A matrix-assisted laser desorption/ionization time of flight mass spectrometry reference database for the identification of Histoplasma capsulatum

Abstract

The isolation of the pathogenic fungus Histoplasma capsulatum from cultures together with the visualization of typical intracellular yeast in tissues are the gold standard methods for diagnosis of histoplasmosis. However, cultures are time-consuming, require level 3 containment and experienced personnel, and usually call for an additional confirmation test. Matrix-Assisted Laser Desorption Ionization Time of Flight Mass Spectrometry (MALDI-ToF MS) has been established as a suitable tool for microbial identification in several clinical laboratories. A reference database has been constructed for the identification of H. capsulatum by MALDI-ToF MS by using six H. capsulatum strains previously identified by molecular methods. For validation, 63 fungal strains belonging to the Collection of the Spanish National Centre for Microbiology were tested against the new reference database combined with other commercial and in-house databases. In a blind assay, all H. capsulatum strains (n = 30) were correctly identified by the database and 86.6% had scores above 1.7. Considering both phases of the fungus for the same strain, the most reliable results were obtained with the mycelial phase, with only 13.3% of isolates having scores below 1.7. The new database was able to identify both morphological phases of the fungus. MALDI-ToF technology yields a prompt and simple identification from H. capsulatum yeast forms and early mycelial cultures. It allows for reducing response time and decreasing risk in fungus manipulation.

MALDI-ToF mass spectrometry, Histoplasma capsulatum, fungal identification, reference database, endemic mycoses

Introduction

Histoplasmosis is an endemic mycosis caused by the dimorphic fungus Histoplasma capsulatum.¹ Three varieties of the fungus are recognized, two of which are pathogenic to humans: Histoplasma capsulatum var. capsulatum, which causes classical or American histoplasmosis and is widely distributed but endemic to the valleys of large American rivers, and Histoplasma capsulatum var. duboisii, which is the causal agent of African histoplasmosis and is endemic in equatorial Africa.^2,3 The third, Histoplasma capsulatum var. farciminosum, is distributed in countries bordering the Mediterranean Sea and causes epizootic lymphangitis in horses.⁴ Although there are no autochthonous cases, the incidence of this disease has increased in recent years in Spain due to migration from endemic areas and the transit of travellers proceeding from those areas.⁵ Clinical features of this fungal disease in humans vary depending on the fungus variety involved in the infection. Histoplasma capsulatum var. capsulatum primarily causes pulmonary infection while H. capsulatum var. duboisii affects mainly the skin and subcutaneous tissues.⁶ In both cases, the infection can disseminate in immunocompromised populations, especially human immunodeficiency virus (HIV) positive patients.⁷

Classic microbiological diagnosis is based on isolating the organism in cultures, microscopic examination of fluids and tissues, and serological techniques. Isolating the fungus from culture is a time-consuming technique requiring 3–4 weeks due to the slow growth of the fungus.⁸ Once the fungus has grown properly, the typical tuberculate conidia must be observed to perform a presumptive diagnosis.⁹ However, a definite test should be performed by using other techniques such as molecular identification.^10,11 Moreover, biosafety level 3 (BSL-3) containment is needed for the manipulation of the mycelial form.

In the last decade, Matrix-Assisted Laser Desorption/Ionization Time of Flight Mass Spectrometry (MALDI-ToF MS) technology has revolutionized the microbial identification field in clinical practice. It has been implemented in many routine microbiological diagnostic laboratories and several protocols have been developed for the identification of pathogen microorganisms, especially for bacteria and yeasts.¹² Many efforts have been made for adapting these protocols to filamentous fungi identification. Different authors have been focused on standardization of the extraction procedure,¹³ increasing discrimination among species complexes,^14,15 improving commercial databases,^16,17 and developing new in-house databases to achieve an accurate identification of clinically relevant moulds.^18–21 However, few studies had been performed with BSL-3 fungal pathogens. To our knowledge, there are no entries of H. capsulatum spectra in most used commercial databases and its inclusion in an in-house database has been reported recently by Lau et al.¹⁸ in a reduced number of strains, only in the mycelial stage of the fungus. The aim of this study was to create a reference MS library for an accurate identification of strains of Histoplasma capsulatum by MALDI-ToF technology and validate it with 63 strains previously identified by molecular methods.

