LAR protein tyrosine phosphatase regulates focal adhesions through CDK1

Traditionally, phosphatases have been considered relatively non-specific regulators of cellular signalling pathways, a view that has been increasingly challenged in recent years. It is now clear that phosphatases play very specific roles in both the positive and negative propagation of signals (Tonks, 2013). Leukocyte common antigen-related (LAR), also known as receptor type protein tyrosine phosphatase F (PTPRF), is a receptor protein tyrosine phosphatase. It is comprised of an extracellular domain that contains three immunoglobulin domains and eight fibronectin type III domains in addition to two cytoplasmic phosphatase domains, D1 and D2 (Chagnon et al., 2004). LAR is expressed in a number of human tissues including brain, heart and bladder, and LAR-deficient mice are characterised by axon guidance defects, diabetes, neuron degeneration, cancer and abnormal development of mammary glands (Chagnon et al., 2004; Van Lieshout et al., 2001). Previous work has identified roles for LAR in neuronal growth and regeneration, mammary gland development and in mitogenic signalling through PDGF receptor β (PDGFRβ) (Schaapveld et al., 1997; Um and Ko, 2013; Zheng et al., 2011).

Cell interaction with the extracellular matrix is mediated in part through complex, dynamic multi-molecular structures called focal adhesions that link the actin cytoskeleton to the extracellular matrix through integrin adhesion receptors (Wehrle-Haller, 2012). Focal adhesions mediate transfer of signals from the extracellular matrix to the cytoplasm and nucleus, regulating many functions including proliferation, apoptosis, differentiation and cell migration (Wehrle-Haller, 2012). Proteins localizing to focal adhesions include structural proteins that link integrins to the actin cytoskeleton as well as many other proteins involved in signal transduction. These include the well-documented regulators of focal adhesions, Src and focal adhesion kinase (FAK, also known as PTK2) but also other kinases including members of the cyclin-dependent kinase (CDK) family (Horton et al., 2015; Robertson et al., 2015; Zaidel-Bar et al., 2007). The extracellular domain of LAR is known to interact with components of the extracellular matrix including the laminin–nidogen complex and syndecan (Johnson et al., 2006; O'Grady et al., 1998), and imaging and proteomic studies have established that LAR is a component of focal adhesions and can regulate the actin cytoskeleton (Serra-Pages et al., 1995; Zaidel-Bar et al., 2007).

However, although LAR is known to localise to focal adhesions and interact with components of the extracellular matrix, its role in cell adhesion has not been established. To analyse the role of LAR in cell adhesion, we used wild-type (WT) mouse embryonic fibroblasts (MEFs) and MEFs in which the LAR phosphatase domains have been deleted (LARΔP) (Schaapveld et al., 1997). Our results reveal a new signalling pathway in which LAR phosphatase signals through CDK1 and c-Abl and Akt family proteins to regulate adhesion complex assembly and optimal adhesion to fibronectin.

LAR localises to adhesion complexes and is known to regulate the actin cytoskeleton (Bateman et al., 2001; Serra-Pages et al., 1995). Therefore, we evaluated the role of LAR in adhesion complex formation and adhesion to extracellular matrix. To do this we used MEFs expressing either WT LAR or a truncated form of LAR lacking phosphatase activity (LARΔP) (Schaapveld et al., 1997). Following serum starvation, cells were either unstimulated or stimulated with PDGF-BB, a well-established regulator of adhesion complex formation (Ridley et al., 1992), for 14 min prior to fixation and staining using an antibody against phosphorylated paxillin (Tyr¹¹⁸) (hereafter phospho-paxillin) to visualise adhesion complexes (Burridge et al., 1992). We observed a significant decrease in phospho-paxillin staining in unstimulated LARΔP cells when compared to WT cells (Fig. 1A,B,E). Treatment with PDGF-BB resulted in increased phospho-paxillin staining in both LARΔP and WT cells but again there was a significant decrease in adhesion complexes in LARΔP cells compared to WT cells (Fig. 1C–E).

