or in secretion of soluble TREM2.
In order to further understand the implications of an R47H signalling deficit in microglia, we cultured iPS-Mg from control and R47H TREM2 het lines as previously described 28,29 and treated them with the TREM2 ligand PS, found on the outer membrane of apoptotic and necrotic cells 30 . The exposure of PS on the surface of heat-shocked SH-SY5Y was confirmed by Annexin-V FACS analysis (Supplementary Fig. 1i, ii). We have previously shown that iPS-Mg harbouring the NHD variant T66M showed a significant deficit in phagocytosis of dead cells 10 and we were able to repeat this under our current conditions (Supplementary Fig. 1iii). We found that in R47H het iPS-Mg, there was no significant reduction in uptake of Vybrant-DiI labelled PS+ cells compared with control iPS-Mg as determined by FACS analysis (Fig. 2a). We also tested this in iPS-Mg cultured from R47H hom (BIONi010-C7) compared with its isogeneic control (BIONi010-C), and determined that the R47H hom iPS-Mg also showed no deficit in their ability to phagocytose PS+ dead cells (Fig. 2a, Supplementary Fig. 1iv). We also analysed the secretion of soluble TREM2 (sTREM2) from unstimulated cells, as previously we showed that this was affected by NHD TREM2 variants expressed in iPS-Mg 10 . Basal secretion of sTREM2 in R47H het or R47H hom variant iPS-Mg was not significantly different across the iPS-Mg lines or compared with control iPS-Mg lines (Fig. 2b). Interestingly we found that when PS+ cells were added to iPS-Mg generated supernatant, this resulted in significantly reduced sTREM2 levels measured in the ELISA (Supplementary Fig. 1v) indicating that PS+ cells were able to "mop up" sTREM2. This was observed in the slight decrease in sTREM2 released following exposure of control iPS-Mg to PS+ cells, not seen in R47H het or R47H hom cells (Fig. 2b). R47H het and R47H hom iPS-Mg show a deficit in pSYK signalling in response to PS+ or DOPS liposomes. Since it is unknown whether an endogenous ligand for TREM2 such as PS can activate TREM2 signalling pathways in iPS-Mg, we investigated whether a 5 min stimulation with either PS+ cells or liposomes containing PS (DOPS; 1,2-dioleoyl-sn-glycero-3-phospho-l-serine) were able to induce pSYK signalling. Exposure of control iPS-Mg to PS+ cells or DOPS induced a significant increased expression of pSYK (Fig. 2c,d). However, as with TREM2 antibody stimulation, R47H het or R47H hom variant iPS-Mg showed no significant increase in pSYK in response to PS+ cells (Fig. 2c,d). This was not due to any differences in the expression levels of TREM2 in the difference cell lines 31 (Supplementary Fig. 1vi). R47H het and R47H hom iPS-Mg display deficits in pERK1/2 downstream of pSYK in response to PS+ or DOPS. Since ERK1/2 and AKT are allegedly activated downstream in the canonical signalling cascade from TREM2 via SYK 15 , we investigated whether the deficits in pSYK signalling observed in Fig. 2, were also observed for ERK and AKT signalling. We found that similar to SYK, pERK1/2 activation was reduced in R47H het iPS-Mg in response to PS+ cells and DOPS liposomes (Fig. 3a). This was confirmed in BIONi010-C control and R47H hom isogenic cells lines (Fig. 3b). Phospho-AKT was also induced by PS+ cells and DOPS, however in R47H het and R47H hom lines, there was no significant reduction in pAKT activation ( Supplementary Fig. 2i,ii). Therefore, we focused on pERK1/2 signalling downstream of pSYK.