Methods

Fungal strains

Six H. capsulatum strains belonging to the Fungal Collection of the Spanish National Centre for Microbiology were used to construct the reference mass spectra database. As shown in Table 1, strains belonging to the three varieties of the fungus and the type strain were included. Mycelial form was represented for all strains and yeast phase was included for three of them. To assess the specificity and to test the accuracy of the MS spectra reference database, a blind set of 63 clinical strains from the above mentioned Fungal Collection were tested. This panel included 33 fungal strains belonging to different fungal species and a total of 30 strains of H. capsulatum. All strains included in the study had been previously identified by sequencing the ITS regions.²²

Table 1.

H. capsulatum strains used to construct the reference database for MALDI-ToF identification.

Strain	IDa	Phaseb	Sample Origin	Geographical Origin	Provenance
CM-2721	Hcc	M/Y	Cervical adenopathy	South America	Clinical origin
CM-4626	Hcd	M	Human leg	Guinea-Liberia border	ATCC 24295
CM-5731	Hcc	M/Y	BAS	Guatemala	Clinical origin
CM-5788	Hcd	M/Y	Biopsy	Africa	Clinical origin
CM-7435	Hcc	M	Soil	Arkansas (USA)	CBS 136.72 (Type Strain)
CM-7436	Hcf	M	Horse	Egypt	CBS 536.84 (Type Strain)

^aIdentification based on molecular analysis and DNA sequencing is given. Hcc, Histoplasma capsulatum var. capsulatum; Hcd, Histoplasma capsulatum var. duboisii; Hcf, Histoplasma capsulatum var. farciminosum. [GenBank accession numbers: KR674032, KR674033, KR674034, KR674035, KR674036, KR674037]

^bBoth morphological phases, M (mycelium) and Y (yeast), are represented in the database for three strains.

BAS, Bronchoalveolar secretions; ATCC, American Type Culture Collection (Manassas, VA, USA); CBS, Centraalbureau voor Schimmelcultures (Utrech, The Netherlands).

Protein extraction protocol from mycelia and yeasts

Protein extraction from all BSL-3 fungi was performed under BSL-3 facilities in compliance with Spanish law (Real Decreto 664/1997). Mycelial forms were subcultured on Sabouraud dextrose agar (SBDA) tubes and grown at 30°C for 2–3 weeks. For yeast phase conversion, mycelium cultures were subcultured on brain heart infusion agar (BHIA) medium containing glutamine (200 nmol/ml) and supplemented with 10% sheep blood for another 2–3 weeks at 37°C. Then, yeast forms were maintained on BHIA medium containing glutamine (200 nmol/ml).

Sample preparation for protein extraction was carried out following manufacturer's instructions (Bruker Daltonics, Bremen, Germany). Briefly, fungal material (a 5 mm diameter piece of mycelium or one to two colonies of yeasts) was carefully collected from the surface of cultures with a sterile inoculating loop and placed in 300 μl of sterile water. Then, 900 μl of absolute ethanol was added. After centrifugation at 13,000 rpm for 3 min, the supernatant was discarded, and the dried pellet was resuspended in 10–20 μl of 70% formic acid plus the same volume of 100% acetonitrile. After a short centrifugation, 1 μl of the supernatant was used to perform mass spectra measurement. When working with BSL-3 fungi, a study of complete inactivation in a set of strains neither included in the in the database nor in the validation panel was performed under BSL-3 facilities after protein extraction. Briefly, protein extracts obtained as explained above were cultured in SBDA tubes at 30°C for 2 months to verify a complete killing of fungi. Once verified the inactivation, all protein extracts were manipulated in a BSL-2 laboratory after protein extraction which was always performed under BSL-3 facilities.

