Interferon gamma peptidomimetic targeted to interstitial myofibroblasts attenuates renal fibrosis after unilateral ureteral obstruction in mice

Renal fibrosis cannot be adequately treated since anti-fibrotic treatment is lacking. Interferon-γ is a pro-inflammatory cytokine with anti-fibrotic properties. Clinical use of interferon-γ is hampered due to inflammation-mediated systemic side effects. We used an interferon-γ peptidomimetic (mimγ) lacking the extracellular IFNγReceptor recognition domain, and coupled it to the PDGFβR-recognizing peptide BiPPB. Here we tested the efficacy of mimγ-BiPPB (referred to as “Fibroferon”) targeted to PDGFβR-overexpressing interstitial myofibroblasts to attenuate renal fibrosis without inducing inflammation-mediated side effects in the mouse unilateral ureter obstruction model. Unilateral ureter obstruction induced renal fibrosis characterized by significantly increased α-SMA, TGFβ1, fibronectin, and collagens I and III protein and/or mRNA expression. Fibroferon treatment significantly reduced expression of these fibrotic markers. Compared to full-length IFNγ, anti-fibrotic effects of Fibroferon were more pronounced. Unilateral ureter obstruction-induced lymphangiogenesis was significantly reduced by Fibroferon but not full-length IFNγ. In contrast to full-length IFNγ, Fibroferon did not induce IFNγ-related side-effects as evidenced by preserved low-level brain MHC II expression (similar to vehicle), lowered plasma triglyceride levels, and improved weight gain after unilateral ureter obstruction. In conclusion, compared to full-length IFNγ, the IFNγ-peptidomimetic Fibroferon targeted to PDGFβR-overexpressing myofibroblasts attenuates renal fibrosis in the absence of IFNγ-mediated adverse effects.


INTRODUCTION
Patients with chronic kidney disease (CKD) are at increased risk of progression to end-stage renal disease (ESRD). Virtually all forms of CKD are characterized by significant amounts of renal fibrosis [1,2]. Renal interstitial fibrosis is the final common process observed in various kidney diseases that leads to renal failure, and is associated with tubular atrophy and dilation, interstitial matrix deposition, accumulation of interstitial fibroblasts, and inflammatory cell infiltration [3]. There is still no effective therapy to halt renal fibrogenesis, and to prevent progression to ESRD of which the latter requires renal replacement therapy eventually [4]. Among the characteristic changes in the kidney during development of fibrosis, interstitial extracellular matrix (ECM) deposition is the key step. Interstitial fibroblasts are the predominant source of ECM constituents. Therefore, dampening of interstitial fibroblast activation and subsequent prevention of differentiation into ECM-producing myofibroblasts, is an attractive approach to attenuate development of renal interstitial fibrosis [5][6][7].
Interferon-γ (IFNγ) is a cytokine primarily produced by T cells and natural killer (NK) cells. It is a pleiotropic cytokine with anti-viral, anti-bacterial, anti-tumor, proinflammatory but also anti-fibrotic activities [8]. Because of the latter, IFNγ is a potential effective biological to treat or prevent renal fibrosis [9][10][11][12]. However, its potential therapeutic value in fibrosis has been limited so far due to the systemic side effects (because of widespread IFNγ receptor [IFNγR] expression) and rapid renal clearance [13][14][15]. In order to overcome these problems as well as to attain high concentrations in the target cells, cell-specific targeting via receptor-mediated uptake of biologicals is an attractive approach.
Activated fibroblasts in fibrotic tissue are characterized by increased platelet-derived growth factor receptor-beta (PDGFβR) expression [16,17], and this prompted us to use the PDGFβR for receptor-mediated uptake of biologicals. To do so, pegylated full length IFNγ was conjugated to PDGFβR-recognizing peptide (PPB) [18] and by using this conjugate (i.e. PPB-PEG-IFNγ) we were able to demonstrate anti-fibrotic effects in the CCl 4 -induced liver fibrosis mouse model [19] and antitumorigenic effects by targeting stromal cells in the B16 melanoma tumor mouse model [20]. Recently we used the same targeting approach as proof of concept in the mouse unilateral ureteral obstruction (UUO) model of renal fibrosis in which anti-fibrotic effects of the PPB-PEG-IFNγ conjugate were demonstrated [12]. Despite the antifibrotic effects of PPB-PEG-IFNγ, systemic side effects were mildly reduced but remained still present compared to full length IFNγ as evidenced by increased brain MHC class II expression. This side effect was most likely due to binding of the full length IFNγ to the ubiquitously expressed IFNγR.
To overcome the problem of IFNγR-mediated systemic side effects we recently designed a new IFNγbased biological which lacks the amino-terminal IFNγRbinding sequence [21,22]. This IFNγ-peptidomimetic (mimγ) was conjugated to the bicyclic platelet-derived growth factor receptor-beta recognizing peptide (BiPPB) for targeting to PDGFβR-expressing cells (schematically depicted in Figure 1a). The mimγ-BiPPB was recently renamed "Fibroferon" [23]. Here we tested the hypothesis that compared to the previously tested non-targeted full-length IFNγ [12], specific delivery of Fibroferon to PDGFβR-overexpressing interstitial myofibroblasts attenuates renal fibrosis and reduces inflammationmediated side-effects in the mouse UUO model.

