Four new cases of stomatin-deficient hereditary stomatocytosis syndrome: association of the stomatin-deficient cryohydrocytosis variant with neurological dysfunction


Dr G W Stewart, Department of Medicine, University College London, Rayne Institute, University Street, London WCIE 6JJ, UK. E-mail:


This report concerns congenitally Na+–K+ leaky red cells of the ‘hereditary stomatocytosis’ class. Three new isolated cases and one new pedigree are described, and one previously reported case is expanded. In all cases, Western blotting of red cell membranes revealed a deficiency in the 32 kDa membrane protein, stomatin. All showed pronounced cation leaks at 37°C with markedly abnormal intracellular Na+ and K+ concentrations, like all other such stomatin-deficient cases. Consistent with recent findings in two previously described British pedigrees, immunocytochemistry demonstrated that the deficiency of stomatin was not complete. On typical blood films, some red cells showed positive stomatin immunoreactivity, while most were negative, although in one case only a minority were negative. All platelets and neutrophils were stomatin positive. The cases differed markedly between themselves with regard to the temperature dependence of the passive leak to K+. Three showed a simple monotonic temperature dependence, while two showed a minimum at around 20–25°C, such that the cells were extremely leaky at 0°C, giving the phenotype known as ‘cryohydrocytosis’. These patients are the only two known cases of stomatin-deficient cryohydrocytosis. Both showed a congenital syndrome of mental retardation, seizures, cataracts and massive hepatosplenomegaly, probably defining a new haemato-neurological syndrome.

The ‘hereditary stomatocytoses and allied disorders’ (HSt) represent a diverse group of dominantly inherited haemolytic anaemias, all of which show a leak to Na+ and K+ across the membrane of the red cell (Delaunay et al, 1999; Stewart & Turner, 1999; Stewart, 2003). The conditions show very marked heterogeneity. The present report focused on two features: the 32 kDa stomatin protein, which can be present or absent (Lande et al, 1982; Hiebl-Dirschmied et al, 1991; Stewart et al, 1992), and the dependence on temperature of the Na+–K+ leak (Coles & Stewart, 1999).

The term ‘cryohydrocytosis’ is used to describe a variant of these leaky conditions in which the cells show very marked swelling and lysis when stored at 0 or 4°C (Miller et al, 1965). To date, this behaviour has always been associated with a U-shaped temperature dependence on the part of the ‘passive leak’ to Na+ and K+, such that there is a minimum in the leak at about 23°C. The leak is much greater at 0°C than it is at 37°C (Coles et al, 1999a; Haines et al, 2001). In the original report of stomatin deficiency in leaky red cells (Lande et al, 1982), one case was labelled cryohydrocytosis and the other not, implying that stomatin deficiency could be associated with different temperature phenotypes. Subsequently, four UK pedigrees with a cryohydrocytosis-type phenotype have not shown stomatin deficiency (Coles et al, 1999a; Haines et al, 2001).

We now present data on four new stomatin-deficient pedigrees and expand the description of Lande's original stomatin-deficient cryohydrocytosis case (Lande et al, 1982). It emerges that three of the pedigrees represent the original, very leaky, overhydrated form of HSt, while one new case, along with that of Lande et al (1982), shows a cryohydrocytosis phenotype. In all new cases, we confirmed that some stomatin was present when peripheral blood was examined by immunocytochemistry using an anti-stomatin antibody, as recently demonstrated for the original UK cases (Fricke et al, 2003a). Further, both of the stomatin-deficient cryohydrocytosis cases show a neurological syndrome of seizures, mental retardation and cataracts associated with hepatosplenomegaly, a combination that has not previously been described. No other leaky variant has shown any evidence of neurological dysfunction and this appears to be a new haemato-neurological condition.

Case reports

The family trees and blood films are shown in Figs 1 and 2, respectively, and typical haematological indices are shown in Table I. The cases in pedigrees A–C represented a simple syndrome of overhydrated stomatocytic haemolytic anaemia, while those in D and E represented cryohydrocytic cases with a neurological syndrome. For pedigrees A–C, all of whom had chronic haemolytic anaemia with a low mean cell haemoglobin concentration (MCHC) and stomatocytes on the film, the specific points were as follows.

Figure 1.

Family trees. Open symbols, not affected; closed symbols, affected.

Figure 2.