Mass spectra acquisition and H. capsulatum reference database creation

In sum, 1 μl of the protein extract was spotted on a metal target plate and allowed to dry. Then, 1 μl of α-cyano-4-hydroxy-cinnamic acid matrix solution was deposited onto the dried spot leading to co-crystallization. Each sample was spotted eight times, and 1 μl of the bacterial test standard (Bruker Daltonics, Billerica, MA, USA) was included to calibrate each experiment. MS-measurement was performed in triplicate using a Microflex LT Mass Spectrometer (Bruker Daltonics). The system was set in standard conditions: the linear positive mode at a laser frequency of 20 Hz within a mass range from 2,000 to 20,000 Da. The instrument parameter settings were: ion source 1 at 20 kV, ion source 2 at 18.5 kV, lens at 8.5 kV, pulsed ion extraction of 250 ns, and no gating. Spectrum acquisition was performed automatically using the FlexControl software (Bruker Daltonics). All collected spectra from 240 laser shots from different positions of the target spot were subjected to analysis with the FlexAnalysis version 3.3 software package (Bruker Daltonics). After performing quality analysis, all suitable spectra were loaded into Biotyper 3.1 (Bruker Daltonics) where a main spectra (MSP) was created for each of the nine isolates (six in mycelial form and three in yeast form) and each MSP was added to the Biotyper database.

Reference MS database validation

Protein extraction from H. capsulatum strains included in the validation assay was performed on different days of culture growth to establish the minimum optimal time to generate suitable spectra (data not shown). From mycelia, protein extracts were obtained from 7-day-old cultures, whereas cultures of 5 days of growth were used for yeast isolates.

Sixty-three clinical strains were included in the assay: 33 fungal strains belonged to different fungal species and 30 strains were H. capsulatum including both human pathogenic varieties (24 H. capsulatum var. capsulatum and six H. capsulatum var. duboisii strains). Six strains belonging to fungal species causing endemic mycoses (Blastomyces dermatitidis, Coccidioides immitis, C. posadasii, and Paracoccidioides brasiliensis) were included to verify the specificity of the new database.

Spectra acquisition was performed in duplicate in a blind assay and was analysed against the MS reference database combined with other commercial and in-house databases: MALDI Biotyper Database (v. 4.0.0.1, 5627 database entries, Bruker Daltonics), MALDI Database for Aspergillus Section Fumigati (I. Quiles, N. Seara, T. Peláez, J. Mingorance, and J. García-Rodríguez, presented at XII National Mycology Congress, Bilbao, SP, 18 to 20 June 2014) and MALDI Database Expansion for Identification of Yeasts.²³

Only those identifications with two or more matches with strains included in reference databases were considered as a correct identification, otherwise the protein extraction procedure was repeated 2–3 times.

Results

Construction of H. capsulatum reference database

MALDI-ToF analysis of the six strains (nine entries in total) used as reference to construct the database gave spectra with sufficient quality to generate main spectra (MSP) and construct the MS reference database. Of note, spectra obtained for all isolates included in the reference database revealed great diversity (Fig. S1). For the analysis, this new reference database was combined with several databases that included a wide number of fungal species (see Reference MS database validation in Methods section).

Validation of H. capsulatum reference database

MALDI-ToF identification results for the 63 strains included in the validation assay are summarized in Table 2. First, strains of fungal species causing endemic mycoses (B. dermatitidis, C. immitis, C. posadasii, and P. brasiliensis) were not identified by the new extended database because they were neither included in commercial nor in-house databases, showing the specificity of this new database. With the exception of these strains, the overall percentage of correct identification was 91.2% (52/57). Briefly, all strains belonging to “non–H. capsulatum” species were correctly identified except for errors at species level for genus Sporidiobulus, Sporothrix, Mucor, Fusarium, and Trichophyton. The scores in all cases were above 1.7, which is considered by the manufacturer as the cutoff score for adequate genus identification. Regarding H. capsulatum strains, all were correctly identified and 86.6% (26/30) had scores above 1.7. When analyzing the results for mycelial and yeast forms from the same strain, the most reliable identification was observed for the mycelial forms. Scores below 1.7 were observed in 50% of yeast isolates (7/14), whereas only 13.3% were observed in mycelial forms (4/30).

Table 2.

Details of the 63 fungal strains included in the specificity panel and their identification by MALDI-ToF MS when analyzed against the MS reference database for H. capsulatum combined with other commercial and in-house MS databases.

