Differences in intracellular localisation of ANKH mutants that relate to mechanisms of calcium pyrophosphate deposition disease and craniometaphyseal dysplasia

ANKH mutations are associated with calcium pyrophosphate deposition disease and craniometaphyseal dysplasia. This study investigated the effects of these ANKH mutants on cellular localisation and associated biochemistry. We generated four ANKH overexpression-plasmids containing either calcium pyrophosphate deposition disease or craniometaphyseal dysplasia linked mutations: P5L, E490del and S375del, G389R. They were transfected into CH-8 articular chondrocytes and HEK293 cells. The ANKH mutants dynamic differential localisations were imaged and we investigated the interactions with the autophagy marker LC3. Extracellular inorganic pyrophosphate, mineralization, ENPP1 activity expression of ENPP1, TNAP and PIT-1 were measured. P5L delayed cell membrane localisation but once recruited into the membrane it increased extracellular inorganic pyrophosphate, mineralization, and ENPP1 activity. E490del remained mostly cytoplasmic, forming punctate co-localisations with LC3, increased mineralization, ENPP1 and ENPP1 activity with an initial but unsustained increase in TNAP and PIT-1. S375del trended to decrease extracellular inorganic pyrophosphate, increase mineralization. G389R delayed cell membrane localisation, trended to decrease extracellular inorganic pyrophosphate, increased mineralization and co-localised with LC3. Our results demonstrate a link between pathological localisation of ANKH mutants with different degrees in mineralization. Furthermore, mutant ANKH functions are related to synthesis of defective proteins, inorganic pyrophosphate transport, ENPP1 activity and expression of ENPP1, TNAP and PIT-1.

with destructive arthritis and may mimic rheumatoid arthritis, gout or osteoarthritis and is the commonest form of inflammatory monoarthritis in the elderly, occurring in up to 40% of those over 65 years of age 1,5 . In contrast, CMD is a rare disorder characterised by hyperostosis/sclerosis of the skull and abnormal modelling of the long bones, and individuals with severe forms of CMD can have reduced life expectancy as a result of compression of the foramen magnum 4 . CMD is associated with decreased ePPi, which allows increased HA deposition and altered bone modelling via chondrogenesis, osteoblastogenesis and osteoclastogenesis 6 . There is currently no specific treatment for CPPDD and CMD.
Mutations near either end of ANKH are mostly associated with CPPDD while mutations in the middle have been reported to cause CMD, though their biological effect and cellular function remain largely unexplored. Previous research shows CPPDD associated P5L (p.Pro5Leu, NM_054027.4:c.14 C > T) to increase expression of ANKH and was reported to increase expression and activity of ENPP1 7,8 . E490del (p.Glu490del, NM_054027.4:c.1468_1470delGAG) deregulated TNAP activity 9,10 . CMD related mutants have largely been restricted to clinical case studies. One case report of S375del (p.Ser375del, NM_054027.4:c.1123_1125delTCC) showed a decrease in ePPi that was consistent with the predicted loss-of-function 11 . G389R (p.Gly389Arg, NM_054027.4:c.1165 G > A), reported in several cases of CMD and recently in CPPDD, where it was predicted to be a loss-of-function variant 4,12 .
Autophagy is a dynamic catabolic mechanism that recycles damaged organelles and non-functional proteins and maintains cellular homeostasis 13 . Previous studies have highlighted the importance of autophagy and its modulation of genes in maintaining healthy chondrocytes in the formation of cartilage and preventing degeneration during osteoarthritis 14,15 .
To investigate the pathogenic mechanisms of ANKH in CPPDD and CMD, we generated four disease-associated ANKH mutants associated with relatively severe clinical phenotypes: two are terminally positioned P5L and E490del associated with CPPDD, two are centrally positioned S375del and G389R associated with CMD. We used confocal imaging to identify ANKH mutant cell localisation dynamics, measured ePPi concentrations and altered mineralization level, evaluated ANKH mutant effect on the function of ENPP1 and gene expression of ENPP1, TNAP and PIT-1. We also investigated the involvement of autophagy for potential mutated ANKH protein recycling in the pathogenesis of CPPDD and CMD.

