Abstract
Metal-on-metal (MoM) hip prostheses are known to release chromium and cobalt (Co), which negatively affect the health status, leading to prosthesis explant. Albumin (ALB) is the main serum protein-binding divalent transition metals. Its binding capacity can be affected by gene mutations or modification of the protein N-terminal region, giving the ischaemia-modified albumin (IMA). This study evaluated ALB, at gene and protein level, as marker of individual susceptibility to Co in MoM patients, to understand whether it could be responsible for the different management of this ion. Co was measured in whole blood, serum and urine of 40 MoM patients. A mutational screening of ALB was performed to detect links between mutations and metal binding. Finally, serum concentration of total ALB and IMA were measured. Serum total ALB concentration was in the normal range for all patients. None of the subjects presented mutations in the investigated gene. Whole blood, serum and urine Co did not correlate with serum total ALB or IMA, although IMA was above the normal limit in most subjects. The individual susceptibility is very important for patients’ health status. Despite the limited results of this study, we provide indications on possible future investigations on the toxicological response to Co.
Introduction
Hip prosthesis surgery is widely used to recover articular functionality. Sliding surfaces can consist in different wear-resistant materials, more commonly a chromium (Cr)-molybdenum (Mo)-cobalt (Co) alloy (for balancing in ISO 5832-12:2007 High Carbon Alloy). Patients with hip prosthesis with metal-on-metal (MoM) coupling represent a category of metal-exposed subjects with an endogenous and continuous exposure source. MoM coupling is subject to wear with consequent release of micro- and nano-metal particles that not only can cause local inflammation but can also produce ionic forms of Co and Cr, 1 which are inducers of oxidative stress 2,3 and have specific toxicity properties. Cr toxicity is mainly related to its hexavalent form (CrVI), which is not released by the prosthesis coupling, but which is considered cancerogenic if inhaled. Co in high doses can induce heart, thyroid and central nervous system alterations. Several studies have investigated the effects of Co accumulation, also known as cobaltism, due to occupational exposure 4 –6 or to ingestion. 7 Co threshold value for professional exposure (mainly by inhalation) is set at 1.5 µg/L in urine, 1µg/L in whole blood and 0.3 μg/L in serum, as measured at the end of a work shift at the end of a working week. 8
Conversely, no clear indications exist on the estimated risk of cobaltism in patients with MoM prosthesis who, according to the Medicines and Healthcare Products Regulatory Agency (MHRA), should be monitored when bivalent Co (Co(II)) values exceed 7 μg/L in whole blood. 9 However, no surveillance threshold is set for urine. In literature, several cases are reported about MoM patients presenting symptoms suggestive of cobaltism, with Co blood levels ranging from 15 μg/L to 625 μg/L, 10,11 thus suggesting that there are specific transport mechanisms, variable between subjects, accounting for the different circulating Co level and different degree of symptoms.
Albumin (ALB; Figure 1) 12,13 is the most abundant plasmatic protein, representing 60–65% of serum proteins. 14 It is the main transport protein for transition metals including Co, 15 –18 as well as having many functions, 19 ranging from maintaining oncotic blood pressure and pH buffering, 14 to binding and transport of hormones, drugs and free fatty acids. 14,20 Genomic and/or post-translational mutations of ALB can reduce its binding affinity with metal ions and, consequently, increase their concentrations in blood and urine. 21 In turn, these ions can take part in reactions leading to the formation of reactive oxygen species (ROS), 22 known to be harmful for cells. 23
Translation of CDS leads to pre-proalbumin, a 609 aa product. In the RER, the first 18 N-terminal aa (lined area) are removed to give proalbumin, a 591 aa product. In the Golgi apparatus, further 6 terminal aa (dotted area) are removed leading to ALB, the active and circulating protein with 585 aa. In squared area are the 4 aa that are sometimes removed leading to the formation of IMA. CDS: coding DNA sequence; aa: aminoacid; RER: rough endoplasmic reticulum; IMA: ischaemia-modified albumin; ALB: albumin.