PDGFβR expression is increased on interstitial myofibroblasts in UUO mouse kidneys
We first analyzed PDGFβR expression in mouse sham and fibrotic (UUO) kidneys. Confirming our previous data, fibrosis in UUO is characterized by increased interstitial PDGFβR expression as revealed by increased mRNA expression (Figure 2a) and protein expression (based on immunohistochemistry) (Figure 2b). w

Fibroferon reduces renal fibroblast activation
In order to determine whether Fibroferon reduces renal fibroblast activation in vivo, α-SMA expression was determined at both mRNA and protein level. As shown in Figure 3a, UUO (receiving subsequent treatment with vehicle) resulted in increased (p < 0.001) α-SMA mRNA expression compared to sham controls. α-SMA mRNA expression after treatment with Fibroferon and non-targeted full length IFNγ was significantly lower compared with vehicle-treated UUO mice (p < 0.05 vs. vehicle), and similar to sham-operated (no UUO) mice. Immunohistochemistry confirmed increased α-SMA expression in UUO kidneys (vehicle treatment, p < 0.01 vs. sham, Figure 3b,c). Representative photomicrographs of α-SMA staining are shown in Figure 3b. Quantitative analysis revealed significantly reduced α-SMA protein expression after Fibroferon treatment when compared to vehicle treatment (p < 0.05 vs. vehicle, Figure 3c).
Since TGFβ1 is known to induce α-SMA expression in fibroblasts [24] we also analyzed TGFβ1 mRNA expression in sham and UUO kidneys. UUO (vehicle) significantly increased renal TGFβ1 expression (p < 0.01 vs. sham), which was not prevented by Fibroferon or IFNγ treatment (Figure 3d).

Fibroferon reduces extracellular matrix (ECM) deposition in UUO kidneys
Next we analyzed the effects of Fibroferon on ECM constituent expression using immunohistochemistry. Clear differences in protein expression were observed when comparing the different treatment groups ( Figure  4). In UUO (vehicle) significantly increased protein expression levels for interstitial fibronectin (Figure 4a Protein expression in non-targeted full length IFNγ did not statistically differ from vehicle-treated mice. Compared with non-targeted full length IFNγ, these data indicate that Fibroferon is more effective in reducing ECM constituent expression.

Effects of Fibroferon on renal inflammation in UUO
IFNγ exerts systemic pro-inflammatory effects but possibly also local intra-renal inflammatory responses. To analyze the latter, we determined whether treatment with non-targeted full length IFNγ and Fibroferon resulted in different intra-renal inflammatory responses as determined by qPCR analyses, and CD3 + T cell infiltration. As shown in Suppl. Figure 1, compared to sham-operated mice, UUO (with subsequent vehicle treatment) resulted in significantly increased MCP-1 (a), TNF-α (b), IL-1β (c) and IL-6 (d) mRNA expressions (all p < 0.01 vs. sham). Although Fibroferon tended to reduce, and reciprocally non-targeted full length IFNγ tended to increase proinflammatory cytokine production (vs. vehicle), no statistically significant differences were observed. UUO (vehicle) significantly increased renal MHC class II expression that was reduced after Fibroferon treatment when compared with non-targeted full length IFNγ (Suppl. Figure 1e). This might be explained by increased IFNγ-induced macrophage activation rather than macrophage influx (Suppl. Figure 1f). Additionally, UUO resulted in increased CD3 + T cell influx as determined by immunohistochemistry (p < 0.01 vs. sham) that was reduced by mimγ-BiPPB (58% reduction) and nontargeted full length IFNγ-treated mice (42% reduction) (Suppl. Figure 1g).