The extracellular matrix protein fibronectin is secreted by fibroblasts and plays a key role in cell–extracellular-matrix interactions during development and in wound healing (Hynes, 1994). In order to investigate whether LAR phosphatase activity contributes to these cell–extracellular-matrix interactions, we used a standard adhesion assay to analyse attachment of WT and LARΔP MEFs to fibronectin. Serum-starved WT and LARΔP cells were allowed to attach to fibronectin in the presence or absence of 20 ng ml⁻¹ PDGF-BB for 30 min. LARΔP cells were substantially impaired in their ability to attach to fibronectin when compared to WT cells, and this was apparent over a range of fibronectin plating concentrations between 0.5 and 10 μg ml⁻¹ (Fig. 2A). Treatment with PDGF-BB did not significantly affect attachment of either WT or LARΔP cells (Fig. 2A). To confirm these data, we used RNA interference (RNAi) to knockdown expression of LAR in WT cells (Fig. 2B). Small interfering RNA (siRNA)-mediated reduction of LAR expression in WT cells was accompanied by a significant decrease in cell attachment to 10 μg ml⁻¹ fibronectin when compared to WT cells transfected with a non-silencing control (NSC) oligonucleotide (Fig. 2C). Having observed that LAR regulated adhesion complex formation and attachment to fibronectin, we analysed cell migration of WT and LARΔP cells, using a standard scratch wound assay. We found that loss of LAR phosphatase activity did not affect cell migration, with both WT and LARΔP cells closing wounds at equivalent rates (Fig. 2D). We conclude that LAR phosphatase regulates adhesion complex formation and attachment to fibronectin in MEFs but has no effect on cell migration.

FAK and Src are non-receptor tyrosine kinases frequently linked to adhesion complex dynamics, cell adhesion and cell migration (Mitra and Schlaepfer, 2006), although recent data also report that adhesion complex composition is unaffected when Src and FAK are inhibited (Horton et al., 2016). Tyr¹¹⁸ of paxillin is a known FAK substrate (Bellis et al., 1995), and we tested whether activation of FAK was affected in the absence of LAR activity. Autophosphorylation of Tyr³⁹⁷ is a key stage in activation of FAK, creating a docking site for the recruitment and activation of the non-receptor tyrosine kinase Src, which subsequently phosphorylates FAK on Tyr⁴⁰⁷ (Calalb et al., 1995). Consistent with immunostaining of adhesion complexes (Fig. 1) we observed substantially reduced paxillin Tyr¹¹⁸ phosphorylation in LARΔP cells when compared to WT cells, both in the absence and presence of PDGF-BB (Fig. 3A). However, when we analysed FAK activity in LARΔP cells, we observed no significant difference in phosphorylation of either Tyr³⁹⁷ or Tyr⁴⁰⁷ when compared to WT cells (Fig. 3B–D). These data are consistent with the previous observation that activation of Src is not regulated by LAR (Zheng et al., 2011) and indicate that the decreased phosphorylation of paxillin and reduced adhesion complex formation seen in LARΔP cells are not a consequence of altered FAK signalling.

To identify key signalling events regulated by LAR we analysed our comparative (WT versus LARΔP MEFs) global phosphoproteomic dataset (deposited in the ProteomeXchange Consortium via the PRIDE partner repository; http://www.ebi.ac.uk/pride/archive, dataset identifier PXD002545; Sarhan et al., 2016). Within this dataset, we identified altered phosphorylation of 270 phosphorylation sites in 205 proteins in PDGF-BB-stimulated LARΔP cells when compared to WT cells. To identify key nodes of regulation within the dataset, the kinase prediction tool GPS (Xue et al., 2008) was used to identify putative kinases upstream of the 270 identified phosphorylation motifs. A cluster of 38 proteins containing predicted CDK1 phosphorylation motifs were identified with significantly reduced phosphorylation in LARΔP cells. Of these, 19 (50%) formed part of the recently characterised meta-adhesome (Horton et al., 2015) (Fig. 4A).

Phosphorylation of Thr¹⁶¹ in the activation loop of CDK1 is a key step in activation of the kinase (Krek and Nigg, 1992; Solomon et al., 1992), therefore we used a phospho-specific antibody to analyse activation of CDK1 in WT and LARΔP cells. In serum-starved WT cells, basal phosphorylation of CDK1 Thr¹⁶¹ was observed and this increased on addition of PDGF-BB (Fig. 4B). In contrast, the basal phosphorylation of Thr¹⁶¹ in LARΔP cells was lower than in WT cells and no increase in phosphorylation of Thr¹⁶¹ was observed when cells were treated with PDGF (Fig. 4B). CDK activity requires interaction with a partner cyclin and CDK1 forms a complex with cyclin B1 during G2/M (Malumbres, 2014) but no difference in expression of CDK1 or cyclin B1, or association between CDK1 and cyclin B1, was observed when LARΔP cells were compared to WT cells (Fig. 4C). This suggests that the decrease in CDK1 activity in LARΔP cells was not a consequence of cyclin availability.