Non-denaturizing electrophoresis analysis of circulating ALB has identified several proteic isoforms 24 –26 caused by single nucleotide polymorphisms (SNPs) in the genic sequence of ALB, with an estimated frequency of 1:1000 and 1:3000 in the population. These cause missense mutations in the aminoacid (aa) sequence of the protein, compromising its binding to metal ions 25,26 directly (SNPs in positions 147, 152 and 154 of the reference genic sequence NM_000477.5) or indirectly (SNPs in positions 140, 141 and 144, leading to proalbumin, an immature form of the protein; Table 1, Figure 1). In literature, no mutation involving cysteine 34 (Cys34) and hystidine 67 (Hys67) (numbering refers to mature ALB) is known, despite these aa are involved in metal ion binding (Table 2, Figure 1) 27 .
aProalbumin mutations.
bAlbumin mutations.
cBase numbering referred to the albumin gene reference sequence (NM_000477).
dAminoacid numbering referred to the pre-proalbumin protein reference sequence (NP_000468.1) and aminoacid numbering (in brackets) referred to the mature albumin (585 aminoacids).
Two aminoacidic residues involved in metal binding but with unknown mutations.27
aBase numbering referred to the albumin gene reference sequence (NM_000477).
bAminoacid numbering referred to the pre-proalbumin protein reference sequence (NP_000468.1) and aminoacid numbering (in brackets) referred to the mature albumin (585 aminoacids).
ALB can also undergo post-translational mutations caused by pathological/traumatic events such as trauma, sclerodermia, diabetes, bacterial or viral infections, end-stage renal disease, liver cirrhosis, brain ischaemia, peripheral arterial disease, myocardial ischaemia and cancer. 28 In these conditions, ALB undergoes a proteolytic cut in the first 4 N-terminal aa of the active protein, known to be involved in the binding with metal ions, 29 leading to the formation of ischaemia-modified albumin (IMA), 30,31 which was the specific target for this study, together with native ALB (Figure 1).
The binding power of ALB and the knowledge of its possible modifications have led to hypothesize that this protein could have a role in the determination of individual susceptibility to metal ions, namely in individual response to accumulation, in transport and elimination of metal ions, and consequently in the onset of systemic symptoms in patients with MoM prosthesis, such to require an explant. 32
We designed an observational, analytical, non-parallel cohort study with prognostic query in patients with MoM hip prosthesis in order to establish, if ALB modifications were linked to a greater risk of explant and could therefore be a marker for individual susceptibility to Co. To this end, circulating and urinary Co levels were measured together with the total serum concentrations of ALB and IMA. Finally, ALB mutations and/or post-translational modifications were analysed as possible cause of a reduced transport capacity of metal ions and therefore higher concentrations of such ions in the blood and urine of the analysed patients.
Methods
Patients’ enrolment
Forty patients with MoM hip prosthesis attending the Orthopaedic Traumatology and Prosthetic Surgery and Revisions of Hip and Knee Implants of Rizzoli Orthopeadic Institute were enrolled.
None of the patients reported symptoms of systemic toxicity. Exclusion criteria were the presence of other medical articular devices and sepsis (suspect or clear). The study population included 24 female and 16 male patients with mean age of 62.7 ± 9.7 (mean ± standard deviation (SD)) and 57.2 ± 10.1 years, respectively. The mean time from implant was 5.3 ± 2.8 years for female patients and 6.8 ± 3.4 years for male patients (mean ± SD). The study was approved by the Ethics Committee of Rizzoli Orthopeadic Institute and all patients signed an informed consent form.
All procedures performed in the study were in accordance with the World Medical Association Declaration of Helsinki.
Samples collection
Peripheral blood samples were harvested from 40 patients using a disposable intravenous cannula, withdrawn and transferred into trace element BD-vacutainer tubes (Becton Dickinson, Franklin Lakes, New Jersey, USA) containing ethylenediaminetetraacetic acid to collect whole blood.
Five millilitre of blood were immediately frozen and stored at −80°C until the analysis for Co determination was performed. An additional 0.5 mL aliquot was stored at −80°C until the RNA extraction. Other 5 mL were centrifuged at 800 relative centrifuge force for 7 min, frozen and stored at −80°C until the Co serum dosage, IMA enzyme-linked immunosorbent assay (ELISA) and ALB total determination were performed.
On the same day, clean-catch urine samples (10 mL) were collected in universal sample pots, frozen and stored at −20°C until the quantification of Co ions and creatinine was carried out.