Fibroferon reduces lymphangiogenesis
New lymph vessel formation (lymphangiogenesis) has been shown in different renal disease models, including UUO [25]. This led us to hypothesize that UUOinduced renal lymphangiogenesis might be attenuated following Fibroferon treatment. Compared to sham, UUO (vehicle treatment) indeed increased numbers of podoplanin + -lymph capillaries (p < 0.01 vs. sham) which were significantly lower in Fibroferon-treated (p ≤ 0.05 vs. vehicle), but not in non-targeted full length IFNγtreated mice (Figure 5a). Depending on the time after UUO induction, both angiogenesis (increase in capillary density) and capillary rarefaction has been described [26]. We analyzed the effect of Fibroferon and non-targeted full length IFNγ on peritubular capillary density. As shown in Figure 5b, UUO (vehicle) was associated with significantly increased numbers of CD31 + peritubular capillaries (p < 0.01 vs. sham), which was not reduced by Fibroferon. Photomicrographs shown in Figure 5a and 5b display representative photos of podoplanin + lymph capillaries and CD31 + peritubular capillaries in UUO (vehicle, Fibroferon, and non-targeted full length IFNγ treatment) and sham kidneys, respectively.

Fibroferon treatment is accompanied by less systemic side effects
To assess whether treatment with Fibroferon reduces systemic side effects compared with non-targeted full length IFNγ, brain MHC class II mRNA expression (as a measure of IFNγ-induced microglia activation) and plasma triglyceride levels (as a measure of IFNγ-induced lypolysis in adipose tissue) were determined. As shown in Figure 6a, in line with our previous observations [12,19,21] treatment with non-targeted full length IFNγ caused significant (p < 0.01) up-regulation of brain MHC class II expression. In contrast, upon Fibroferon treatment brain MHC class II expression was similar to vehicle-treatment. Compared with non-targeted full length IFNγ, Fibroferon treatment significantly reduced plasma triglyceride levels (p < 0.05, Figure 6b). Compared with vehicle, plasma triglyceride levels were not significantly increased after non-targeted full length IFNγ treatment (not shown), which might be explained by the fact that we used nonpegylated full length IFNγ. Finally, we determined weight loss and recovery after induction of UUO in the various treatment groups. As shown in Figure 6c, mice receiving wFibroferon displayed less weight loss and accelerated weight gain after UUO when compared to non-targeted full length IFNγ treated mice, which might be explained either by attenuating the renal damage and/or by reducing side effects in mice receiving Fibroferon.

