illustrates the experimental scheme for the 2D-DIGE approach used to identify proteins differentially expressed in RBC cytosols. Routine evaluation of samples began with non-proteomic approaches including testing for unstable hemoglobin and evaluation of enzymatic activity for the most common causes of HNSHA (G6PD, PK, and Hexokinase deficiency), if not previously reported in the patient’s record. RBC preparation included passage of cells over microcrystalline cellulose to remove leukocytes followed by hypotonic lysis of RBC, removal of membranes and hemoglobin depletion from the resulting hemolysate. The resulting cytosolic protein fraction was then prepared for DIGE. Gel images were evaluated, and differentially expressed ‘spots’ were excised and processed for protein identification by LC/MS.
Schematic illustrating the experimental work flow from diagnosis of an anemia patient as HNSHA of unknown etiology to identifications of proteins potentially involved in the pathology of the disease.
summarizes relevant laboratory evaluation of the 4 HNSHA patients included in this study: HA09, HA19, HA21 and HA24 (HA for hemolytic anemia). HA09 is a 10-year old male who presented with severe chronic hemolytic anemia at birth, and who remains transfusion dependent. Bone marrow aspirate revealed erythroid hyperplasia with prominent basophilic stippling. Prior workup was negative for hemoglobinopathy, common enzyme defects or RBC membrane abnormalities. Lab values concurrent with sampling for this study included hematocrit (27%), hemoglobin 8.9 g/dl, RBC count 3.09x1012/l, reticulocyte count of 3.5%, bilirubin (11.0 mg/dl) and ferritin (639 ng/ml) levels. Because of the patient’s transfusion dependence, the sample was obtained at the nadir just prior to transfusion (in this case 6 weeks following the most recent transfusion) to maximize the fraction of patient derived RBC for analysis. Glucose-6-phosphate dehydrogenase (G6PD), pyruvate kinase (PK), hexokinase (HK), gluthathione peroxidase (GPx), glutamate-oxaloacetate transaminase (GOT) and glucose phosphate isomerase (GPI) activity as well as a screening test for pyrimidine 5' nucleotidase (P5'N-1) were evaluated to rule out known enzyme defects. Reduced Glutathione (GSH) level was normal. Lacking a clear diagnosis, this sample was selected for proteomic analysis comparing HA09 (patient), his mother, one unrelated control, and a common pooled control sample. The same, pooled control sample (designated ‘standard’) was used in each subsequent HA DIGE experiment as a common control for comparison between experiments.
Sample characterization (Blood Parameters and Enzyme Assays).
HA19 is a 2-year old female who presented with chronic, hemolytic anemia accompanied by splenomegaly, leucopenia and thrombocytopenia. Initial work up was negative for common enzyme deficiencies (G6PD and PK), hemoglobinopathy or membrane defects, and there was no family history of chronic anemia. Lab values concurrent with sampling for this study include hematocrit of 23%, hemoglobin of 7.3, RBC count of 3.1x1012
/l and reticulocytes of 11.6%. Evaluation in our lab revealed no abnormalities in G6PD, HK, PK, GOT, GPI, GPx, triose phosphate isomerase (TPI) or phosphoglycerate kinase (PGK), with GSH in the normal range (Table S1
). Because no diagnosis for the etiology of anemia was made, this sample was selected for proteomic analysis comparing HA19 (patient), both parents and one unrelated control sample.
HA21 is a 15-year-old female with mild chronic anemia of unknown cause and a history of a previous acute hemolytic event. Prior evaluation showed no evidence of common enzyme deficiency (G6PD, PK, HK, GPI, TPI and GPx reported as normal), hemoglobinopathy or membrane defect, with normal osmotic fragility. Lab values concurrent with sampling for this study include hematocrit of 34%, hemoglobin of 12.7 g/dl, RBC count of 3.57x1012/l and reticulocytes of 18%. Enzyme analyses in our laboratory were negative for deficiencies, but showed elevated GPx in both the patient and her mother. In the absence of an identified defect, proteomic analysis was performed comparing HA21 (patient), her mother and two unrelated control samples.
HA24 is a 6-month old female with severe, chronic anemia requiring transfusion approximately every 5 weeks. As with HA09, a sample for analysis was obtained at the nadir, just prior to transfusion for analysis. At the time of sampling, hematocrit was 22.3%, hemoglobin 7.4 g/dl, RBC count 2.49x1012
/l and reticulocytes 6.21%. Prior workup for common enzyme deficiencies, hemoglobinopathy and membrane defects were negative. HK activity and GSH levels, were in the normal range. Despite a report of normal PK activity as part of the patient’s original evaluation, enzyme analysis in our laboratory was suggestive of PK deficiency, particularly considering the patient’s reticulocyte count and history of transfusion, with patient, mother and father all showing decreased activity (Table S1
). Despite this ambiguity, proteomic analysis was begun in this case comparing HA24 (patient), mother, father, sister and one unrelated control. Subsequent sequence analysis of the PKLR gene revealed that HA24 is homozygous for the R479H mutation 
; the parents are heterozygous carriers, while the sister is unaffected. A subsequent western blot analysis utilizing a PK specific antibody demonstrated low, but detectable PK expression in the patient, and intermediate expression in both parents compared to unaffected controls (Figure S1
). Proteomic analysis of these samples was completed, providing a comparison of differential protein expression in the context of a defined molecular cause of HNSHA.