Circulating and urinary Co measurement
Inductively coupled plasma mass spectrometry (ELAN DRC II; Perkin Elmer, Waltham, Massachusetts, USA) equipped with dynamic cell reaction was used for the measurements. Blood and serum samples were diluted with Triton X-100 0.05% while urine samples were diluted with bi-distilled water, for inorganic trace analysis (Merck KgaA, Darmstadt, Germany).
The calibration curve was prepared by dilution of a standard solution ranging from 0.5 mg/L to 1000 mg/L (Co in nitric acid – HNO3 2% mono elemental standard solution; Carlo Erba Reagenti, Milano, Italy). The accuracy of the method was determined on the basis of the mean values obtained on certified reference materials (Environmental and Occupational; German External Quality Assessment Scheme (G-EQUAS)) for blood. The coefficients of variation ranged from 4% to 8% and the limit of detection was 0.05 µg/L in all matrices. Our laboratory participates in the inter-comparison programme for toxicological analysis in biological materials for the determination of Co (G-EQUAS of the German Society of Occupational and Environmental Medicine).
Urinary creatinine was determined by a modified Jaffè reaction (ILab 350 Clinical Chemistry System; Instrumentation Laboratories SpA, Bedford, Massachusetts, USA). The exclusion criteria of the American Conference of Governmental Industrial Hygienists recommendation for very diluted (creatinine concentration less than 0.3 g/L) or very concentrated (creatinine concentration greater than 3.0 g/L) urine samples were adopted. 33 Co ions levels measured in all three matrices were used in the correlation analyses, as value measure at the time of collection (μg/L) and as Co released over a period of a year (value at sampling/duration of implant; in μg/L/year), assuming that the release of the ion from the prosthesis was constant in time.
Total serum ALB measurement
Total ALB measurement was performed using the colorimetric assay BCG (Bromocresol Green) ALB Assay kit (Sigma-Aldrich Co., St Loius, Missouri, USA).
Serum samples stored at −80°C were thawed at room temperature and the assay was performed following the manufacturer’s instructions.
ALB gene mutations assay
RNA extraction, reverse transcription PCR and ALB cloning
The 0.5 mL blood aliquots were thawed out with 1 mL of TRIzol (Invitrogen, Carlsbad, California, USA) in ice. The aqueous phase containing RNA was isolated following the TRIzol manufacturer’s protocol (Invitrogen) and the total RNA was purified following the clean-up protocol of the RNeasy Mini Kit (QIAGEN, Valencia, California, USA). RNA quantity was determined by RNA gel electrophoresis in 1% agarose gel in TAE 1X (Merck & Co., Whitehouse Station, New Jersey, USA) stained with 0.5 µg/mL etidium bromide (Sigma-Aldrich Co.) and visualized with UV light, using RiboRuler High Range Ladder (Thermo scientific, Waltham, Massachusetts, USA).
RNA was subjected to RT using the following conditions: 1 µg total RNA, Moloney murine leukaemia virus reverse transcriptase (Promega, Madison, Wisconsin, USA; used with companion buffer) 200 U, oligo dT-15 2.5 µM, random hexamers 2 µM and dNTPs 500 µM each. The RT reaction was performed in a final volume of 20 µL for 60 min at 37°C. In order to verify that the RT reaction was successful, amplification of the human Glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene was performed, using specific primers (GAPDH forward: 5′-GAAATCCCATCACCATCTTCCAG-3′; GAPDH reverse: 5′-AGGAGACCACCTGGTGCTCAGTGTAGC-3′). GAPDH amplification was performed in a final volume of 25 µL, using the following conditions: 1 µL cDNA, 0.2 µM each primer, 12.5 µL BioMix Red (Bioline, Taunton, Massachusetts, USA); initial denaturation for 2 min at 94°C; 25 cycles of 30 s at 94°C, 30 s at 61°C, 30 s at 72°C followed by a final extension for 7 min at 72°C.
In order to clone ALB sequence, specific primers were designed using Amplify software 34 and the NCBI (National Centre for Biotechnology Information) primer design tool, 35 following standard criteria 36 : ALB forward primer, 5′-TTTCTCTTCTGTCAACCCCACA-3′ and ALB Reverse primer, 5′-GAAGCATTCATTTCTCTCAGGTTC-3′.
ALB cloning was obtained using polymerase chain reaction (PCR) standard conditions (as described above), except that there were 5 µL cDNA and 45 cycles with an annealing temperature of 63°C. GAPDH and ALB products were analysed in a 1.5% agarose gel as described above.