DISCUSSION
In this study we demonstrate pronounced antifibrotic effects of targeted IFNγ peptidomimetic to PDGFβR-expressing cells without identifiable systemic www.impactjournals.com/oncotarget  side effects in the UUO mouse model of renal fibrosis. This strategy of targeting a chimeric anti-fibrotic biological to PDGFβR provides new opportunities for the treatment of renal fibrosis as well as other conditions associated with enhanced PDGFβR expression such as cancer and vascular remodeling.
Renal fibrosis is the result of tissue injury and remodeling and is a common hallmark and cause of ESRD [27]. Myofibroblast activation and proliferation play a fundamental role in fibrosis progression [28]. Recent progress in understanding the mechanisms underlying the pathogenesis of fibrosis has led to the expectation that various cytokines such as IFNγ could be potentially useful therapeutics to halt or even prevent fibrosis [29]. However, the major obstacle in IFNγ-based therapy is the manifestation of clinical side effects, which has led to abortion of clinical trials or restricted diverse clinical use of IFNγ [14,15,30,31]. These adverse effects often preclude administration of anticipated therapeutic effective dose, leading to lack of efficacy in patients. In line with this, we previously showed that systemic delivery of non-targeted full length IFNγ in mice results in leukocyte and endothelial cell activation, flu-like symptoms, neurotropic effects, hyperlipidemia, elevated TNF-α and IL-6, inflammation of the central nervous system, and elevated triglyceride levels [19], effects most likely mediated through binding to the widely expressed IFNγR. To overcome this hurdle, cell-specific delivery of IFNγ provides a new approach to improve its efficacy and to ameliorate potential systemic side effects. To prevent canonical IFNγ signaling, we used mimγ, which lacks the INFγR-binding domain. Mimγ targeted to liver stellate cells was recently shown to exert anti-fibrotic activity in vivo [21,22]. Here we tested the anti-fibrotic effects of mimγ targeted to PDGFβR-expressing cells in a model of renal fibrosis. Mimγ was directed to interstitial myofibroblasts using a bicyclic PDGFβR-recognizing peptide (Bi-PPB). The mimγ-BiPPB was recently renamed Fibroferon [23].
Interstitial myofibroblasts are characterized by increased expression of PDGFβR in the human and mouse fibrotic kidney [12]. Bi-PPB is a bi-cyclic peptide that binds to the dimeric PDGFβR and may thus be used to deliver the mimγ peptide to the target cells. To determine its anti-fibrotic effects, we first examined α-SMA expression both at the mRNA and protein levels since increased α-SMA levels indicate renal myofibroblast activation, the hallmark of fibrogenesis. We indeed observed a clear inhibitory effect of mimγ-BiPPB on myofibroblast activation (i.e. reduced α-SMA expression). ECM composition is significantly altered and expanded during fibrogenesis, which can severely compromise resident cell function (e.g. tubular epithelium) [5,32]. We therefore analyzed expression of the most prominent ECM components fibronectin, and collagens I and III. Again, a clear pharmacological effect of Fibroferon was noted, and compared to non-targeted full length IFNγ the efficacy of Fibroferon in reducing ECM deposition was significantly higher, illustrating the potency of our cell-specific targeting strategy. Of note, although mRNA expression levels of fibronectin and collagens I & III were lower in Fibroferon-treated mice compared with vehicletreated mice, these differences were not significant (data not shown). These data suggest that the protective effects of Fibroferon are (partly) mediated by enhanced fibrolysis as we demonstrated previously using another interferon-γ based conjugate in a model of liver fibrosis [33]. TGFß1 is a pro-fibrotic protein which instigates fibroblast activation and transformation into myofibroblasts, resulting in collagen synthesis and ECM contraction [24]. However, we did not observe altered TGFß1 expression after IFNγ or Fibroferon treatment.
Inflammation commonly precedes and accompanies fibrosis [34]. There is evidence for a direct role of T cells in the pathogenesis of renal interstitial fibrosis which can be explained through different mechanisms [35,36]: (1) direct effects on fibroblast activation and ECM deposition, (2) indirect effects on fibroblasts via induction of a profibrotic phenotype of infiltrating macrophages resulting in fibroblast activation through profibrotic cytokines and growth factors (e.g. PDGF, CTGF, TGFß1), and (3) indirect effects on fibroblasts via tubular epithelial cell activation and subsequent production of cytokines and growth factors that act on fibroblasts. We indeed observed an almost two-fold reduction in the number of infiltrating T cells after mimγ-BiPPB treatment, possibly mediated via reduced expression of T cell chemoattractants that are shown to be produced by myofibroblasts [37,38].