Results
We found that ANKH mutations altered cellular localisation dynamics and led to biochemical changes at different levels by comparing with wt.ANKH. Our detailed findings are summarised in Table 1  www.nature.com/scientificreports www.nature.com/scientificreports/ human adult fibroblasts (HAF), adenocarcinomic human alveolar basal epithelial cells (A549), HeLa and monkey Cos7 cells (Figs. 1A, S1) 16,17 . We observed this specific cell membrane localisation in the majority of transfected cells at all-time points after transfection, unlike any other mutant. Further clarification using 3D confocal Z-stack rotating rendering also showed the localisation was primarily found in the cell membrane (Movie S1). CH-8 cells displayed auto-fluorescence that masked overexpression of control eGFP and wt.ANKH-eGFP ( Fig. S2 and Movie S2). Overexpressing wt.ANKH in HEK293 cells increased ePPi following 48 hours after transfection (Fig. 1B). Interestingly, wt.ANKH overexpression in CH-8 cells significantly decreased mineralization compared to GFP only control (Fig. 1C). Wt.ANKH in HEK293 cells showed gene expression of ENPP1 significantly increased at 48 and 72 hours, peaking at 48 hours (Fig. 1D), this pattern was mirrored in the ENPP1 activity of CH-8 cells (Fig. 1G). A similar, but non-significant trend, was measured for TNAP at 24, and 72 hours after P5L ANKH slowed initial membrane recruitment and increased ePPi and mineralization via increased ENPP1 activity. We observed an initial diffuse expression of P5L within the cytoplasm with perinuclear aggregation and formation of vesicle like structures at 24 hours, with only 34% of HEK293 cells displaying specific cell membrane localisation, but returned to wild-type levels at 48 and 72 hours ( Fig. 2A). P5L significantly increased ePPi double that of wild-type control at 72 hours in HEK293 cells and mineralization increased 20% in CH-8 cells (Fig. 2B,C). It also increased TNAP in HEK293 cells but did not change PIT-1 (Fig. 2E,F). In CH-8 cells we observed insignificant increases in ENPP1 expression, but ENPP1 activity as consistently high at all-time points (Fig. 2D,G).
S375del completely abolished ANKH localisation to the cell membrane, decreased ePPi and increased mineralization. S375del completely rendered ANKH unable to enter the HEK293 cell membrane and remained diffusely expressed within the cytoplasm at 24, 48 and 72 hours after transfection Fig. 3A). We measured a trending decrease in ePPi at 72 hours in HEK293 cells and observed a significant 40% increase in mineralization in CH-8 cells compared to wt.ANKH (Fig. 3B,C). In HEK293 cells we measured non-significant increases in ENPP1 and ENPP1 activity in CH-8 cells was found to increase at 24 hours but decrease at 48 hours ( Fig. 3D,G). PIT-1 increased at 48 hours ( Fig. 3F).
G389R slowed initial ANKH to insert in the membrane and trended to decrease ePPi, increased mineralization and was potentially targeted as a non-functional protein. G389R expressed mostly within the cytoplasm with just 28% of HEK293 cells displaying specific cell membrane localisation at 24 hours, half that of the wild-type control, but show similar localisation after 48 hours, though there was novel vesicle like puncta (Fig. 4A). In HEK293 cells we found a non-significant decrease in ePPi at 48 and 72 hours. In CH-8 cells we measured small, though significant, increases in mineralization (Fig. 4B,C). G389R overexpressed in HEK293 cells was found to significantly increase of over two-fold ENPP1 early compared to wildtype at 24 hours (Fig. 4D), increased ENPP1 activity in CH-8 cells at 24 hours was also seen dropping from 48 hours (Fig. 4G). We also observed co-localisation of G389R-eGFP with RFP-LC3 under autophagic and rescue conditions in HEK293 cells (Fig. 4H, zoomed and intensity over distance plots and Movie S3). We did not observe co-IP of LC3 with mutant G389R (Fig. 4I).
E490del mutated ANKH remained mostly cytoplasmic and formed punctate like structures that co-localised with LC3, and increased mineralization through increased expression of ENPP1, TNAP and PIT-1. E490del localised primarily to the cytoplasm at all time points, showing less than half of the membrane expression after 72 hours of wild-type control, and like G389R formed vesicle like puncta after 48 hours (Fig. 5A). The ePPi slowly increased over time at similar levels to wild-type with no significant difference. In CH-8 cells we observed a significant 15% increase in mineralization (Fig. 5B,C). Strikingly E490del increased expression of ENPP1 by over five-fold compared to wt.ANKH in the HEK293 cells, reflected by a five-fold increase in ENPP1 activity in CH-8 cells (Fig. 5D,G). This early expression at 24 hours is also seen in the two-fold increases of TNAP and PIT-1 (Fig. 5E,F). Similar to G389R we observed co-localisation of E490del-GFP with RFP-LC3 under autophagic and rescue conditions (Fig. 5H, zoomed and intensity over distance plots and Movie S4). We did not observe co-IP of LC3 with E490del (Fig. 5I).