Blood films in patients A-II-3, B-II-1, C-I-2 and D-II-2.

Table I.  Haematological data.
PatientRoutine haematologyIntracellular electrolytesK influx at 5 mmol/l external [K+] at 37°C
Hb (g/dl)MCV (fl)MCHC (g/dl)Retics (%)Time stored at 0°C [Na+]i [K+]iNaK pumpNaK2Cl co-transportLeak
  1. Hb, haemoglobin; MCV, mean corpuscular volume; MCHC, mean cell haemoglobin concentration; retics, reticulocytes.

  2. Intracellular [Na+] and [K+] were measured on washed cells by flame photometry as described in Methods. For measurement of intracellular [Na+] (mmol/l cells) and [K+] (mmol/l cells), the elapsed time (h) between venesection and analysis is shown in column 6. K+ influx was measured using 86Rb as a tracer in a medium containing (mmol/l): K+, 5; Na+, 145; Cl, 150, MOPS (pH 7·4 at RT), 15; glucose, 5; and ouabain and/or bumetanide, 0·1, if required (Coles et al, 1999b). ‘NaK pump’ is equivalent to the ouabain-sensitive fraction; ‘NaK2Cl cotransport’ to the bumetanide-sensitive; and ‘leak’ to the ouabain + bumetanide-resistant.

A-II-310·8 96·230·4 8·8518 72·840·7 4·20·163·40
B-II-18·6119·925·4 9·62411132·620·90·242·30
C-I-210·112128·913·572 95252·25
Travel control    7215·993·3
D-II-212·7121·331·4 1·4fresh10631·7 7·30·450·69
E-II-111·010334 3–30fresh 3858
Normal male13–17 80–9930–35<2fresh  5·9–11·388–105 0·8–2·00–10·06–0·10

Patient A-II-3 was of Somalian descent. She is the sole affected member of the family. Patient B-II-1 was a male baby born in 1999 from haematologically healthy French parents. Patient C-1-2, was born in 1977, of Hispanic descent. A splenectomy was performed at age 17 years, but despite this procedure, she has never developed a thrombotic complication as reported for other hereditary stomatocytosis cases (Stewart et al, 1996). The patient's father was said to be affected with chronic anaemia, but was not available for testing. In adulthood, she had two affected children, C-II-1 and C-II-2 (a male and female, respectively), now aged 2 and 6 years.

Patient D-II-2 was born of French parents in 1964. He has been followed up since his first year of life for neurological problems including seizures, cerebellar ataxia, communicant tetraventricular hydrocephalus, and has been treated with a series of anti-convulsant medications. At 6 years of age, he underwent ventriculo-peritoneal shunting. He had delayed growth and was slightly overweight. He had a dysmorphic appearance with macrocephaly (+3 SD), large ears (>97 percentile), a short and wide neck and chest, and brachidactily. A congenital zonular cataract and slight mental retardation were present. Massive hepato-splenomegaly associated with chronic jaundice without haemoglobinuria was regularly noted. The liver function tests typically showed bilirubin, 114 μmol/l (normal, 3–22 μmol/l); alanine transaminase, 72 U/l (7–56 U/l); alkaline phosphatase, 72 U/l (30–125 U/l). After surgery for ventricular hydrocephalus, he suffered a severe haemolytic crisis and was transfused; he has had two further haemolytic crises. Red cell indices, when the patient was not in crisis, are shown in Table I. The blood film is shown in Fig 2. The platelet count was low (120 × 109/l). Other parameters were as follows: bilirubin, 97 μmol/l; haptoglobin, 0·05g/l; and ferritin, 349 μg/l. Since his childhood, he has regularly been found to have in vitro hyperkalaemia (5–13 mmol/l), without electrocardiographic abnormalities. The red cell enzymes and haemoglobin were normal. The Pink test (Vettore et al, 1984) was normal at 21% (normal, <30%), while there was a decreased incubated osmotic resistance of red blood cells. The somatic karyotype (RGH bands) showed no abnormalities. Both parents and his brother and sister were clinically asymptomatic and had normal blood counts with no biochemical sign of haemolysis.