Ten SNPs comprising from position 140 to 342 of NM_000477.5 reference sequence were summarized in Table 1 and investigated in the samples. The first five listed mutations referred to proalbumin, while the others to ALB. 25,26
In Table 2, two aminoacidic residues and their corresponding codons in NM_000477.5 were reported. No mutations referred to these aa are known.
Amplicon purification, DNA sequencing and sequence analysis of ALB
ALB amplicons were purified with GenElute PCR clean-up kit (Sigma-Aldrich Co.) following the manufacturer’s recommendations. All sequences were determined using the Big Dye Terminator Cycle Sequencing-Ready Reaction kit and automated DNA sequence analyser ABI-PRISM 3730 (Applied Biosystems, Foster City, California, USA). Each sample was sequenced twice and sequences were analysed by Basic Local Alignment Search Tool family programs – accessed via the NCBI homepage 37 with default parameters. Sequence electroferograms were analysed using 4Peaks software. 38
IMA measurement by ELISA assay
Human IMA serum concentrations were determined with a commercially available ELISA kit and performed according to the manufacturer’s instructions (CUSABIO BIOTECH Co., ltd, Wuhan, P.R. China).
Approximately 100 μL of standards or samples were added to a microtiter plate pre-coated with an antibody specific to Human IMA, and incubated for 2 h at 37°C. After aspiration, Biotin-antibody was added to each microplate well and incubated for 1 h at 37°C. Aspiration and washes were performed. Horseradish peroxidase-avidin was added into each well for 1 h at 37°C and, then, washed again. Finally, 3,3′,5,5′-tetramethylbenzidine substrate solution was incubated for 15 min and stopped with the Stop solution provided (sulphuric acid). The colour changes were measured spectrophotometrically at a wavelength of 450 nM. Quantifications were achieved by the construction of standard curves using the known concentrations of IMA and the data analysis was performed by using a four-parameter logistic curve fit, obtaining results expressed in IU/mL.
The kit has a detection range of 3.12–200 IU/mL, with an intra-assay precision < 8% (coefficient of variability, CV%) and inter-assay precision < 10% (CV%). The normal value of IMA assessed by this kit was estimated to be 10 IU/mL.
Statistical analysis
Patients were divided in two groups depending on the Co levels measured in whole blood: Co > 7 μg/L and Co < 7 μg/L (named for the purpose of this study as high Co and low Co). The Mann–Whitney U Test with independent samples was used to evaluate the differences in total ALB and IMA values in patients stratified by high Co and low Co groups. Pearson correlation was performed to study the correlation between Co in the different matrices and the variables analysed (total ALB and IMA). Correlation analyses were performed both on Co annual values in the three matrices and on the values measured on the day of sampling.
The statistical analysis was performed using SPSS versus 14.0 software (SPSS Inc., IBM, Chicago, Illinois, USA). Results were considered statistically significant with p value < 0.05.
Results
Measurement of blood and urine Co
In our sample, 22 subjects of 40 had a Co value in whole blood higher than 7 μg/L, which is the alert value in patients with MoM prosthesis according to MHRA. 9 For this reason, we divided the patients in two groups, high Co and low Co. Table 3 shows the mean Co levels at sampling in whole blood, serum and urine (normalized by creatinine concentration) and the calculated annual values, all stratified by high Co and low Co group. In order to characterize our patients sample, we also considered as reference value for normal Co concentration, the one used in occupational medicine for non-exposed population which is 1 μg/L in whole blood and 0.3 μg/L in serum. According to these values, all patients analysed (37/37, because of 3 missing serum samples) had Co serum values above this cut-off, while whole blood measurement showed 32 out of 40 patients with Co concentration above the cut-off value (data not shown).
Distribution of Cobalt concentrations in different matrices.
Co: cobalt; B: blood; S: serum; U: urine; n = number of subjects with blood Co > 7 μg/L (high Co group) or < 7 μg/L (low Co group) for which the measurement was possible.
aMean ± standard deviation.