Discussion
ANKH function at the cell membrane to transport PPi into the ECM is well established, however. We initially show ANKH localisation to the cell membrane to be non-homogenous and a striking distinction between the localisation dynamics of each ANKH mutants. P5L and G389R both localised more slowly to the cell membrane, E490del showed some cell membrane localisation but was mostly cytoplasmic and S375del did not at any time. A recent paper on novel mutant ANKH-F377del, also shows evidence of the mutated protein to be dispersed in the cytoplasm 18 . Although we see similar wt.ANKH localisation as reported by Kanaujiya, J. et al. 18 and many other non-bone cell lines, we cannot discount the effects that this may be specific to bone. These very different localisation dynamics, even between mutants causing similar clinical features, highlights the complex and diverse function of ANKH in the pathogenesis of CPPDD and CMD.
We quantified each ANKH mutant effect on ePPi concentration and potentially non-physiological, mineralization, and as expected overexpression of wt. ANKH increased ePPi level compared to endogenous ANKH, but the CPPDD linked gain-of-function P5L caused a remarkable increase of over double ePPi compared to wt.ANKH once a majority of the mutant had inserted into the cell membrane at 72 hours, suggesting a slowed response to ANKH changes in the pathophysiology, or even limited with sustained mutation expression. Conversely, both CMD linked loss-of-function S375del and G389R trended to decrease ePPi. In the ECM ePPi plays a key role in inhibiting HA accretion and bone mineralization 2,7 . In CMD, there is thought to be abnormal skeletal modelling due to excessive HA mineralisation because of the decrease in ePPi 19 , a phenotype confirmed by our novel molecular data. In contrast, increased ePPi leads to excessive CPPD mineralization in the ECM of those suffering with CPPDD. Our data highlighted that all tested mutants had pathologically increased mineralization. Furthermore, the quantity of mineralization ranks the mutations in their severity of clinical phenotypes with S375del being the most severe, second P5L, third E490del and fourth G389R. Taken together, our data suggest a direct and novel link between differential cellular localisation, altered ePPi concentration and mineralization associated with mutated forms of ANKH.  ePPi and discerned the complexity of these relationships, especially in the context of disease progression and causality. Increased ENPP1 and PIT-1 of wt.ANKH corresponded to the increase in ePPi, and the reduction in TNAP was concordant with decreased mineralization. Therefore, simply increasing the quantity of ANKH did  www.nature.com/scientificreports www.nature.com/scientificreports/ not replicate a gain-of-function phenotype. This is likely due to saturation of the ANKH -ENPP1 interaction, where ENPP1 is thought to provide most of the ePPi present in the ECM [24][25][26] . The decrease in mineralization can be attributed to the mild increase in ePPi, enough to further inhibit HA accretion but insufficient for CPPD formation.
P5L increased expression of TNAP at 24 hours as a response to increased ePPi, but not sustained; however the level of ENPP1 and PIT-1 were similar to wt.ANKH. Interestingly, we measured an increase in ENPP1 activity at 24, 48 and 72 hours, and therefore we suggest that P5L raised ePPi via increased ENPP1 activity rather than increased ANKH function or increased ENPP1 expression -enlightening the previous observation showing P5L to increase activity of ENPP1 8 . Furthermore, the increased ePPi and non-changed PIT-1 indicated that the increased mineralization caused by P5L was specific to CPPD formation rather than HA. Inability of S375del to insert into the membrane was so severe that even increased expression of ENPP1, TNAP, PIT-1 and corresponding ENPP1 activity as compensation was not enough to rescue the phenotype. S375del showed by far the largest mineralization profile and decrease in ePPi over time indicated the mineralization to be likely HA rather than CPPD. In the long term, reduced ANKH membrane presence would exacerbate a condition reliant on extracellular ePPi control and this explains the severe phenotype. The milder phenotype G389R increased expression and corresponding activity of ENPP1 but did not change TNAP and PIT-1. Interestingly, as G389R was recruited to the cell membrane similar to wt.ANKH but still increased ENPP1 activity and expression as compensation for decreasing ePPi levels, this suggested a severe reduction in transport capability for G389R but needs further exploration. E490del drastically increased early expression of ENPP1, TNAP and PIT-1 and corresponding ENPP1 activity as compensation for lack of ANKH at the cell membrane, however as this mutant did not completely abolish ANKH localisation and hence, the compensatory effect seemed to be overpowered and might indeed be causal in the pathogenesis of CPPDD in sufferers affected by the E490del mutation, requiring further study.