The fact that the red cells of patient E-II-1 were deficient in stomatin has previously been reported (Lande et al, 1982). Of Irish origin, she was born in 1972. Like patient D-II-2, she presented with a predominantly neurological syndrome of cataracts, seizures, spastic quadriplegia and mental retardation associated with massive hepatosplenomegaly and blueberry muffin spots on the skin. A diagnosis of congenital rubella syndrome was suggested after birth, but all serological tests for this infection were negative. Short stature was noted, below the third centile for age. The neurological syndrome remains constant, and at a recent review, aged 29 years, she was awake, smiled on greeting, but did not speak fluently and was unable to follow simple commands. She showed diffuse spastic rigidity with some flexion contractures. She had some active movement in the upper limbs but none in the lower. Sensation was grossly intact. She could neither stand nor walk. The deep tendon reflexes were increased; Hoffman's sign was positive bilaterally, and the plantars were equivocal. Her extraocular movements were grossly conjugate, but were nystagmoid or rotatory when she moved her head.

Lifelong haemolysis was also present. A series of haemolytic crises occurred, usually associated with infection. Cholecystectomy was performed at the age of 9 years. The direct antiglobulin, Donath-Landsteiner, Ham's and sugar water tests were negative. The morphology and MCHC of fresh erythrocytes were regarded as normal. It was noted that if EDTA blood samples were stored for some hours on ice, haemolysis occurred, an observation that suggested a diagnosis of ‘cryohydrocytosis’. Osmotic gradient ektacytometry on fresh cells showed a slightly dehydrated normal pattern (Omin, 126 mOsm/kg, expected 114–164; O’, 348, expected 357–396). (Ektacytometry was also performed on members of pedigrees A, B and D but overnight storage of the cells made it difficult to interpret the results.) No karyotype was available.


For Na+/K+ studies, blood was usually transported in citrate–phosphate–dextrose–adenine (CPDA) anticoagulant, usually on ice, and used within 48 h. CPDA markedly ameliorates lysis in these leaky cells (Jarvis et al, 2003). Storage times and temperatures are listed in Table I. The following methods were carried out as described: osmotic gradient ektacytometry (Clark et al, 1983); sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) and Western blotting using anti-stomatin antibodies (Coles et al, 1999b); immunocytochemistry employing anti-stomatin antibody and flow cytometry (Fricke et al, 2003a); determination of the intra-erythrocytic cation concentrations; measurement of the cation leak across the membrane in the presence of ouabain and bumetanide (inhibiting the NaK-ATPase and the NaK2Cl cotransporter, respectively) (Coles et al, 1999b). For quantifiable measurement of stomatin expression on Western blots, an iodinated second antibody (donkey anti-rabbit, Amersham, IM134; Amersham, UK) was employed. Bound 125I activity was measured using a Fuji FLA-3000 phosphorimager. In semi-quantitative-enhanced chemiluminescence (ECL) studies, a goat anti-rabbit conjugated to horseradish peroxidase (Amersham NA934V) was used. ECL activity was measured in a Bio-Rad Chemi-Doc system (Bio-Rad, Hemel Hempstead, UK). The dynamic range of this imager is limited compared with that of the phosphorimager (it saturates at high signal strength) and assessments of expression by this means can only be regarded as semi-quantitative. For case E-II-1, K+ efflux was measured by serial measurements of internal [K+] at different temperatures, in the following medium (mmol/l): Na+, 140; Mg2+, 1; Cl, 127; POinline image, 1; HEPES, 40 (pH 7·4 at temperature of efflux expt); glucose, 5; ouabain and bumetanide, 0·1 each, as described (Clark, 1988). The seven exons of the stomatin gene were amplified from genomic DNA and sequenced as described (Fricke et al, 2003a).


Measurements of intracellular [Na+] and [K+] and tracer ion flux at 37°C are shown in Table I. In each case, intracellular Na+ and K+ levels were markedly abnormal. Cases A-II-3, B-II-1, D-II-2 and all showed intracellular [Na+] of >70 mmol/l and a reciprocal decrease in intracellular [K+]. Case E-II-1 showed a less marked but nevertheless major abnormality. Where measured (cases A-II-3, B-II-1, C-I-2, D-II-2), the isotopic K+ fluxes at 37°C showed an increase in both the ouabain + bumetanide-resistant, ‘leak’, component and the ouabain-sensitive, NaK pump. Although the leak components in cases A-II-3, B-II-1 and C-I-2 were very markedly increased at more than 20 times normal [as in other stomatin-deficient overhydrated hereditary stomatocytosis (OHSt) cases (Stewart et al, 1992)], the NaK pump acceleration in A-II-3 was limited to only about three times that of normal, while it was 15 times that of normal in B-II-1. The NaK pump rate was not measured in C-I-2. The leak at 37°C was not so marked in case D-II-2, but was nevertheless about twice that seen in non-stomatin deficient cryohydrocytosis (Coles et al, 1999a; Haines et al, 2001).