In the setting of renal fibrosis, injured renal cells and inflammatory cells secrete mediators and growth factors that promote angiogenesis and/or lymphangiogenesis [37][38][39][40][41]. Since anti-lymphangiogenic effects of IFNγ have been reported [42], we analyzed the potential inhibitory effect of Fibroferon on lymphangiogenesis. Renal lymphangiogenesis in UUO kidneys was significantly reduced following Fibroferon treatment, which was not observed after treatment with non-targeted full length IFNγ. Angiogenesis was also shown to occur during the progression of fibrosis in UUO [43], which was confirmed in our study. A reduction, though non-significant, was observed in the percentage surface area of CD31 + peritubular capillaries following Fibroferon treatment. Reduced lymphangiogenesis and angiogenesis most likely reflect indirect effects secondary to the direct anti-fibrotic effects of Fibroferon on myofibroblasts.
Recently we reported on the anti-fibrotic effects of full length IFNγ targeted to PDGFβR-expressing myofibroblasts in UUO mice [12]. When comparing the anti-fibrotic effects of targeted full length IFNγ [12] and Fibroferon, the attenuating effects on αSMA and ECM constituents are comparable. However, clear differences were observed in favor of Fibroferon with respect to CNS and renal inflammation. Compared with targeted full length IFNγ [12], brain MHC class II expression was 79% reduced upon Fibroferon treatment. Similarly, renal F4/80 and MHC class II expression were significantly reduced with 33% and 53%, respectively. We used MHC class II expression as a marker of macrophage activation. Macrophages are the main target cells for native INFγ. For anti-fibrotic therapies, generalized macrophage activation is an off-target effect leading to serious adverse effects, and the brain is an off-target organ for renal diseases.
Our data indicate that Fibroferon lacks this off-target effect, in contrast to targeted full length IFNγ, which can be explained by the lack of Fibroferon binding to the extracellular domain of IFNγR. A potential weakness of current study is that we did not include data demonstrating a direct comparison between targeted full length IFNγ (PPB-PEG-IFNγ) and Fibroferon. However, we recently published on the anti-fibrotic effects of PPB-PEG-IFNγ [12], and, compared with the vehicle control group, the beneficial anti-fibrotic effects of Fibroferon are similar. The mechanism by which Fibroferon exerts its protective effects is not yet fully elucidated. We previously demonstrated mimγ playing a direct role in the activation and nuclear translocation of the transcription factor STAT1α via the nuclear signaling sequence (NLS) present in the C-terminal region of IFNγ. This region is able to modulate IFNγ-responsive genes through activation of the JAK/STAT pathway [21]. Mimγ does not recognize the extracellular domain of IFNγR but rather binds to the cytoplasmic domain of the receptor chain 1 (IFNγR1), and thereby initiates cellular signaling [44][45][46][47][48][49]. Binding of Bi-PPB to the PDGFβR has been demonstrated in previous studies 19,20 . We propose that mimγ-BiPPB binds to the PDGFβR and following internalization, the construct is able to interact with the intracellular part of IFNγR1 and initiates phosphorylation of STAT1 thereby eliciting IFNγrelated biological effects. Fibroferon just has the IFNγ signaling moiety and is only biologically active once it is internalized. This premise needs further study though.
In conclusion, in this study we demonstrate the benefits of (targeted) Fibroferon over non-targeted full length IFNγ with respect to its anti-fibrotic effects and the development of off-target systemic side effects. We demonstrate a proof-of-concept strategy for the use of targeted IFNγ-based biologicals to reduce renal fibrosis by removing the extracellular receptor recognition sequence and directing it to myofibroblasts. The small size and increased stability of Fibroferon compared to the full length targeted IFNγ will provide better therapeutic benefits and more possibilities for subcutaneous or muscular administration. In addition, based on these and our previous data we believe that Fibroferon offers high specificity and low toxicity with full agonistic activity, therefore has highly favorable characteristics for the prolonged administration. This strategy is one of the first examples of re-direction of a cytokine to the kidney by creating a chimeric peptide. Our data warrant further exploration of the precise underlying mechanism and mode of action of Fibroferon as well as of the possibility to use Fibroferon to target fibrosis in a clinical setting.