Strain	IDa	Phaseb	NOTES	MALDI-ToF Analysisc
				ID	Log score
CL-52	Candida albicans		ATCC 64551	Candida albicans	2.23
CL-182	Sporidiobulus salmonicolor			Sporobolomyces salmonicolor	1.939
CL-199	Candida kefyr			Candida kefyr	2.154
CL-212	Cryptococcus gattii			Cryptococcus gattii	2.101
CL-1258	Candida krusei		ATCC 6258	Candida krusei	2.364
CL-1787	Candida glabrata		ATCC 90030	Candida glabrata	2.389
CL-1789	Candida parapsilosis		ATCC 22019	Candida parapsilosis	2.383
CL-1791	Cryptococcus neoformans		ATCC 90112	Cryptococcus neoformans	1.897
CL-1844	Candida lusitanie			Candida lusitanie	2.258
CL-2671	Saccharomyces cerevisiae			Saccharomyces cerevisiae	1.864
CL-3736	Trichosporon dermatis			Trichosporon mucoides	2.296
CL-6328	Candida orthopsilosis		ATCC 96139	Candida orthopsilosis	2.234
CL-7127	Candida guilliermondii			Candida guilliermondii	2.279
CL-8796	Candida tropicalis			Candida tropicalis	2.209
CM-10	Trichophyton rubrum			Trichophyton rubrum	1.98
CM-2580	Aspergillus fumigatus		ATCC 204305	Aspergillus fumigatus	2.089
CM-2908	Paracoccidioides brasiliensis	M		No DB
		Y		No DB
CM-2911	Coccidioides posadasii	M		No DB
CM-3112	Mucor circinelloides			Mucor ramosissimus	1.934
CM-3113	Blastomyces dermatitidis	Y		No DB
CM-3114	Blastomyces dermatitidis	M		No DB
CM-3115	Blastomyces dermatitidis	M		No DB
		Y		No DB
CM-3197	Fusarium oxysporum			Fusarium proliferatum	1.865
CM-3508	Aspergillus terreus			Aspergillus terreus	2.095
CM-3509	Aspergillus flavus			Aspergillus flavus	1.711
CM-3578	Microsporum canis			Microsporum canis	2.041
CM-4244	Rhizopus microsporus			Rhizopus microsporus	2.273
CM-7111	Sporothrix schenckii	M		Sporothrix schenckii	1.423
CM-7229	Sporothrix globosa	M		Sporothrix schenckii	1.42
CM-7274	Scedosporium apiospermum			Scedosporium apiospermum	2.039
CM-7095	Scedosporium aurantiacum			Scedosporium aurantiacum	1.829
CM-7325	Coccidioides immitis	M		No DB
CM-7576	Trichophyton mentagrophytes			Trichophyton interdigitale	1.92
CM-2854	Hcc	M		Hcc (Y)	1.714
CM-2906	Hcc	M		Hcf (M)	1.734
CM-2907	Hcc	M		Hcc (M)	2.011
CM-3036	Hcc	M		Hcf (M)	2.09
CM-3153	Hcc	M		Hcc (M)	1.786
		Y		Hcc (Y)	1.768
CM-3221	Hcc	M		Hcc (M)	1.919
CM-3576	Hcc	M		Hcf (M)	1.717
CM-4892	Hcc	M		Hcc (M)	1.706
		Y		Hcc (Y)	1.46
CM-4921	Hcc	M		Hcc (M)	1.896
		Y		Hcc (Y)	1.66
CM-5659	Hcc	M		Hcc (M)	1.815
		Y		Hcf (M)	1.684
CM-5678	Hcc	M		Hcc (M)	1.784
		Y		Hcf (M)	1.51
CM-5692	Hcc	M		Hcc (M)	1.684
CM-5787	Hcc	M		Hcd (M)	1.722
		Y		Hcf (M)	1.823
CM-6015	Hcc	M		Hcd (M)	1.821
		Y		Hcc (M)	1.548
CM-6066	Hcc	M		Hcc (Y)	1.744
		Y		Hcc (Y)	1.998
CM-6305	Hcc	M		Hcf (M)	1.75
CM-6306	Hcc	M		Hcf (M)	1.793
		Y		Hcc (Y)	1.654
CM-6307	Hcc	M		Hcf (M)	1.809
CM-6409	Hcc	M		Hcc (M)	1.854
		Y		Hcc (Y)	1.611
CM-6556	Hcc	M		Hcc (M)	1.642
CM-7001	Hcd	M		Hcf (M)	1.95
		Y		Hcc (Y)	1.926
CM-7255	Hcd	M	CBS 215.53	Hcf (M)	1.592
CM-7256	Hcd	M	CBS 175.57	Hcc (M)	1.473
CM-7257	Hcd	M	CBS 114388	Hcd (M)	1.786
CM-7434	Hcc	M	CBS 287.54	Hcf (M)	1.883
		Y		Hcc (Y)	1.865
CM-7441	Hcc	M		Hcc (M)	1.873
CM-7450	Hcc	M		Hcc (M)	1.721
		Y		Hcc (Y)	2.108
CM-7514	Hcc	M		Hcc (M)	1.743
		Y		Hcc (M)	1.703
CM-7704	Hcd	M		Hcd (M)	1.995
CM-7767	Hcd	M		Hcd (M)	1.99