We confirmed that pERK1/2 signalling activated by PS+ cells was indeed due to PS exposed on the cell surface; heat-shocked cells were pre-incubated with 5 µg recombinant Annexin-V for 1 h before the PS+ cells were added to iPS-Mg. We found that Annexin-V significantly reduced the pERK1/2 response to PS+ cells, indicating that PS is responsible for activating pERK1/2 (Fig. 3c). To confirm that ERK phosphorylation was downstream of pSYK we used pharmacological inhibition of SYK. The specific SYK inhibitor, BAY61-3606 (10 µM) significantly inhibited the subsequent PS-induced pERK1/2 activation (Fig. 3d). We thus investigated whether TREM2 signalling might impact on NLRP3 inflammasome-mediated caspase-1 activation. We examined whether there were differences in ASC speck formation and subsequent caspase-1 activation in control iPS-Mg compared with R47H iPS-Mg in response to PS+ cells. As a positive control for NLRP3 inflammasome activation, cells were treated with LPS, a classical initial priming signal, and subsequently with ATP, a second activating signal for ASC-speck formation and downstream NLRP3-mediated caspase-1 activation (reviewed in 32 ). We found that significant ASC-speck formation was induced by exposure to PS+ cells or LPS/ATP in control iPS-Mg, but that R47H het or R47H hom iPS-Mg showed no significant increase in ASC-speck formation above basal (Fig. 4a) www.nature.com/scientificreports/ Following downstream of ASC-speck formation and activation of the NLRP3 inflammasome, inactive procaspase-1 is subsequently cleaved to the active caspase-1. Measurement of caspase-1 activity with a Caspase-Glo®1 Inflammasome assay revealed that whilst control iPS-Mg showed significant caspase-1 activation in response to PS+ cells, R47H iPS-Mg showed a deficit in caspase-1 activation in response to PS+ cell stimulation (Fig. 4b). A specific caspase-1 inhibitor, YVAD-CHO was used to ensure that activation by PS+ cells was indeed attributable to caspase-1 ( Supplementary Fig. 3i). We further confirmed that the activation of caspase-1 by PS+ cells was www.nature.com/scientificreports/ downstream of NLRP3 and SYK activation with the NLRP3 inhibitor BAY11-7082 and SYK inhibitor PRT-060318 (Fig. 4c). These inhibitors also blocked caspase-1 activation by LPS/ATP treatment (Fig. 4d). Pre-incubation with the TLR4 inhibitor E5564 inhibited the LPS/ATP evoked caspase activation (Fig. 4d), but not the PS+ cellinduced caspase-1 response, (Fig. 4c), suggesting that PS+ cells signal to activate the NLRP3 inflammasome independently of TLR4 activation. In R47H het and R47H hom iPS-Mg, the inhibitors had no effect on caspase-1 activity, which showed no significant activation of caspase-1 with PS+ cells or LPS/ATP (Supplementary Fig. 3ii-iv).
We also investigated IL-1β secretion, since the inactive precursor of IL-1β is cleaved by caspase-1 (also termed interleukin-1 converting enzyme or ICE) and subsequently secreted; a deficit was observed in the form of reduced IL-1β secretion in R47H het and R47H hom iPS-Mg compared with controls in response to PS+ cells (Fig. 5a). Interestingly stimulation with LPS/ATP evoked IL-1β release from R47H het cells but not from R47H hom expressing iPS-Mg although the overall level of IL-1β secretion in response to PS+ cells or LPS/ATP was less in R47H het compared with Cv ( Fig. 5a) In the canonical inflammasome pathway, LPS induces transcription of inflammasome genes, including IL-1beta and NLRP3 in a priming step. Using qPCR to test whether PS+ cells also primed the iPS-Mg for NLRP3 inflammasome activation, we found that unlike LPS, neither IL-1beta nor NLRP3 was transcriptionally regulated by addition of PS+ cells (Fig. 5b,c). Furthermore, LPS stimulation was less effective at induction of inflammasome gene expression in R47H iPS-Mg compared with Cv (Fig. 5b,c). Genes such as the alternative inflammasome NLRP1 or CASP1, which has been shown not to be transcriptionally regulated, were not induced with PS+ or LPS stimulation (Fig. 5d,e). We also found that NFκB was upregulated by LPS, but not PS+, and again this was reduced in R47H iPS-Mg compared with Cv, suggesting that transcriptional responses are affected in R47H TREM2 cells that could ultimately impact on the signalling capacity of the cell (Fig. 5f).

ECAR and OCR deficits associated with R47H are modified by PS+. Activation of the NLRP3
inflammasome is strongly linked to alterations in cell metabolism 33,34 and we have shown previously that TREM2 variant iPS-Mg exhibit deficits in their ability to regulate cellular metabolism 29 . In order to further understand the reduced capacity of R47H TREM2 iPS-Mg to induce an NLRP3 inflammasome response, we investigated glycolytic capability and oxidative phosphorylation by Seahorse analysis of the extracellular acidification rate (ECAR) and oxygen consumption rate (OCR), respectively, in iPS-Mg after 24 h treatment with PS+ cells. As previously reported, glycolytic and mitochondrial stress parameters revealed deficits in glycolysis and oxidative phosphorylation (Fig. 6a,b). Glycolytic capacity, the maximum ECAR rate, was significantly reduced in R47H het iPS-Mg (Fig. 6c) and maximum respiration was reduced in R47H variant iPS-Mg (Fig. 6d). Interestingly, addition of PS+ cells rescued both the glycolytic capacity and maximum respiration in R47H het iPS-Mg, but not significantly in R47H hom although a similar trend was observed. These data suggest that the PS+ dead cells might act as an energy source that can replenish the metabolic deficiencies in R47H cells or provide lipids that feed into metabolic pathways.