In all cases, Coomassie-stained SDS-PAGE (not shown) gels showed a reduction in stomatin at 32 kDa. Coomassie gels for case E-II-1 have previously been published (Lande et al, 1982). Figure 3(A) shows Western blots of normal and abnormal red cell membranes from patients A-II-3 and D-II-2 probed with a primary anti-stomatin antibody and a secondary iodinated goat anti-rabbit-IgG antibody, which enabled quantifiable assessment of the amount of stomatin present. Lanes representing known stomatin-deficient OHSt cases (Lock et al, 1961; Meadow, 1967) and a non-stomatin-deficient cryohydrocytosis (Coles et al, 1999a) are shown for comparison. The study shows that in the new cases, stomatin was not completely absent, but present at a variable level in different patients: at 23% of normal in patient A-II-3, 5·6% in B-II-1, and 3·4% and 15·7% in the Stockport (Lock et al, 1961) and Brighton (Meadow, 1967) cases respectively. In further studies, in which the semi-quantitative technique of ECL was used for detection (Fig 3B.i. and B.ii.), these findings were confirmed and extended. Figure 3B.i. shows that stomatin expression in case B-II-1 was about 6%, like Stockport and D-II-2. A-II-3 showed a stomatin expression level of about 39% of mean normal, while cases B-II-1 and D-I-2 showed about 6% and 3% of normal respectively. Two previously described cryohydrocytosis cases are also shown on this blot: Hemel (Coles et al, 1999a) and Hurstpierpoint (Haines et al, 2001). Both were strongly positive, re-affirming the difference between these forms of cryohydrocytosis. Blot B.ii. shows a lane from case C-I-2, in which stomatin expression was about 17% of normal.

Figure 3.

Western blotting of red cell membrane preparations probed with polyclonal anti-stomatin antibody. Prior to Western blotting, total protein loadings in lanes were normalized by comparison of Coomassie staining levels in standard sodium dodecyl sulphate (SDS) gels (not shown). Panel (A), quantifiable detection using 125I-goat anti-rabbit IgG. Five normal samples were compared with patients A-II-3 and D-II-2, and previously described patients from the cryohydrocytosis (CHC) Hemel (Coles et al, 1999a), OHSt Stockport (Lock et al, 1961) and Brighton (Meadow, 1967) pedigrees. Levels of expression vary between 3–23%. Panels (B.i) and (B.ii) show two blots probed with the semi-quantitative-enhanced chemiluminescence (ECL) technique. Blot (B.i) confirms that the expression in B-II-1 is very low, like D-II-2 and OHSt Stockport, and also includes lanes from CHC cases Hemel (Coles et al, 1999a) and Hurstpierpoint (Haines et al, 2001), which are clearly not stomatin-deficient. Panel (B.ii) shows data on patient C-I-2 compared with three normal samples, A-II-3 and OHSt Stockport. Expression of stomatin was quantitated at about 17% of normal by ECL.

Immunocytochemistry using a polyclonal anti-stomatin antibody in cases A-II-3, B-II-1, C-I-2 and D-II-2 was conducted (Fig 4). Cases B-II-1 and C-I-2 showed a picture similar to that previously seen in two previously published cases (Fricke et al, 2003a), in that only a minority of cells were stomatin-positive, and even then stomatin immunoreactivity (stomatin-IR) was only weak. It is noteworthy that in cases B-II-1 and C-I-2, nearly all of the stomatocytes were stomatin-negative. All of the neutrophils and platelets and about half of the lymphocytes were positive, as in the previous cases (Fricke et al, 2003a). Stomatin-IR in case A-II-3 was different, with only a minority of stomatin-deficient red cells. This was consistent with the Western blot.

Figure 4.