In the absence of LAR phosphatase activity, we have observed decreased phosphorylation of CDK1 consistent with decreased activity and have identified a cluster of potential adhesion-related CDK1 substrates. CDK1 has recently been linked to cell adhesion (Robertson et al., 2015), thus we next examined the role of CDK1 in LAR-dependent cell adhesion. Treatment of WT MEFs with the CDK1 inhibitor, RO-3306, resulted in significantly fewer focal adhesions, consistent with CDK1 regulating adhesion complex formation in this cell type (Fig. 4D,E). To confirm that CDK1 regulates adhesion complexes in MEFs downstream of LAR we transiently expressed a constitutively active form of CDK1 (CDK1 AF) in LARΔP cells. This resulted in a significant increase in adhesion complex formation (Fig. 4F,G). These data establish CDK1 as a protein kinase acting downstream of LAR to regulate adhesion complex formation.

Previous work has shown that LAR regulates PDGF-mediated activation of Akt (Zheng et al., 2011) and loss of Akt activity is associated with decreased phosphorylation of CDK1 at Thr¹⁶¹ (Nogueira et al., 2012), suggesting that Akt proteins might be part of the pathway downstream of LAR leading to activation of CDK1 and regulation of adhesion complex formation.

To investigate whether Akt is involved in LAR-dependent cell adhesion, we analysed adhesion complex formation in WT cells treated with an Akt inhibitor (InSolution Akt Inhibitor VIII) and observed a significant (P<0.001) decrease in adhesion complex formation compared to control vehicle (DMSO)-treated cells (Fig. 5A). Activation of Akt requires phosphorylation of Thr³⁰⁸ in the activation loop (Alessi et al., 1996), and consistent with previously published data showing that PDGFRβ-induced activation of Akt is dependent on LAR (Zheng et al., 2011), we observed a rapid increase in phosphorylation of Akt Thr³⁰⁸ in control, serum-starved WT cells, treated with PDGF (Fig. 5B,C) and significantly reduced phosphorylation of Thr³⁰⁸ in LARΔP cells at all time points (Fig. 5B,C). Treatment of WT cells with the Akt inhibitor, in the absence or presence of PDGF-BB, also resulted in a significant reduction in Thr¹⁶¹ phosphorylation of CDK1 (Fig. 5D) confirming Akt-dependent regulation of CDK1 activity in MEFs. This provides further evidence for a new role of LAR in regulation of cell adhesion through CDK1.

Previous work has reported hyper-phosphorylation of the tyrosine kinase c-Abl protein family in LARΔP cells and Abl has been identified as a potential substrate for the Drosophila LAR ortholog (Wills et al., 1999; Zheng et al., 2011). In mammalian cells a large number of c-Abl substrates have been identified, many of which are involved in regulation of cell adhesion and the actin cytoskeleton (Wang, 2014; Woodring et al., 2003). To establish whether c-Abl is a LAR substrate, we transfected 293T cells with cDNA constructs expressing either WT LAR or one of two LAR substrate-trapping mutants, LAR C/S (C1548S) or LAR D/A (D1516A) (Fig. 6A) (Wang et al., 2007). These mutations within the intracellular phosphatase domain D1 allow LAR to interact with substrates but reduce its dephosphorylation activity (Wang et al., 2007). Using this approach, we identified that both the LAR trapping mutants and WT LAR interacted with c-Abl (Fig. 6B). When compared to cells transfected with either trapping mutant, a reduction in tyrosine phosphorylation was observed in c-Abl bound to WT LAR (Fig. 6B). These data identify c-Abl as a LAR substrate. To determine whether c-Abl is involved in LAR-mediated regulation of CDK1 we treated LARΔP cells with AG957, a selective inhibitor of c-Abl (Anafi et al., 1992), and examined the effect of inhibiting c-Abl on activation of Akt and CDK1. Inhibition of c-Abl in LARΔP cells resulted in significantly increased phosphorylation of both Akt Thr³⁰⁸ and CDK1 Thr¹⁶¹ (Fig. 6C–F). These data indicate that c-Abl is a key component of the LAR pathway upstream of Akt and CDK1.