We previously reported that iPS-Mg expressing the R47H variant showed a reduced maximal respiratory capacity and a reduced ability to undertake a glycolytic switch when preincubated with an inflammatory stimulus of LPS and IFNγ 29 . Here we found that pre-incubation with PS+ significantly enhanced glycolytic capacity and maximum respiration in R47H het expressing iPS-Mg to similar levels found in Cv cells (Fig. 6c,d). Glycolytic capacity and maximal respiration were not enhanced further in Cv and were not significantly rescued in R47H hom expressing iPS-Mg (Fig. 6c,d). Exposure of iPS-Mg to PS+ cells and then analysis of the cellular metabolic potential under stressful conditions (by the addition of FCCP), indicated that Cv iPS-Mg were able to increase their metabolic potential to stress when compared with CV with no PS+ cells present. The R47H variants were able to increase their metabolic potential under stressful conditions if primed with PS+ cells, but the potential was significantly below that observed in CV cells (Fig. 6e).

Discussion
Here we show the consequences of a reduced signalling response downstream of TREM2 in patient-derived iPS-Mg cells harbouring the R47H het TREM variant, and in an R47H hom line including reduced pSYK, pERK, and pAKT evident with either exposure to a PS+ ligand/DOPS liposomes or by antibody crosslinking. This reduced pSyk response is supported by findings in R47H-expressing BMDMS 5 and R47H-expressing 293 T cells 7 .
Whilst downstream phosphorylation events were reduced in R47H het/hom iPS-Mg, phagocytosis of PS+ cells and shedding of sTREM2 at baseline were not affected. A major role for microglia in the brain is to phagocytose dead cellular material and AD is characterised by extensive neuronal loss. Whilst a role for TREM2 in phagocytosis has been proposed in a number of animal and cell models 8,14,35 , there is however, conflicting evidence as to whether TREM2 is involved in the phagocytosis of dead neurons; thus Takahasi et al. 14 showed that shRNA knockdown of TREM2 led to reduced phagocytosis of apoptotic neurons whilst Wang et al. 8 reported that microglia from Trem2-/-mice showed no difference in phagocytosis of dead cells compared with wild-type. Our findings for R47H variant iPS-Mg are different from our previous report on the more severe NHD T66M-/-patient iPS-Mg where PS+ phagocytosis was reduced 10 , which we confirmed here. These cells do not process TREM2 to maturity and have no TREM2 on their surface whilst R47H het and R47H hom iPS-Mg do express mature TREM2, and these levels were not affected by PS+ exposure (Supplementary Fig. 1vi). These findings are likely to underlie the differences in phagocytic responses. Phagocytosis is driven by glycolysis in macrophages 36 and we reported previously that phagocytosis of Aβ was significantly affected in the R47H variant as well as T66M and W50C 29 . However, here we find that phagocytosis of PS+ is not affected by TREM2 variants, suggesting that the effects of TREM2 variants are highly selective. Taken together, these findings would suggest furthermore that K/O models of R47H more closely resemble the T66M variant microglial phenotype than the more subtle phenotype observed in R47H het or R47H hom microglia which both displayed no PS+ phagocytic deficit. Additionally, our www.nature.com/scientificreports/ current data suggest that the activation of signalling pathways through TREM2-PS interactions are independent of their ability to phagocytose dead cellular material. We sought to determine the consequences of the downstream signalling deficit coupled to R47H. Signalling from pSyk is linked to activation of the NLRP3 inflammasome and caspase-1 23,25 . A number of different stimuli acting via specific immunoreceptor tyrosine-based activation motif (ITAMs) can induce NLRP3 inflammasome activation in microglia or macrophages 23,37,38 and this includes dead cells 20 . Here we found a reduced ability of R47H variant iPS-Mg to activate components of the NLRP3 inflammasome in response to the TREM2 ligand PS+. Thus ASC speck formation, caspase-1 activation and IL-1β secretion, part of the canonical NLRP3 pathway (reviewed in 32 ) were reduced. Another cytokine, IL-18 is also induced upon NLRP3 activation but we were unable to detect the secretion of this cytokine (data not shown). Taken together, these findings were unexpected as TREM2 is suggested to be anti-inflammatory and thus an overall reduction in IL-1β release might be a beneficial response to dying neurons. However, transcript levels of the inflammatory cytokines IL-1β and IL-6 have been shown to be reduced in mouse models expressing mutant APP and PSEN1 in which TREM2 function was lost 39 . This study would be strengthened by the use of a TREM2 knock-out iPS-Mg cell line to establish a role for TREM2 in NLRP3 inflammasome activation and by using genetic rescue experiments to show that the effect of R74H TREM2 on can be reversed.
