Background
Gaucher disease (GD) [
1] is a recessively inherited lysosomal storage disorder caused by deficiency of a lysosomal enzyme, glucocerebrosidase (EC 3.2.1.45), which leads to insufficient clearance of the enzyme’s substrate, cellular glucosylceramide. Pathologic accumulations of glucosylceramide (or other substrates, such as glucosylsphingosine) in the lysosomes of tissue macrophages (Gaucher cells) results in splenomegaly, hepatomegaly and multiple forms of skeletal disease [
2]. Three clinical phenotypes have been described: type 1, the prevalent form usually defined by the absence of central nervous system impairment; and types 2 and 3, both rare and severe, have central neurological involvement [
3]. GD diagnosis is confirmed by the detection of low glucocerebrosidase activity, usually less than 30% of the normal value in peripheral leukocytes. Genotyping can sometimes provide prognostic information [
4]. More than 250 mutations of the GBA1 gene encoding lysosomal glucocerebrosidase have been reported as being associated with GD, but the predominant mutation in type 1 GD is called N370S (or c.1226A > G) [
5]. The N370S mutation is usually protective against neuronopathic disease. GD can be treated by enzyme replacement therapy (ERT). The first enzyme preparation used to treat GD consisted of placenta-derived glucocerebrosidase (alglucerase available in 1991, Genzyme Corporation) with modified mannose-terminated glycans, allowing more selective uptake by tissue macrophages, the prominent storage cells in GD [
6‐
8]. This preparation was replaced in 1996 by recombinant enzyme (imiglucerase), which was therapeutically equivalent in terms both of safety and efficacy [
9]. Two new biosimilar agents are now available: velaglucerase-alfa (Shire) [
10] and taliglucerase-alfa (Pfizer) [
11]. ERT reduces macrophagic substrate accumulation, but no routine substrate assay is currently available. A substrate reduction therapy (miglustat, Actelion) that can be prescribed in a few indications has been available since 2002.
The levels of several biomarkers (e.g., in our article, ferritin and chitotriosidase, but also tartrate-resistant acid phosphatase or angiotensin-converting enzyme, not analyzed in this study) change during the clinical course of GD [
12‐
14] due to macrophagic activation: their concentrations rise with disease progression and generally decrease during ERT [
12,
15,
16]. These variations can predict bone complications [
17]. High blood ferritin during the course of type 1 GD may reflect macrophage activation triggered by substrate accumulation, demonstrated by increases in CCL18 and macrophage inflammatory protein-1α or 1β [
18,
19]. ERT is associated with a dramatic decrease of blood ferritin [
16,
17], which is more pronounced in patients with an intact spleen [
20]. Chitotriosidase is massively produced by storage cells and there is a linear relationship between chitotriosidase and glucosylceramide levels, as shown in spleen sections from patients with GD [
21]. Chitotriosidase values drop sharply during ERT, when substrate accumulation decreases, coinciding with clinical improvements [
12].
Hematological abnormalities (anemia and thrombocytopenia) are common in GD, because Gaucher cell infiltration leads to hypersplenism (increased destruction or sequestration of red blood cells or platelets) and bone-marrow insufficiency (decreased production) [
22]. Splenectomy, performed essentially before 1991, increases baseline platelet count and decreases the slope of platelet clearance during ERT [
17]. Anemia and thrombocytopenia can be used as biomarkers to manage GD patients.
Grabowski et al. [
23] developed an Emax model to describe changes in hemoglobin and platelets and in splenic and hepatic volume during ERT of patients in the International Collaborative Gaucher Group Registry. Biomarkers of French GD patients from a single center was modeled before and during ERT by Stirnemann et al. [
17], but no physiological model was proposed to analyze changes in biomarkers levels during ERT. Nonlinear mixed effects models [
24] are widely used to analyze biological processes described by repeated longitudinal data. They allow estimation of the mean value of the parameters and their inter-individual variability. These models allow a sparse sampling design with few data points per individual in a large set of individuals.
The aim of this study was to develop a pathophysiological model explaining the response of biomarkers (ferritin, chitotriosidase, hemoglobin, platelets) to ERT and to analyze the influence of several covariates.