Evaluation of ALB SNPs
We were able to amplify by reverse transcription-PCR, in 30 of 40 samples, a region of 412 base pairs (bp) of ALB, corresponding to the sequence in position 39–451 of NM_000477.5. In all cases, the PCR products electrophoresis showed amplicons of the expected length. By sequencing, we evaluated the presence of possible SNPs reported in literature, and point mutations in the sequence coding for Cys34 and Hys67, of the aa sequence of mature ALB (Tables 1 and 2).
All sequences have been analysed in duplicate. All samples analysed showed a correspondence with NM_000477.5 in all considered positions (Tables 1 and 2). However, two samples showed heterozygosis and transversion not present in the NCBI (dbSNP) database for ALB.
Measurement of total ALB and IMA
Serum total ALB and IMA were measured in 37 of 40 subjects because of missing serum samples for three of them. Table 4 shows the concentration of total ALB and IMA, as mean ± SD for the high Co (n = 20) and low Co (n = 17) groups. IMA values, normalized according to Lippi et al. 39 (individual serum ALB concentration/median ALB concentration of the population × IMA value), are also reported in the table.
Distribution of serum albumin and ischaemia-modified albumin values in patient with MoM prosthesis, grouped in high Co (>7 μg/L) and low Co (<7 μg/L).
ALB: albumin; IMA: ischaemia-modified albumin; Co: cobalt.
aMean ± standard deviation.
Total ALB did not show significant variation compared to the normal population (normal range 3.5–5.2 g/dL; Table 4) and no statistically different distributions were found between high Co and low Co groups (p = 0.270; Mann–Whitney test).
The manufacturer of the IMA ELISA kit indicated 10 IU/mL as normal reference value for human; in a parallel study from our group in a population without prosthesis, we confirmed this value (10.8 IU/mL, data not shown). Considering this value as cut-off, 15 patients of 17 (88%) in the low Co group, and 15 patients of 20 (75%) in the high Co group, had IMA values above the normal value. The different distribution of IMA concentrations between the two groups was not statistically significant (p = 0.257), even when considering the normalized value (p = 0.117; Mann–Whitney test).
These analyses were carried out taking the occupational medicine Co limit values into account, and 29 of 37 patients with serum Co > 0.3 μg/L (78%) also had an IMA value above the normal limit. Likewise, considering the whole blood limit of 1 μg/L (32 subjects), 21 of 29 for which it was possible to perform the assay (72%), had an IMA value above the normal limit.
Correlation between Co in the different matrices, total ALB and serum IMA
Pearson test was used to correlate total ALB and the concentrations of Co in whole blood, serum and urine, both values at sampling and annual values. No correlation was found. Likewise, no correlations were found between IMA values (raw and normalized) and Co concentrations in whole blood, serum and urine, both values at sampling and annual values.
Discussion
Hip prosthesis with MoM coupling represents an important and specific source of metal ions release, such that patients with MoM prosthesis are recognized between the categories at risk of metallosis, and in particular of cobaltism. In this population, the source of release is endogenous and continuous, therefore the individual susceptibility is paramount to understand and estimate the individual response to metal ion accumulation. Moreover, up to date, the toxicity of this endogenous and specific way of exposition has not been well characterized and defined.
Physiologically, Co is bound to ALB, which has high affinity binding sites for metals 27 and for this reason has been identified as a potential marker of individual susceptibility, 32 intended as the ‘managing’ protein of released Co. Gene point mutation of ALB as well as post-translational modification of this protein, can reduce ions ‘captation’; consequently, free circulating ions could take part in the formation of ROS and cause cellular oxidative damage. Despite this, the effects of SNPs and the presence of IMA do not generally affect the health status of patients macroscopically. 25,26
In this study, a mutational screening has been performed from position 39 to 451 of ALB gene in 30 out of 40 patients with MoM hip prosthesis. In fact this region of the gene contains positions sensitive to point mutations: these involve triplets coding for the N-terminal aa of mature ALB deputed to metal binding, and triplets involved in the stability of the protein (Table 1). This screening allowed also to evaluate if SNPs were present in some nucleotidic position correlated to proalbumin, the immature form of ALB (Figure 1, Table 1). The presence of such SNPs reduces the possibility of a proteolytic cleavage of proalbumin aa, therefore the circulating form is different from the mature form, indirectly affecting the binding ability of ALB itself. 25,26 The analysis of the sequence did not reveal the presence of SNPs in all positions considered (Table 1). Similarly, codons for Cys34 and Hys67 (Table 2), involved in metal ion bindings, 27 corresponded to the reference sequence, in line with the absence in literature of known SNPs in those positions.