We also found that LPS/ATP was unable to induce the same degree of inflammasome response in R47H expressing microglia as Cv microglia, both at the level of priming (transcription of IL-1β and NLRP3) and ASC speck/caspase-1 /IL-1β release, suggesting a convergence in the signalling pathways of TLR4 and TREM2. Evidence of such interaction has been reported at the level of NFκB signalling which is upstream of NLRP3 activation, whilst down-regulation of TREM2 has been shown to inhibit release of inflammatory factors from LPS-stimulated microglia by inhibiting NFκB signalling pathway activity 40 . Interestingly NFκB did not show enhanced expression in Cv or R47H het variant iPS microglia exposed to PS+ (Fig. 5f) but was significantly increased following LPS stimulation in Cv only, suggesting that NFκB may not be involved in PS+ stimulated pathways in human iPS Mg.
Our present findings may reflect the effect of the deficit we identified in mitochondrial respiration and glycolysis in R47H-expressing iPS-Mg. Gene transcription requires energy and microglia typically undergo a switch to glycolysis from oxidative phosphorylation to enable this to occur 41 , and this may explain why LPS/ATP priming is also ineffective in R47H-expressing iPS-Mg. NLRP3 inflammasome activation is linked to alterations in cell metabolism 42 and an increased glycolytic rate facilitates this activation (reviewed in 43 ). NLRP3 activation can also modulate glycolysis 41 possibly by increasing glucose uptake driven by IL-1β 44 . We reported previously that microglia expressing the R47H variant are unable to undertake this glycolytic switch 29 . Interestingly, despite our finding that PS+ cells exposure in the R47H expressing cells can ameliorate some of the mitochondrial deficits, the microglia still showed signalling deficits in response to PS+, suggesting that the phagocytic pathway which we found not to be affected by R47H, may be primarily responsible for this amelioration. Therefore, in disease paradigms where lipid signalling is used to compensate for lack of energy, carrying an R47H variant may prevent 'efficient' use of the lipid energy source. Evidence suggests that TREM2 can regulate the expression of many genes influencing lipid transport such as ApoE. Our data suggest that R47H variant microglia may use PS+ cells to overcome metabolic dysfunction and in turn drive phagocytosis (reviewed in 45 ). This may be detrimental to nearby cells transiently expressing PS on their surface as ectopic PS exposure on live neurons can cause engulfment of distal neurites 46 and viable neurons (reviewed in 47 ). One limitation with the use of PS+ cells is that is becomes difficult to uncouple specific TREM2 activation from their ability to provide an energy source. Future work using cyclocreatine to provide extraneous ATP to the microglia without activating TREM2 signalling pathways or using specific TREM2 activating antibodies to achieve the opposite, will help determine how TREM2 signalling pathways directly impact the metabolic capacity of microglia and how this relates to disease progression.
Taken together our findings show that R47H variant microglia express a deficit in upstream and down-stream signalling responses, culminating in their inability to activate the NLRP3 inflammasome in response to a TREM2 ligand, PS, and a reduced ability to activate the inflammasome in response to a classic inflammatory/priming signal, both of which have ramifications for microglial responses in AD. (a) Left, western blot analysis of pERK1/2 signalling in iPS-Mg following 5 min stimulation with PS+ SH-SY5Y cells and DOPS liposomes in control and TREM2 R47Hhet patient lines. Right, quantification of pERK1/2 protein normalised to beta-actin; data show mean ± SEM, n = 5-13 cell lines from 4 individual experiments, ****P < 0.0001, *P < 0.05 (two-way ANOVA with Tukey's correction), ns, non-significant. (b) Left, western blot analysis of pERK1/2 signalling in iPS-Mg following 5 min stimulation with PS+ SH-SY5Y cells and DOPS liposomes in control and TREM2 R47H hom lines. Right, quantification of pERK1/2 protein normalised to beta-actin; data show mean ± SEM, n = 3-7 individual experiments, ***P < 0.001, *P < 0.05 (twoway ANOVA with Tukey's correction). (c) Left, western blot analysis of pERK1/2 signalling in iPS-Mg following 5 min stimulation with PS+ SH-SY5Y cells with or without pre-incubation of recombinant Annexin-V. Right, quantification of pERK1/2 protein normalised to beta-actin; data are mean ± SEM, n = 3 individual experiments, *P < 0.05 (two-way ANOVA with Tukey's correction). (d) Left, western blot analysis of pERK1/2 signalling in iPS-MG following 5 min stimulation with PS+ SH-SY5Y cells with or without treatment with pSYK inhibitor 10 μM BAY61-3606 in control and TREM2 R47H het patient lines. Right, quantification of pERK1/2 protein normalised to beta-actin; data show mean ± SEM, n = 7-8 cell lines from 4 individual experiments, ***P < 0.0001, *P < 0.05 (two-way ANOVA with Tukey's correction) ns, non-significant. Full length blots are presented in Supplementary Fig. 4.   28,29 . All iPSCs were maintained and routinely passaged in Essential 8 medium (Gibco). Karyotype analysis was performed by The Doctors Laboratory, London, UK 29 . Using our previously described protocol, iPS-microglia (iPS-Mg) were generated 28,29 . All patient R47H lines used here are heterozygous, and thus will be referred to throughout the paper as R47H het whilst the gene-edited line will be referred to as R47H hom .