Much of what we currently know about LAR function focuses on its role in the nervous system and comes from work on the Drosophila ortholog (Um and Ko, 2013). Drosophila LAR localises to synaptic junctions and regulates synaptic growth and axonal guidance through regulation of the actin cytoskeleton through Rho family GTPases (Bateman et al., 2000; Kaufmann et al., 2002; Pawson et al., 2008). Outside of the nervous system, relatively little is known about LAR function, although Drosophila LAR has been reported to regulate the actin cytoskeleton in Drosophila epithelia (Bateman et al., 2001; Conder et al., 2007). In addition, in mammalian cells an interaction between LAR and EphA2 has been reported to be important for cell migration (Lee and Bennett, 2013). Our data, showing decreased adhesion complex formation and decreased cell adhesion to extracellular matrix in the absence of LAR phosphatase activity, identify LAR as a regulator of cell adhesion to extracellular matrix in mammalian cells. Furthermore, we identify that LAR affects cell adhesion through regulation of CDK1 activity.

Kinase prediction analysis of LAR-regulated phosphosites identified a node of cytoskeleton- and adhesion-related proteins centred on CDK1 (Fig. 4), 50% of which form part of the recently characterised meta-adhesome (Horton et al., 2015). Most work has focused on nuclear functions for CDKs, and the role of CDK1 in regulating the G2/M cell cycle transition is well established (Malumbres, 2014). However, there is increasing evidence from both imaging and proteomic studies to support the hypothesis that CDKs have significant non-nuclear roles. CDK1 has been identified as being present in focal adhesions and recent data indicates that CDK1, along with other CDKs, plays a role in integrin-based cell adhesion and motility (Manes et al., 2003; Robertson et al., 2015). During G2/M transition in vertebrates, phosphorylation of CDK1 Thr¹⁶¹ is a function of CDK7, a CDK-activating kinase (CAK), and one possibility is that LAR might also regulate CDK7 activity (Desai et al., 1995; Fisher, 2005). However, analysis of our phosphoproteomics dataset (Sarhan et al., 2016) did not reveal any significant change in phosphorylation of Ser¹⁶⁴, which is required for activation of CDK7 (Larochelle et al., 2001). This suggests LAR might regulate phosphorylation of Thr¹⁶¹ and activation of CDK1 in a CDK7-independent manner.

We have identified that the tyrosine kinase c-Abl is a direct substrate for LAR in mammalian cells and a key component linking LAR phosphatase to CDK1. Consistent with these findings, c-Abl localises to adhesion complexes and the Drosophila ortholog of LAR is known to interact with Abl to regulate axon guidance, an actin- and adhesion-dependent process (Wills et al., 1999; Zaidel-Bar et al., 2007). The question remains as to how c-Abl is regulating CDK1 activity, and our data indicate this is achieved through Akt as we observed that inhibition of Akt blocked LAR-dependent activation of CDK1 (Fig. 5D). As LAR has been reported to regulate PDGF signalling through c-Abl (Zheng et al., 2011) and Akt is well-established as a signalling component downstream of receptor tyrosine kinases including PDGFR (Andrae et al., 2008), one possible explanation for our findings is that LAR regulates CDK1 and hence cell adhesion through the PDGF pathway. The decreased Akt Thr³⁰⁸ phosphorylation seen in LARΔP cells following PDGF stimulation (Fig. 5B) would support this idea and Akt has also been reported to regulate PDGF-dependent phosphorylation of FAK at Tyr³⁹⁷ (Higuchi et al., 2013). However, we find no evidence that LAR regulates FAK activity and phosphorylation of CDK1 Thr¹⁶¹ in LARΔP cells is not affected by PDGF (Fig. 4B). In addition, we observed that LAR can regulate adhesion complex formation and cell adhesion independently of PDGF stimulation (Figs 1 and 2). One surprising observation was that although inhibition of c-Abl restored phosphorylation of CDK1 Thr¹⁶¹ in unstimulated LARΔP cells, thus identifying c-Abl as a component of the pathway linking LAR to activation of CDK1, stimulation with PDGF resulted in a steady decrease in CDK1 Thr¹⁶¹ phosphorylation over time. One possible explanation for this is that in the absence of LAR and c-Abl activity, a phosphatase is activated that can dephosphorylate CDK1. Irrespective of this, our data clearly demonstrate a role for c-Abl in regulating LAR-mediated phosphorylation of CDK1 and taken together, our data indicate that although PDGF signalling can act to modulate LAR-dependent effects on adhesion, LAR can also regulate CDK1 and cell adhesion independently of PDGF signalling (Fig. 7).