Immunocytochemistry using anti-stomatin antibody and avidin-biotinylated peroxidase complex (ABC) detection. The positive immunoreactive product is coloured brown. Key: n, neutrophil; spl, stomatin-positive lymphocyte; p, platelet; e+, stomatin-positive erythrocyte; e-, stomatin-negative erythrocyte; s, stomatocyte. In the normal blood film (bottom left panel), all of the red cells and all platelets are stomatin-immunoreactive. In the patients, the non-erythroid cells show the same pattern of immunoreactivity as the normal blood film, but stomatin immunoreactivity in the red cells is different. Only a few red cells are weakly positive in cases B-II-1 and C-I-2, while in case A-II-3, nearly all of the red cells are positive and only a few are negative, consistent with the Western blots in Fig 3. All of the stomatocytes are stomatin-negative.

Stomatin expression in fixed and permeabilized cells was compared with thiazole orange staining in case A-II-3 (Fig 5). Normal red cells (Fig 5A) showed high levels of stomatin expression, as expected. A-II-3 red cells showed generally lower levels, and there was a skewing in the distribution into region 2 (top right) such that the greater the stomatin expression, the greater the thiazole staining, suggesting that the young cells were richer in stomatin than the older, as previously seen in the Stockport and Brighton cases (Fricke et al, 2003a). In numerical terms, 37% of the stomatin-positive cells were also thiazole positive, whereas only 12% of the stomatin-negative cells were thiazole-positive.

Figure 5.

Flow cytometry in case A-II-3. Cells were fixed and permeabilized as described in Methods (Fricke et al, 2003a). Stomatin immunoreactivity (y-axis) was compared with intensity of thiazole orange staining (x-axis). The discriminatory boundaries were determined in control experiments such as that shown in panel C, where both the anti-stomatin antibody and thiazole orange were omitted. In the normal sample, (panel A), all red cells were stomatin positive (regions 1 and 2, above horizontal line) and only a minority were thiazole positive (to right of vertical boundary, regions 2 and 4). There is less stomatin immunoreactivity in the patients’ red cells (panel B) and a skewing of the distribution towards the upper right box (region 2), indicating that the greatest stomatin immunoreactivity is associated with the greatest thiazole staining.

Temperature studies of the K+ leak as a function of temperature (Fig 6) were striking. Panel A shows measurements of K+ influx on patients A-II-3, B-II-1, C-I-2 and D-II-2, while panel B shows K+ efflux data on patient E-II-1. At 37°C, the flux in patients A-II-3, B-II-1 and C-I-2 exceeded that in patient D-II-2, consistent with the data in Table I. At lower temperatures, patients A-II-3, B-II-1 and C-I-2 showed a steady fall in flux rate with a flattening at 5°C in A-II-3 but not in C-I-2. By contrast, the curve in patient D-II-2 showed a U-shaped profile such that the flux at 0°C vastly exceeded that at 37°C, far in excess (at this temperature) of any other cases that we have seen. Case E-II-1, in which K+ efflux was measured, showed a similar U-shaped profile. These profiles are consistent with a ‘cryohydrocytosis’ phenotype.

Figure 6.

Temperature dependence of the K+ leak. Panel (A), ouabain + bumetanide-resistant K+ influx was measured in patients A-II-3, B-II-1, C-I-2, D-II-2 and a control, using conditions described in the legend for Table I. Cases A-II-3, B-II-1 and C-I-2 show a monotonic decrease with temperature; case D-II-2 shows a flat temperature dependence between 37 and 27°C, after which the flux increases with decreasing temperature. Panel (B), ouabain + bumetanide K+ efflux in case E-II-1 and control, showing a similar curve to that of D-II-2 in the ouabain + bumetanide K+ efflux, with a minimum at about 20°C.

Sequencing of the stomatin gene was undertaken in D-II-2 and E-II-1. The seven exons were amplified and sequenced. At position +2779 in the cDNA (+2125 in exon  7), patient E-II-1 was heterozygous for the sequence variation 2779A > G (…aacagtA/Gctttt…) while D-II-2 was had the genotype A/A at this position, as in the published sequence (Unfried et al, 1995). This is in the 3′-untranslated region and seems most unlikely to be significant.