DOPS liposome preparation and treatment of iPS-Mg. DOPS lipids (#840035C; Avanti Polar Lipids)
at 10 mg/ml in chloroform, were dried under a nitrogen gas stream in glass vials then stored under vacuum for a further 4 h prior to use. The dried lipids were hydrated with 250 µl pre-chilled PBS buffer in the range of 2-10 mg/ml and sonicated in a water bath for 30 min, maintaining a temperature of ~4 °C. The lipids were then processed using 21 passes through an Avanti extruder (#610000-AVL) with 100 nm polycarbonate membranes, until the solution turned from cloudy to clear. The liposomes were prepared at a 20 mM concentration in glass vials and kept at 4 °C for up to 2 weeks. Functional TREM2 agonist activity was confirmed by comparing luciferase induction in the GloResponse NFAT-RE-luc2 Jurkat Cell Line (Promega, CS176401) against the same line ectopically expressing full length human TREM2. 16-24 h before the assay, Jurkat cells were plated at 5 × 10 5 /ml in 10% serum medium. On the assay day Jurkat cells were harvested into 1% serum medium at 106 µl /ml and 25 µl was dispensed per well of a 384 well plate. A liposomes dilution range was dispensed at 2.5 µl per well in triplicate and luciferase activity was assessed after 16-24 h incubation. DOPS liposomes were added to iPS-Mg at 50 ng/ml for 5 min. SYK signalling was blocked by preincubation of iPS-Mg for 1 h with 10 µM BAY61-3605 prior to addition DOPS liposomes.
Immunoblotting. Immunoblotting of Syk and ERK signalling was carried out using conventional western blotting techniques. Briefly iPS-Mg were plated at 500,000 cells/well on 6 well plates and following treatment were lysed in modified RIPA buffer (25 mM Tris pH7.4, 50 mM NaCl, 0.5% NP-40, 0.25% Na doexycholate, 1 mM EDTA) containing 1 × Halt protease and phosphatase inhibitor cocktail (Life Technologies). Cell lysates  (Table 1), washed with PBS-T and incubated with corresponding LiCor compatible 680/800 nm conjugated secondary antibodies ( Table 1). The membranes were washed and visualised using an Odyssey detection system (LiCor). Protein bands were subsequently quantified using ImageJ software (www. imagej. nih. gov/ ij). www.nature.com/scientificreports/ Immunostaining. Immunostaining was carried out for the localisation of ASC specks. iPS-Mg were plated at 50,000 cells/well on glass coverslips in 24 well plates and following treatment, fixed with 4% paraformaldehyde (PFA) for 20 min, followed by PBS washing, permeabilisation with 0.2% Triton for 10 min, and block (2% bovine serum albumin, BSA) for 1 h at RT. Coverslips were incubated with primary antibody (

Cellular respiration analysis.
For the real-time analysis of extracellular acidification rates (ECAR) and oxygen consumption rates (OCR), iPS-Mg were plated and matured on Seahorse cell culture microplates and analysed using a Seahorse XFe96 Analyser (Agilent Technologies) as previously described 29 . iPS-Mg were incubated overnight with or without PS+ cells. Glycolytic stress kits were used to analyse cellular glycolysis and Mito stress kits were used to analyse mitochondrial respiration. Data were analysed using Wave v2.4.0.6 software (Agilent Technologies).

Statistical analysis.
Statistical analysis was carried out in GraphPad Prism using ANOVAs (one-way or two-way) with Tukey's multiple comparison tests from at least 3 independent experiments. All methods were performed in accordance with the relevant guidelines and regulations.
Ethics. MTAs  www.nature.com/scientificreports/  www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.