^aIdentification based on molecular analysis and DNA sequencing is given. Hcc, Histoplasma capsulatum var. capsulatum; Hcd, Histoplasma capsulatum var. duboisii; Hcf, Histoplasma capsulatum var. farciminosum.

^bBoth morphological phases, M (mycelium) and Y (yeast), are detailed for dimorphic fungi.

^cBest match log score and identification provided by the system for MALDI-ToF analysis are included for each strain. Each strain was tested in duplicate and 2–3 repetitions of the assay were performed. In case of H. capsulatum strains, the morphological state of the best match log score is also specified between brackets. For strains not represented in any database, neither identification nor log scores are listed.

ATCC, American Type Culture Collection (Manassas, VA, USA); CBS, Centraalbureau voor Schimmelcultures (Utrech, The Netherlands); No DB, strain not represented in databases analyzed.

Considering all isolates of H. capsulatum included in the validation assay (yeast and mycelial forms), identification scores obtained for the 44 isolates assayed were distributed around a mean of 1.78 with a standard deviation of 0.15 (Fig. S2). Depending on the selected cutoff, sensitivity values varied from 75% (cut-off of 1.7) to 100% (cutoff of 1.4); however, specificity was 100% in all cases (Table 3). Of note, we were not able to achieve yeast-phase conversion for 16 H. capsulatum strains.

Table 3.

Sensitivities and specificities according to different MALDI-ToF cut-off values for all H. capsulatum isolates (yeast and mycelia).

Cut-off value	Sensitivity (%)	Specificity (%)
1.4	100	100
1.5	95.5	100
1.6	88.6	100
1.7	75	100
>1.7	40.9	100

Discussion

Classical diagnosis of histoplasmosis has been based on culture and histopathological analysis of tissue biopsies. Culture is time-consuming and needs to be confirmed by other methods. Histopathological analysis may also lead to misidentification with other microorganisms such as Leishmania spp.²⁴ Moreover, both methods require skilled personnel for a correct identification of fungal structures. Serological tests are definitively not informative since immunosuppressed patients show a high number of false negatives and people coming from endemic regions could give false positive results. Antigen detection is a useful method in disseminated infection but is not available in many clinical laboratories.²⁵ Several molecular methods have also been developed recently, but there is no consensus among laboratories and there are not up-to-date commercial methods available.²⁶ In recent years, MALDI-ToF technology has proven to be a reliable and adequate tool for the identification of pathogenic microorganisms. Furthermore, this technique does not require experienced technicians and it reduces costs as well as time of response. Many research groups have focused their work on standardizing fungal species identification by MALDI-ToF technology, especially for species of the genera Candida and Aspergillus,^27,28 but there are some areas that have not yet been explored. A large number of fungal pathogenic species is not represented in most commonly used MS databases and discrimination among some species is not sufficiently defined.

Regarding BSL-3 fungal pathogens, studies on MALDI-ToF applications are scarce, probably due to the difficulty of working under BSL-3 facilities with these fungi, which are also slow-growing. In this work, all protein extracts from H. capsulatum cultures were obtained in BSL-3 facilities and an inactivation test was performed before handling them in BSL-2 facilities. To the best of our knowledge, we have developed the first reference MS database for the identification of H. capsulatum using MALDI-ToF technology that includes both morphological stages of the fungus. Strains used for reference database construction had been previously identified by sequencing the ITS region. Some of these strains were acquired from international reference collections of microbial cultures. This database has proven to be specific and reliable for identifying H. capsulatum by MALDI-ToF. All strains of H. capsulatum were correctly identified, and 86% reached the reliability score for genus identification indicated by the manufacturer. Closely related and dimorphic fungal species, such as B. dermatitidis, P. brasiliensis, and C. immitis, did not produce any results since they were not represented in the new extended database. This fact supported the high specificity achieved for H. capsulatum species identification. Regarding mycelial and yeast forms from the same strain, the most reliable results were obtained for mycelial forms, as 86.6% of isolates had scores above 1.7. Fifty percent of yeasts gave low scores between 1.4 and 1.7.