We hypothesized that the deficiency of stomatin might be a direct result of the very abnormal Na+ and K+ content of these cells. We took non-stomatin-deficient cryohydrocytosis red cells (Haines et al, 2001), incubated them at 0°C for 2 h, during which time these cells will equilibrate with the surrounding plasma (Jarvis et al, 2003), and incubated them with ouabain at 37°C to allow any process of protein trafficking to occur, but no change in stomatin content was observed by immunocytochemistry (not shown).


These cases confirm that stomatin is deficient only in very Na+–K+ leaky examples of the hereditary stomatocytoses. It is not the abnormality in intracellular Na+ or K+ as such that is important, as equally high Na+ levels can be seen in other forms without stomatin deficiency (Jarvis et al, 2001). However, all of the stomatin-deficient cases show higher leak K+ flux than the non-stomatin-deficient cases. In each case where immunocytochemistry was performed (A-II-3, B-II-1, C-I-2, D-II-1), the blood film showed variation in stomatin expression among the red cells. Flow cytometric studies in patient A-II-3 confirmed that greater stomatin expression was associated with younger red cells, as previously found in other cases (Fricke et al, 2003a). As in other cases, stomatin expression was normal in leucocytes and platelets, as seen in normal blood.

In contrast to the cases that showed monotonic temperature dependence, cases D-II-2 and E-II-1 showed the U-shaped temperature profile typical of the cryohydrocytosis-type variant. The presence of stomatin deficiency distinguishes these cells from the haemolytic (Coles et al, 1999a; Haines et al, 2001) and non-haemolytic (Gore et al, 2002) cases that show this U-shaped profile. Cases D-II-2 and E-II-1 are, to our knowledge, the only two patients to show the haematological picture of stomatin-deficient cryohydrocytosis. Both of these cases, and only these cases, showed the associated systemic syndrome of mental retardation, seizures, cataracts, short stature and massive hepatosplenomegaly (and in case D-II-2, dysmorphic features) (see Table II). This may define a new haemato-neurological syndrome. Further cases are required to confirm this suggestion.

Table II.  Comparison of neurological abnormalities in cases D-II-2 and E-II-1.
Spastic quadriplegia+
Cerebellar ataxia+
Mental retardation++
Developmental anomalies: dysmorphia, brachydactyly, macrocephaly+
Growth retardation++
Massive hepatosplenomegaly++

Sequencing of the stomatin gene revealed no significant abnormality in either D-II-2 and E-II-1, as previously found in overhydrated cases, confirming that the gene was normal in both forms of these leaky red cell diseases (Fricke et al, 2003a). This was not a surprise, as stomatin is a very widely distributed protein (Fricke et al, 2003b; Stewart & Fricke, 2003). It was present even on the platelets and neutrophils of peripheral blood films of all of these patients. This normal sequence is in stark contrast to the splicing abnormality in the stomatin gene that was found in a case of recessive multi-system disease, quite different from any form of hereditary stomatocytosis (Argent et al, 2004).

The function of stomatin remains unknown. It is associated with sphingomyelin + cholesterol-rich ‘rafts’ in eukaryotic cells (Snyers et al, 1999). Stomatin may be associated with the control of surface expression of membrane proteins (Tavernarakis et al, 1999). The cells in both the recessive multi-system case (Argent et al, 2004) and the knock-out mouse (Zhu et al, 1999) were not leaky to Na+ and K+, and the lack of stomatin seems most unlikely to be the cause of the leak.

All of the cases described here, except those in pedigree C, are isolated. Given past experience of these leaky cell diseases, in which cases that were originally isolated have subsequently given rise to affected children (Townes & Miller, 1980; Morleéet al, 1989; Haines et al, 2001; Jarvis et al, 2001), it seems likely that they all represent de novo mutations of dominant disorders. Cases D-II-2 and E-II-1, which show neurological dysfunction, are different, and these cases could be recessive, but there were no other affected siblings in either family, nor was consanguinity present.


We thank the patients for their co-operation. We thank the Wellcome and Sir Jules Thorn Trusts and INSERM (Unité 473) and INSERM/AFM project no. 4MRo9F for support and the patients for their co-operation. We are grateful to the National Institutes of Health (grants nos HL20985, HL27059, DK 32094) for funding, and to Dr Jonathan Artz for allowing us to report his patient. We thank Prof. Geoff Laurent for access to the phosphorimager, and Dr David Rees for permission to report his patient. We thank Luzie Augustinowski, Katja Rumpf and Stephanie Pastors for excellent technical assistance.