We have observed that LAR positively regulates attachment to fibronectin and, although there is no evidence for LAR interacting directly with fibronectin, LAR has been reported to interact with the laminin–nidogen complex (O'Grady et al., 1998). This would suggest that LAR might play a role in regulating both ‘inside-out’ and ‘outside-in’ signalling in a manner analogous to integrins (Hynes, 2002) and it would be interesting to analyse whether different extracellular matrix substrates elicit distinct LAR-dependent signalling events. Interestingly, although we have clear evidence that LAR regulates cell adhesion, loss of LAR activity had no effect on cell migration (Fig. 2). This would be consistent with our data showing that LAR did not affect the activity of FAK (Fig. 3), a known regulator of cell migration in fibroblasts (Sieg et al., 1999). Given the ability of LAR to regulate both PDGF-dependent mitogenic signalling and adhesion to extracellular matrix, it is tempting to speculate that LAR plays a pivotal role in integrating signals required for cell cycle progression. Similarly, LAR might also be important in anoikis, where detachment of a normal adherent cell is a prelude to programmed cell death (Reddig and Juliano, 2005).

In summary, we have identified a new mechanism whereby a membrane-associated protein tyrosine phosphatase positively regulates adhesion complex formation and cell adhesion through a signalling pathway involving c-Abl, Akt and CDK1.

For immunoblotting, antibodies were diluted 1:1000 and, for immunocytochemistry, antibodies were diluted 1:100. Antibodies against Akt (#9272), phospho-Akt Thr³⁰⁸ (#9275; antibody recognises equivalent epitopes on Akt1, Akt2 and Akt3), cyclin B1 (#4135), paxillin (#2542), phospho-paxillin Tyr¹¹⁸ (#2541), phospho-CDK1 Thr¹⁶¹ (#9114) and c-Abl (#2862) were obtained from Cell Signaling Technology (Danvers, MA). Antibodies against CDK1 (#sc-54), phospho-FAK Tyr⁴⁰⁷ (#sc-16664) and PY99 phosphotyrosine (#SC-7020) were obtained from Santa Cruz Biotechnology (Dallas, TX). Antibodies against FAK (#610087) and phospho-FAK Tyr³⁹⁷ (#611722) antibodies were purchased from BD Biosciences (Franklin Lakes, NJ). FLAG M2 antibody (#F3165) was from Sigma-Aldrich (Poole, UK). Anti-LAR antibody (#73-193) was purchased from Neuromab (UC Davis, CA). The anti-Alix antibody was a gift from Carl Hendrik Heldin (Karolinska Institute, Sweden) (Lennartsson et al., 2006). Alexa-Fluor-594-conjugated phalloidin and DAPI were from Life Technologies (Paisley, UK). Goat anti-mouse-IgG IRDye-conjugated antibody and goat anti-rabbit-IgG HRP-conjugated antibodies were from LI-COR Biosciences (Lincoln, NE).

All cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, 10 mg ml⁻¹ streptomycin, and 250 μg ml⁻¹ amphotericin B and periodically tested for, and shown to be free of, mycoplasma contamination. WT and LARΔP MEFs were kind gifts from Wiljan Hendriks (Radbound University Medical Centre, The Netherlands) (Schaapveld et al., 1997). Cell culture reagents were purchased from Gibco Life Technologies (Paisley, UK). For growth factor stimulation experiments, cells were grown to 80% confluence and starved for 16 h in serum-free DMEM before being treated with 20 ng ml⁻¹ recombinant PDGF-BB (Cell Signaling Technology) at 37°C. In some experiments, cells were incubated with 10 μM InSolution Akt Inhibitor VIII (Barnett et al., 2005) (Merck Millipore, Watford, UK) to inhibit Akt, 10 μM AG957 (Sigma-Aldrich, Poole, UK) to inhibit c-Abl or RO-3306 (Merck Millipore, Watford, UK) to inhibit CDK1. The carrier DMSO was added to control cells.