Identification scores obtained for the 44 isolates from H. capsulatum were distributed around a mean of 1.78 with a standard deviation of 0.15 (Fig. S2). Scores between 1.7 and 2.0 are considered reliable genus-only identification by the manufacturer. However, several authors have considered these scores to be very conservative for fungi,^23,29–33 proving an increase of the sensitivity when lowering the “secure species” score, whereas the accuracy of the technique was always preserved. Based on these studies, an analysis of sensitivity and specificity was performed testing different cutoff scores (Table 2). When decreasing the cutoff to 1.6, sensitivity rose drastically to 93.3% in H. capsulatum strains.

The poor scores obtained with some yeast isolates, even when repeating protein extraction, could be explained by the difficulty to obtain a pure conversion from mycelium to yeast. In fact, yeast phase conversion was achieved only for nearly 50% of all H. capsulatum strains included in this work. Furthermore, the limitations of working under BSL-3 facilities, do not allow changing either culture conditions or protein extraction procedures. On the other hand, even if the extraction method probably was not the most adequate for the best results, it allowed the complete killing of the fungi which is essential to assure the safety of the worker.

The new reference database was not able to perform a proper discrimination between the three varieties of the fungus: only 28 of the 44 isolates (63.6%) included in the specificity panel matched with their correct variety. This fact can be explained by the complex phylogenetic relationships among the three varieties, as described previously by other authors^34,35 who suggested that the traditional classification in three varieties is meaningless. In addition, as explained previously, the standard extraction method used was not optimal for this kind of fungi and could affect the discrimination power of the technique. In clinical settings, the identification of the varieties is irrelevant as treatment would be the same regardless.

Spectra from yeast and mycelial stages of the fungus were included in the database with the purpose of identifying both phases properly. In early experiments, we found that both phases from the same strain gave completely different spectra (Fig. S1). That may be caused by differences in gene expression profiles in yeasts and mycelial forms which has been described by Edwards et al.³⁶ Looking at the results, there was partial cross-identification between both phases: seven isolates out of 44 (15.9%) in a morphological phase matched with strains in the other phase represented in the database. The reason of those observations could be that these strains were not completely converted to the proper phase.

In this work, young cultures (7 days of growth for mycelia and 5 days for yeast) have been used in the validation assay with suitable results. This represents a meaningful decrease of time compared to classical identification methods based on microscopic observation of cultures. At 7 days of growth, most mycelial isolates included in the assay did not present the characteristic tuberculate macroconidia for their identification.

Finally, despite the difficulty of working with this kind of slow-growing fungi in BSL-3 conditions, we achieved an excellent specificity by performing the validation in the most similar conditions to a real clinical setting (blind assay, testing only two duplicates, using manufacturer's instructions for extraction procedure, etc.). This goal and the capacity of this method to anticipate H. capsulatum identification from cultures in contrast to classical methods, highlight the potential of MALDI-ToF technology to improve diagnosis of histoplasmosis and other BSL-3 organisms.

In conclusion, our reference MS database is robust and suitable for H. capsulatum identification in both morphological phases of the fungus and allows advancing in the diagnosis of histoplasmosis. Furthermore, minimal handling of H. capsulatum cultures is needed for performing protein extraction, reducing biosafety risks. This technique could be performed in clinical and microbiological diagnostic laboratories of hospitals and other research institutions with proper biosafety containment. However, more studies about the taxonomy of the fungus and the refinement of several parameters to improve the construction of MALDI-ToF MS reference spectra libraries are required.

Supplementary material

Supplementary data are available at MMYCOL online.

Acknowledgements

We cordially thank Frank Hodgkins for his careful reading and editing of the manuscript.

This work was supported by research projects PI11/00412 and PI14CIII/00045 from the Spanish Fondo de Investigaciones Sanitarias of the Instituto de Salud Carlos III.

C.V. was supported by research fellowships from the Fondo de Investigaciones Sanitarias of the Spanish Ministry of Economy and Competitiveness (FI12/00095).

S.G. was supported by a research fellowship from the Fondo de Investigaciones Biomedicas of the Spanish Ministry of Science and Innovation (FI10/00464).

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and the writing of the paper.