Experimental animals and surgical procedures
Male C57Bl/6 mice (9-12 wks; 28-30 g) were obtained from Harlan (Zeist, the Netherlands). Animals were housed in cages with free access to food and water. During the operation the abdomen was opened by a midline incision under general anesthesia (2% isoflurane/O 2 ). All mice were subjected to unilateral ureteral obstruction (UUO) by a double-ligation (using 6-0 silk suture) of the left ureter proximal to the kidney. Right kidneys were used as sham controls and were manipulated but not ligated. After surgery, mice were i.v. injected (penile vein) at days 2, 4 and 6 after surgery (see Figure 1b) with either 1) UUO, vehicle (saline), 2) UUO, Fibroferon (mimγ-BiPPB, 5 µg/mouse/injection), 3) UUO, non-targeted full-length IFNγ (5 µg/mouse/injection). Total follow-up time was 7 days during which mice were weighed daily. Group size was n = 6/group/time point. The experimental animal studies have been approved by the local Institutional Review Board of the University of Groningen (DEC 6066B) and adhered to the national guidelines for the care and use of laboratory animals.

Immunohistochemistry
Immunohistochemistry for α-SMA, fibronectin, type III collagen, CD31, podoplanin, and CD3 was performed on 2 µm paraffin sections. Sections were deparaffinized in xylene and rehydrated in graded alcohol series and distilled water. Antigen retrieval was achieved by overnight incubation at 80ºC in 0.1 M Tris/HCl buffer (pH 9.0) for fibronectin and type III collagen staining, and treatment in the microwave for 15 min at 300W with 10 mM citrate buffer (pH 6:0) for podoplanin and CD31 staining. No antigen retrieval was performed for α-SMA staining. Type I collagen and PDGFβR staining was performed on 4 µm acetone-fixed frozen sections. Endogenous peroxidase activity was blocked with 0.03% H 2 O 2 (in PBS) for 30 min. Primary antibody binding was detected by sequential incubations with horseradish peroxidase (HRP)-labeled appropriate secondary and tertiary antibodies (all obtained from DAKO, Glostrup, Denmark). Peroxidase activity was visualized using 3, 3'-diaminobenzidine tetrahydrochloride (DAB, DAKO, Glostrup, Denmark) as chromogen (10 min. incubation). Sections were counterstained with hematoxylin for 1 min and mounted with Kaiser's glycerin gelatin.

Quantification of immunostaining
The sections were scanned using a NanoZoomer HT (Hamamatsu Photonics K.K., Shizuoka Pref., Japan). The extent of cortico-interstitial α-SMA, fibronectin, type I and III collagens, and CD3 expression was measured as numbers of positive pixels, and the numbers of podoplanin-positive vessel-like structures were counted in 30 cortical areas using Aperio ImageScope software (version 9.1.772.1570, Aperio Technologies Inc, Vista, CA, USA) at 200x magnification. For CD31 staining, 10 cortical areas at 400x magnification were selected, and the percentage of CD31 positive area was measured using ImageJ 1.46r (Rasband, W.S., U.S. National Institutes of Health).

Quantitative real-time PCR
Total RNA from kidney tissues was extracted using RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The RNA concentrations were measured on a NanoDrop ND-1000 spectrophotometer (Thermo Scientific, USA). Singlestranded cDNA was synthesized using Superscript II and random hexamer primers (Invitrogen, Carlsbad, CA) in a volume of 20 μl. cDNA's were diluted to a concentration of 2 ng/μl and 2.5 μl/reaction (5 ng) was used for qRT-PCR analysis. PCR reactions were performed in a 10 μl reaction volume containing 1 x qPCR master mix (EUrogentec, Liege, Belgium), 1x Taqman Gene Expression Assay mix (Applied Biosystems, Forster City, CA). The primer reference numbers were as follows: Ywhaz: Mm03950126_s1, α-SMA/Acta2: Mm01546133_ m1, MHC-II/CD74: Mm00658576_m1, and PDGFβR: Mm00435546_m1. qRT-PCR reactions were performed on a ABI7900HT thermal cycler (Biosystems, Forster City, CA). Relative gene expression was calculated using the 2 -ΔCt method with Ywhaz (tyrosine 3-monooxygenase/ tryptophan 5-monooxygenase activation protein ζ) as housekeeping gene. Data are expressed as fold-change compared to sham-operated kidneys.

Plasma triglyceride measurements
Whole blood was collected into commercially available EDTA-coated tubes (Item No: 450474, Greiner Bio-One, GmbH, Austria). The tubes were kept on ice during processing. Cells were removed from plasma by centrifugation for 10 minutes at 1000-2000 x g using a refrigerated centrifuge. Triglyceride levels were measured on a Vitalab Selectra-E chemistry workstation (Vital Scientific BV, Spankeren, The Netherlands).

Statistical analyses
Statistical analyses were performed using GraphPad Prism 5.0 (GraphPad Software Inc, La Jolla, CA). Differences between multiple groups were calculated using the non-parametric one-way ANOVA (Kruskal-Wallis test) with Dunn's Multiple Comparison Test posttest. Comparison of two groups was performed using the Mann Whitney test (or Student's t-test where indicated). Comparison of weight loss and recovery in the various groups was performed using Repeated Measures ANOVA with Bonferroni post-test. p ≤ 0.05 was considered statistically significant. Data are expressed as mean ± standard error of the mean (or mean ± standard deviation where indicated).

CONFLICTS OF INTEREST
K.P. is co-founder and shareholder of BiOrion Technologies BV. All other authors declare no competing financial interests.