This study therefore allows us to exclude any correlation between Co level and the presence of SNPs in the ALB region studied. However, in view of our experimental design and the small sample size due the characteristics of enrolled subjects, we cannot exclude the presence of other mutations, on other regions of the ALB gene that would disrupt the mechanism of binding and ion transport. The cloning performed in the present study is in fact limited to the 5’-end region of the gene ALB, which is, according to the literature, the portion directly involved in metal binding. Therefore, nothing excludes that mutations in other not investigated portions could hamper, for example, the proteic folding that allows the interaction between aa far apart in the primary sequence for the protein. Consequently, if there were gene mutations beyond position 451 of NM_000477.5, the change of proteic domain and/or tertiary structure of ALB could be modified.
Functional alterations of ALB can be of endogenous origin (genetic) or induced, such as in some diseases, in inflammatory conditions and stress. The experimental design of this study, in addition to investigating the genic aspect of individual susceptibility to the management of Co release from MoM prosthesis, also had the aim to quantify IMA, a form of ALB in which the N-terminal aa (Asp-Ala-His-Lys) have been altered so as to be unable to bind Co and the other transition metals. 40 The design of the study also aimed at analysing the correlation between Co levels in the studied matrices and modified ALB, considering blood, serum and urinary Co level at the date of sampling and the calculated annual release. This last analysis was based on the assumption that Co was released from the prosthesis with a constant modality. By evaluating IMA distribution in the patients groups with high and low Co, defined by MHRA guidelines, there were no statistically significant differences in IMA between the two groups, despite these levels were, in average, higher of the limit indicated by the manufacturer of the ELISA kit (Table 4). By considering Co occupational medicine reference values (i.e. 1 μg/L in whole blood and 0.3 μg/L in serum), we have demonstrated that, where Co level was beyond the specific cut-off, often IMA level was above the normal range, suggesting a correlation between Co and IMA concentrations. However, no statistically significant correlation was found between IMA and Co levels in blood, serum and urine at the time of sampling or annually derived, excluding the hypothesis that, in our sample, the presence of this protein, modified and unable to bind metal ions, would affect circulating Co concentrations.
The same kind of investigation was carried out with serum total ALB, hypothesizing that an altered concentration of this protein would affect Co transport and therefore its concentration in the body matrices considered. Even in this case, no correlation was found between ion concentrations and serum total ALB. In addition, serum total ALB, as well as IMA values, did not show statistically significant differences between patients with high Co and low Co, excluding an involvement of these two proteins in determining blood Co level. Despite the lack of correlation between ALB/IMA concentrations and Co levels, the presence of IMA above the normal limit in most of the patients analysed, does not allow to rule out the role of IMA for the transport of other transition ions that have not been investigated in this study, but correlated with the presence of the prosthesis. More in general, we cannot exclude that modified ALB could play a role on patients health status, within molecular pathway not yet identified and investigated.
In conclusion, the analyses carried out in our sample of patients have highlighted that the measured levels of Co in whole blood, serum and urine do not correlate with genic mutation nor post-translational modification in the considered region of ALB and do not correlate with serum ALB concentrations. However, it remains evident the individual susceptibility plays a key role in the patients’ health status and therefore, the results of the present study, despite their limitations, can give indications on possible future investigations on different aspects, as well as on a wider gene and protein analysis of ALB itself.
Footnotes
Acknowledgements
The statistician Dr Barbara Bordini is gratefully acknowledged for her contribution. We also would like to thank the orthopaedic surgeons and nursing staff of the Prosthetic Surgery and Revisions of Hip and Knee Implants Division of Rizzoli Orthopaedic Institute of Bologna (Italy) and Dr Marilina Amabile for their contribution for collecting the samples. The authors are grateful to Gabriella Mattei and Michela Bonaguro for their excellent technical assistance with automated sequencing. The authors wish to thank Dr Lucia Mancini for her help in the English revision of the manuscript.
Authors’ Contribution
Federica Facchin and Simona Catalani equally contributed to the study.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Italian Ministry of Health, ‘Early diagnosis of pending failure in hard bearings’ (grant no. RF-2009-1472961) and by the Fondazione del Monte di Bologna e Ravenna.