Abnormal or defective proteins are often targeted by autophagic machinery for recycling 27 . Co-localisation of the autophagy marker LC3 with E490del under normal, autophagy induction and rescue conditions, and with G389R under autophagy induction and rescue conditions could suggest that these mutated proteins were targeted as non-functional or defective proteins. We also checked for direct physical interaction of ANKH with LC3, which may indicate a role in the autophagy process, however we did not observe co-IP of these mutants with LC3, making it more likely that G389R and E490del were targets of autophagy as non-functional or defective proteins rather than facilitators in the autophagy process. The absence of co-localisation for P5L and S375del does not suggest they are not recycled as this process is not limited to autophagy, for example an endoplasmic-reticulum-associated protein degradation (ERAD) ubiquitination pathway that is preferential for insoluble aggregates within the cytoplasm could explain no co-localising of S375del with LC3 27 .
We summarize the possible mechanisms in the pathogenesis of each tested mutants associated with CPPDD or CMD as follows: P5L. This proline to leucine mutation may delay ANKH membrane uptake and the longer term increasing in ePPi could be attributed to an overreaction from increased ENPP1 activity rather than ENPP1 expression directly. This is likely that the rise in ePPi is not due to increased ENPP1 amounts at the cell membrane but rather to increased activity of ENPP1 that is already at the cell membrane facilitated by the P5L mutation as suggested by previous work and now also our ENPP1 activity assay. Since TNAP expression increased initially and has a role in reducing excess ePPi, we suggest that there should be a feedback mechanism to upregulate TNAP under conditions of high ePPi but cannot compensate for the mutation effect.
S375del. Our data suggests that deletion of serine at residue 375 completely abolishes ANKH localisation to the cell membrane and is a loss-of-function. We observed almost cytoskeletal-like localisation over 72 hours and measured a corresponding decrease in ePPi and increase in mineralization. The phenotype of S375del CMD is severe and at the molecular level even the compensatory increase in ENPP1, TNAP and PIT-1 could not rescue the biochemical phenotype. Mechanistically, we predict the serine deletion results in inappropriate trafficking from the ER and abolishes ANKH function.
G389R. This mutant displayed equivocal molecular data highlighting its variable clinical associations with both CMD and CPPDD. Its cell membrane localisation resembled the CPPDD-associated mutant P5L but we measured a sustained decrease in ePPi similar to the CMD associated mutant S375del. Early upregulation of ENPP1 under low ePPi concentrations further supports this to be a loss-of-function mutation. Hence, since the mutant successfully localised to the cell membrane, the glycine to arginine change at residue 389 appears to cause a loss of transport channel function, which caused the decrease in ePPi and increase in mineralization, although small, reflecting its milder phenotype. The involvement of G389R in autophagy also suggests that it undergoes mis-folding or some other abnormal processing, but perhaps to a lesser degree than E490del.
E490del. The deletion of a glutamate residue at the C-terminus renders it unable to insert effectively into the cell membrane. However, this mutant functioned similar to wt. ANKH in regulating ePPi by increasing ENPP1, TNAP and PIT-1 but was still pathogenic as it increased mineralization. We suggest that this mutant is recognised as a non-functional protein and is constantly recycled by the autophagy machinery, which consequently due to its loss also upregulates ENPP1, TNAP and PIT-1 as compensation for the loss of ANKH at the cell membrane. It is noteworthy that the clinical presentation of CPPDD associated with the E490del is less severe than other forms that might be due to some of the ANKH getting to the membrane but at lower efficiency.
In summary, the corresponding pathogenetic mechanisms for each of the ANKH terminal CPPDD and central CMD associated mutations appear to be unique even between mutants causing similar phenotypes which