provides an overview of DIGE experiments, including a summary of the number of spots detected on gels from each sample set after spot filtering, and the number of filtered spots differentially expressed by analysis of variance (ANOVA). We excluded spots from further evaluation when they were not matched correctly between the gels, when spot intensity appeared too low to allow identification by mass spec, and when a spot contours drawn by software corresponded to a smear rather than a spot with defined boundaries. A total of 243 spots were picked and a total of 213 proteins were identified (several proteins being repeatedly identified) when pooling data from all four experiments (Table S2
). 21 of these proteins have not been previously identified in RBC proteome databases 
(listed in Table S3
). Despite good resolution on 2-D gels, proteins with similar molecular weight and isoelectric point (pI) comigrate, resulting in identification of more than one protein in some spots analyzed.
Summary of spots detected by Same Spots Software.
As an overall data quality metric, Principal Component Analysis (PCA, ) of all spots included () shows that replicate samples (run with dye reversal) are highly similar in all cases, and that protein expression patterns in HA patient samples are easily distinguished from control samples run in the same experiment. Replicate samples remained tightly grouped when PCA was performed on all detected spots or on all differentially expressed spots (ANOVA). Plotting of the spots showed that in all cases the patient was an “outlier”, separated from more closely grouped family members and controls.
Principal Component Analysis was calculated by Same Spots software.
Assessment of expression profiles of HA09 showed down-regulation of proteins to be more common than up regulation, with more than twice as many spots showing lower expression in the patient (, A). For HA19 and HA21 we found a comparable number of differentially expressed protein spots with elevated or reduced expression in the patient versus controls (, B-C). 25 spots showed reduced expression in HA24 compared to control samples in this set (, D, lower panel). We found 9 spots with increased expression in the patient accompanied by intermediate expression levels in the mother (, D, upper panel). In each DIGE comparison, we observed several protein spots with expression intensities for one or both parents intermediate between patient and control samples (as in , D above), a pattern that might be expected if parents are heterozygous for a polymorphism affecting expression of the respective protein. Spots with this pattern are considered in more detail below (see ).
Expression profiles of spots showing up- or down-regulation in patient samples.
Proteins with parental expression intermediate between patient and unrelated control samples.
By superimposing images of Coomassie stained gels from all four experiments including the positions of the picked spots, it was possible to identify differentially expressed spots that were picked in more than one experiment. Across all gels, protein identification was overlapping with the same major protein species present in all cases, and additional peptides from minor species being variably present. As an initial method of protein ID verification the location of the excised spot was compared with the molecular weight (kD) and presumed pI of the identified protein(s). When the discrepancy was > 20% of theoretical molecular mass or > 1 pI unit, it was often found that peptides were derived from the previously picked spot (representing contamination at the level of the spot picking head) and could thus be excluded from further analysis. In some instances we found multiple peptides mapping to a protein larger than the spot location could only be aligned to one part of that protein, suggesting that the protein fragmented prior to gel separation, either as an artifact of processing or a physiologic cleavage prior to sample processing. A representative Coomassie stained 2-D gel is shown in Figure S2
presents an overview of this data highlighting proteins that were most highly or most frequently differentially expressed. These include exportin 7 (XPO7), fumarate hydratase (FH), purine nucleoside phosphorylase (PNP), chaperones, cytoskeletal and ribosomal proteins, proteasome subunits and additional proteins involved in protein degradation. Proteins showing lower expression in individual patients are listed in Supplementary Table S4
Schematic depiction summarizing proteins and groups of proteins found to be differentially expressed in patients HA09, HA19, HA21 and HA24.
XPO7 is a member of the importin-beta superfamily of nuclear transport receptors thought to cycle between the cytoplasm and the nucleus 
. XPO7 was identified in multiple co-migrating spots in three experiments (, A and Figure S3
) presumably representing isoforms with distinct post-translational modifications. Patients HA19, HA21 and HA24, all showed consistently increased levels of XPO7 relative to control samples.
Exportin 7, Fumarate Hydratase and Purine Nucleoside Phosphorylase protein expression.
The tricarboxylic acid cycle (TCA) enzyme fumarate hydratase, which normally localizes to mitochondria, was unexpectedly found to be present in RBC cytosol. A possible explanation for the presence of this enzyme would be as a component of residual mitochondria in circulating reticulocytes. In patients HA19 and HA21, this explanation appears incorrect, as fumarate hydratase protein was reduced more than 2 fold in these patients compared to controls (, B and Figure S4
) despite these patients having the highest reticulocyte counts among samples examined.