Cell lysates were prepared, separated by SDS-PAGE, transferred to nitrocellulose membranes and immunoblotted using previously described protocols (Zheng et al., 2011). Immunoblots were visualised using fluorescence detection on the Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln NE). Densitometric analysis was carried out using ImageJ. Immunoprecipitation of cyclin B1 or CDK1 using Dynabeads protein G (Novex, Life Technologies) was performed as per the manufacturer's protocol.

Adhesion assays were performed as described previously (Lock and Hotchin, 2009) with the modification that adherent cells were quantitated by addition of 5 mg ml⁻¹ of Thiazolyl Blue tetrazolium bromide (MTT) (Sigma-Aldrich, Poole, UK) for 1 h at 37°C followed by addition of 100 μl of isopropanol for 15 min at room temperature, after which absorbance was measured at 570 nm. For the scratch wound assays, cells were plated in 96-well cell culture plates and cultured to confluency. Scratch wounds were made using an automated wound maker (Essen BioScience, Hertfordshire, UK) and images collected at hourly intervals post-wounding using an IncuCyte ZOOM imaging system (Essen BioScience, Hertfordshire, UK).

WT and LARΔP cells were grown on acid-etched glass coverslips before being cultured overnight in serum-free medium. Where indicated, seeded cells were incubated with either 10 μM CDK1 RO-3306 specific inhibitor for 1 h or 10 μM InSolution Akt Inhibitor VIII or in DMSO overnight. Cells were stimulated with 20 ng ml⁻¹ PDGF-BB for 14 min and then incubated on ice for 5 min. Fixed and permeabilised cells were immunostained as previously described (Lock and Hotchin, 2009). Cells were visualised using TIRF Nikon A1R confocal microscope with NIS-Elements Software and images were analysed using ImageJ 1.48. To quantify the adhesion complex area, background-subtracted images of a minimum of four cells per experiment were thresholded to identify and measure adhesion pixels using ImageJ software using a previously published method (Webb et al., 2004).

Cells were transiently transfected with 80 pmol of SMARTpool ON-TARGETplus siRNA for murine LAR (L-042444-00-0005; Dharmacon, Lafayette, CO) using Lipofectamine RNAiMAX (Life Technologies, Paisley, UK) as described elsewhere (Lock and Hotchin, 2009). Non-silencing control (NSC) siRNA was used as a control for each experiment.

To ‘rescue’ CDK1 activity, LARΔP cells were transiently transfected with a constitutively active form of CDK1 (CDK1-AF) (Hagting et al., 1998) using Lipofectamine 2000 (ThermoFisher Scientific) according to the manufacturer's instructions. In the LAR phosphatase trap experiment, 293T cells were transiently transfected using Lipofectamine 2000 with cDNA constructs expressing either WT c-Abl (WT c-Abl) (Johannessen et al., 2010) or WT LAR, LAR-C1538S (LAR C/S) or LAR-D1506A (LAR D/A) (Wang et al., 2007). The LAR constructs were kindly provided by Ruey-Hwa Chen (National Taiwan University, Taipei, Taiwan). The CDK1-AF construct was obtained from Addgene (Cambridge, MA).

Potential CDK1 substrates were identified using the Group-based Prediction System (GPS; version 2.1.2) (Xue et al., 2008). To minimise false positives, the highest threshold was applied. The protein–protein interaction (PPI) network analyses were performed using Cytoscape (version 3.3.0) (Shannon et al., 2003). GO enrichment analyses were performed using (DAVID; version 6.7) (Huang et al., 2009).

We acknowledge the help of Alessandro Di Maio (Birmingham Advanced Light Microscopy facility) and Cleidiane Zampronio (Functional Genomics, Proteomics and Metabolomics Facility, University of Birmingham). We are also grateful to Justyna Szyroka and Peter Noy for help with the scratch wound assay. We would like to dedicate this paper to the memory of our friend, colleague and co-author, Carina Hellberg, who sadly died in 2014.