Discussion
Our pathophysiological model predicts changes in biomarkers on ERT and estimates the rate constant of normalization. For the final model, we estimated a normalization half-life of 0.5 years during ERT for all four biomarkers. Only 2 studies have modeled the changes in biomarkers and they used different models: Emax and linear mixed models. In the study of Grabowski et al. [
23], the ERT dose effect had a significant impact on biomarkers with a large sample size. Emax is an empirical model with a hyperbolic function; it is not the result of a physiological model. These models are used in pharmacometrics [
27] to study the link between dose and concentration. However, the drug often acts on a biological quantity, modifying its production or elimination. We modeled this quantity by a model with a production rate and a rate constant of elimination. Using physiological knowledge, we obtained differential equations to explain changes in biomarkers over time. For instance, 90% response is obtained after 9 T
50 for an Emax model, whereas it is reached after only 3.3 half-lives (similar to T
50) for an exponential model.
Our results show improvement in all biomarkers under ERT (decrease in ferritin and chitotriosidase and increase in hemoglobin and platelets). Stein et al. [
20] also highlighted an increase in ferritin and a renormalization under ERT. Hollak et al. [
31] reported a decrease of chitotriosidase of 32% in 1 year, and we found a 95% response in 2 years and a 36% response in 1 year. De Fost et al. [
32] reported that 53% of patients had anemia at baseline and 58% had thrombocytopenia, with renormalization under ERT, and showed a similar pattern of response after 1 year under ERT.
Patients <15 years of age have lower initial concentrations of ferritin and hemoglobin but higher platelet counts; at initiation of ERT, women have a lower concentration of hemoglobin and splenectomized patients have a higher platelet count and ferritin. For GD children from the International Collaborative Gaucher Group Registry, Kaplan P et al. [
33] noted that 50% had platelet counts less than 120 × 10
3/mm
3 and 40% had anemia at the time of diagnosis. Hemoglobin and ferritin tend to be lower in GD women, as for other women, probably because of menstruation, with no link to GD. In our model, no covariates had a significant impact on the chitotriosidase changes. In contrast, Stirnemann et al. [
17] found that splenectomy and GD genotype affected chitotriosidase activity, but their study was limited by a small number of samples, and they used a different model.
Our model allows estimation of the rate constant of biomarker improvement using a pathophysiological model. Bone events are the most debilitating and disabling complication of GD. With substrate overload, Gaucher cells activate and induce proinflammatory cytokine synthesis which can modify the activity of the osteoblast-osteoclast system and promote lytic phenomena and intraosseous vascular complications [
34,
35]. Further analysis of the interaction between biomarkers and bone events is needed.
Recently, plasma glucosylsphingosine has been proposed as a biomarker for GD [
36,
37] and could be used as a reflection of intracellular glucosylceramide. Our model may also predict changes in intracellular glucosylceramide and/or glucosylsphingosine.
Our model could be used to study the effect of biomarker changes on complications such as the occurrence of bone events. Using a few measurements of biomarkers post-treatment, we can estimate the rate constant of normalization and individual amplitude of variation in biomarkers in order to predict further response to treatment. Our model could help to define an individual risk for complications and to refine the best ERT regimens.
Parameters of this pathological model could be used to predict value at steady state with the dosages in the first months of ERT. These predicted values could be used as an individual objective of treatment to modify dosage for each patient. Clinical studies are needed to confirm this hypothesis.
We developed a model which could in the future be used to manage patients, but further studies are needed to confirm clinical applications. This model with 2 parameters (in addition to the baseline value) predicts the level of improvement of biomarkers and the time to 95% response, using only a few measurements per patient.
Acknowledgments
The authors wish to thank the RCLD assistant of the RCLD, Monia Bengherbia, the RCLD secretary, Samira Zebiche, the GD patients and their physicians, who contributed to the FDGR. The development of the original software for the French Gaucher Disease Registry (FGDR) was funded by a grant from the nonprofit organization VML (Vaincre les Maladies Lysosomales). The FGDR was funded, in part, by INSERM and InVS (Institut national de Veille Sanitaire).
Competing interests
FM had consulting fees and INSERM UMR1137 received a grant from Sanofi. JS had travel fees from Sanofi-Genzyme. NB had travel fees, consulting fees, fees for speaking and received grants (Genzyme, Shire, Actelion, Pfizer) donated to the Department of Clinical Research of the Assistance-Publique Hôpitaux de Paris. The other authors declared no financial disclosures.
Authors’ contributions
MV, JS and FM designed the research and analyzed and interpreted the data. MV, JS and FM wrote the first draft of the paper, which was then corrected and approved by all authors.