Purine nucleoside phosphorylase (PNP) was identified in several spots in all experiments. PNP deficiency is a rare autosomal recessive disorder characterized by autoimmunity that may include hemolytic anemia 
. Patients HA19 and HA21 both show lower PNP abundance in several spots when compared to control samples. Interestingly, the extent of decreased expression varies among the spots: in HA19 from 3.1 fold to 7.7 fold and in HA21 from 2.23, to 9.1 fold, respectively (Table S4, B
), while the overall expression pattern remains similar (, C, D).
Abundance of specific cytoskeletal proteins differed in all four patients (Supplementary Table S5
–and ) compared to controls. Actin was increased in patients HA09 and HA24, while tubulin was increased in HA21, and a subunit of the dynactin/dynein complex (ACTR1A) 
showed higher levels of expression HA09. Expression of an actin filament capping protein (CAPZB), a regulator of actin filament growth 
was found to be lower in HA19. A small GTPase, RhoA, a regulator of actin dynamics 
showed lower expression in HA19 and HA21. We also found diminished expression levels in these patients for two GDP dissociation inhibitors that influence the GDP-GTP exchange of rab GTPases, known to be involved in vesicle trafficking 
Two proteins thought to be involved in ribosome assembly and/or stability were differentially expressed in all four patients. Proliferation-associated protein 2G4 (P2G4) is part of a pre-ribosomal ribonucloeprotein complex and has been implicated in growth regulation in human fibroblasts 
and cancer cells 
. In HA09 PA2G4 expression is reduced 1.5 to 1.61 fold (Table S4, B
). Ribosomal protein SA (RPSA), also called Laminin-receptor 1, was more highly expressed in HA19, HA21 and HA24 (Table S4, A
). In addition to being a receptor for laminin, RPSA is required for assembly and stability of the 40S ribosomal subunit.
Chaperones were also differentially expressed in all patients (Table S6
and ). Conspicuously, in HA09 multiple T-complex subunits (TCP) showed reduced expression, while several heat shock proteins (HSP) were present in higher amounts when compared with control samples (). In HA19 and HA21 several TCP subunits were increased; HA19 also had decreased expression of a single heat shock related protein (HSPA8, , B). Several chaperone proteins showed reduced expression in HA19 and HA24, although only two spots in HA21 and one spot in HA24 met the threshold for a change above 1.5-fold.
Figure 6 Expression profiles of spots predominantly containing Chaperone proteins are shown (see also Supplementary Table S6).
A prominent common pattern observed in all patient samples was reduced abundance of multiple proteasome subunits as well as reduced expression of additional proteins involved in protein degradation (, Table S7
), indicating differences in protein turnover between HA patient RBC and control samples. There was a single exception to this pattern among the proteasome subunits (increased expression of PSMC5 in patient HA21, , C green line). In HA19 and HA21 two additional proteins associated with the proteasome, thioredoxin-like 1 
and PITHD1 (C-terminal proteasome-interacting domain of thioredoxin-like) domain-containing protein 1, showed higher expression levels compared to controls.
Figure 7 Expression profiles of proteins involved in protein degradation (see also Supplementary Table S7).
In order to focus upon protein expression changes that could play a role in etiology of HNSHA, we searched for spot patterns where the volume measured in the patient sample was most distinct from the control samples, and one or both parents showed intermediate expression. We reasoned that such a pattern might be indicative of recessive inheritance of a lesion for which both parents were heterozygous ( and )–particularly in those cases where expression was decreased in the HA patient sample. Such patterns of possible recessive inheritance were observed in patients HA09, HA21, and HA24.
Examples of protein expression patterns where one or both parents are intermediate between the levels measured in the patient sample and those measured in the controls (see also ).
In HA09 (, A) two spots containing G6PD (#82 and #160) and two spots identified as PA2G4 (#232 and #307), were identified with the lowest expression measured in the patient sample, with mother or mother and father showing intermediate expression relative to unrelated controls. Expression profiles for HA19 are shown in , B. 14-3-3 epsilon (#94) showed reduced expression in both the patient and mother, and HSPA8 (#89) showed low expression in the patient with intermediate expression in both parents. PITHD1 (#231) showed increased expression in the patient, and intermediate expression in the mother. In HA21, proteasome subunit PSMB4 (#61) and superoxide dismutase 1 (#204) showed low expression in the patient and intermediate expression in the mother (, C, left panel). Four spots with higher expression in the patient and intermediate expression in the mother compared to controls (, C, right panel) were identified: EIF2S3 (eukaryotic translation initiation factor, #28), ACTR1A (alpha-centractin, #67), and two spots both identified as VCP (valosin containing protein, #162 and #200).
In HA24, CCT8 (T-complex chaperone subunit, #97) was identified as a protein with low expression in both the patient and mother (, D, left panel). 8 spots with increased expression in the patient and intermediate expression in one parent were identified: RPSA (ribosomal protein SA, #13), XPO7 (exportin 7, #14 and #55), EEF2 (elongation factor 2, #4) BPGM (bisphosphoglycerate mutase, #36 and #43), ADSL (adenylosuccinate lyase, #132), PRDX2 (peroxiredoxin 2, #159) and PGAM1 (phosphoglycerate mutase 1, #167).