Increases in beta-lipoproteins in hyperlipidemic and dyslipidemic dogs are associated with increased erythrocyte osmotic fragility


  • Erica Behling-Kelly,

    Corresponding author
    1. Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA
    • Correspondence

      Erica Behling-Kelly, Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, S1-062 Schurman Hall, Tower Road, Ithaca, NY 14853, USA


    Search for more papers by this author
  • Rebekah Collins-Cronkright

    1. Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell University, Ithaca, NY, USA
    Search for more papers by this author



The capacity of RBCs to traverse endothelium and deliver oxygen to tissues is dependent on a mechanically stable yet flexible plasma membrane. The lipid composition of the RBC membrane is crucial in maintaining its structure and fluidity. Lacking a nucleus and having minimal synthetic capacity, RBCs are exquisitely sensitive to changes in plasma lipids.


The objective of this study was to investigate the effects of plasma lipids on RBC fragility in hyperlipidemic and dyslipidemic dogs.


Osmotic fragility of RBCs, plasma lipoprotein fractions, and cholesterol and phospholipid content of RBC membranes were measured in hyperlipidemic, dyslipidemic, and healthy control dogs. Osmotic fragility of normal canine RBCs incubated in phosphate-buffered saline and in both intact and lipid-depleted plasma from diabetic dogs was also measured.


RBCs from hyperlipidemic and dyslipidemic dogs with diabetes mellitus and dogs treated with glucocorticoids were significantly more fragile than RBCs from healthy control dogs. RBCs from hyperlipidemic dogs with cholestatic disease tended to be more stable relative to RBCs from controls. RBC osmotic fragility was positively correlated with beta-lipoprotein levels, but was only weakly correlated with serum cholesterol concentration. Incubation in plasma from hyperlipidemic diabetic dogs rendered RBCs from healthy dogs osmotically fragile, whereas lipid-depleted plasma from the same diabetic dogs had no effect.


RBCs from hyperlipidemic and dyslipidemic dogs are osmotically fragile, and fragility is highly correlated with increases in beta-lipoproteins. Future studies are planned to address the consequences of lipid-induced fragility and subclinical hemolysis on endothelial cells, platelets, and coagulation.


In dogs, hyperlipidemia is characterized by an increase in lipoproteins, including chylomicrons, very low-density lipoproteins (VLDL), low-density lipoproteins (LDL), and high-density lipoproteins (HDL). Intermediate-density lipoproteins (IDL) are short-lived and are not found in plasma of healthy dogs.[1] Dyslipidemia occurs when total lipoproteins are within the reference interval (RI) and the relative proportion of classes of lipoproteins is altered. Lipoprotein classes are differentially susceptible to oxidation, bind different membrane receptors, and have divergent kinetic profiles in their relative ability to transfer cholesterol and phospholipids between cells, including RBCs.[2-4] The ratio of phospholipid to cholesterol in cell membranes has a profound impact on cell rigidity.[5, 6] Importantly, RBCs lack the ability to synthesize cholesterol and have minimal capacity for active removal of cholesterol from their membrane.[7] This limitation makes RBCs particularly susceptible to membrane modification by plasma lipoproteins.

The detrimental effects of hyperlipidemia on RBC morphology and life span have been documented in a number of species, including rats, rabbits, and dogs with profound changes in RBCs resulting from simply feeding these animals a high-fat diet. Rats fed a cholesterol-enriched diet develop hemolytic anemia attributed, in part, to RBC membrane loss of phospholipids and thus increased rigidity of the RBC membrane.[8] RBCs from cholesterol-fed rats had visible holes detectable by electron microscopy.[9] Laboratory Beagles fed a high-cholesterol diet developed anemia characterized by acanthocyte, or spur cell, formation.[10] Likewise, feeding a high-fat diet to New Zealand White laboratory rabbits resulted in a hyperlipidemic state, anemia, leukocytosis, thrombocytosis, decreased oxyhemoglobin percentage, increased methemoglobin percentage, and an increase in RBC osmotic fragility, proposed to result from lipid-triggered production of reactive oxygen species and free-radical oxidation of hemoglobin.[11] Although capable of precipitating hemolytic anemia in dogs, hyperlipidemia often is not considered a cause of clinical anemia. However, it is possible that hyperlipidemia frequently may contribute to membrane alterations, subsequent RBC fragility, and subclinical compensated hemolysis in dogs.

We hypothesized that RBCs from hyperlipidemic and dyslipidemic dogs have increased osmotic fragility in the absence of overt evidence of hemolysis or RBC morphologic abnormalities. Our objective was to study the relationship between RBC fragility and plasma lipid composition in dogs by measuring RBC osmotic fragility, characterizing plasma lipoprotein fractions, and measuring cholesterol and phospholipid content in RBCs from hyperlipidemic, dyslipidemic, and control dogs, and by investigating the effects of hyperlipidemic plasma on normal RBCs.

Materials and Methods

Study design and study population

This prospective study was conducted from June 2012 to November 2013 and involved two separate rounds of selection of canine patients presented to Cornell University Hospital for Animals. The study was conducted following all institution guidelines and with an approved institutional animal care and use protocol. In the first round, inclusion criteria were the presence of hypercholesterolemia, regardless of underlying etiology, based on a measured serum cholesterol concentration > the upper limit of the RI (332 mg/dL), the availability of results of a CBC and routine serum biochemical profile (Roche Hitachi 917, Indianapolis, IN, USA), and the availability of RBCs and serum. Only RBC osmotic fragility and lipoprotein fractions were measured in these dogs. Exclusion criteria were absence of reticulocytosis, as reticulocytes can alter osmotic fragility due to their increased surface area and retained cation pumps[12, 13], and intravascular hemolysis. Reticulocytosis was identified by measuring reticulocytes using an automated hematology analyzer (ADVIA 2120, Siemens Healthcare Diagnostics, Norwood, MA, USA) and evaluating peripheral blood smears for polychromasia. Absence of hemolysis was verified by a hemolytic index < 20 as measured by the ADVIA 2120. In the second round of patient selection, dogs with specific diseases that frequently result in hyperlipidemia or dyslipidemia were identified and placed into disease groups based on the following criteria: (1) dogs with diabetes mellitus that were at least 30 days beyond their initial diagnosis, were hyperglycemic, and lacked evidence of any additional disease process; (2) dogs with hyperadrenocorticism confirmed by ACTH stimulation or low-dose dexamethasone suppression tests; (3) dogs being treated with an immunosuppressive dose of glucocorticoids (at least 1 mg/kg prednisone); and (4) dogs with cholestatic liver disease based on hepatobiliary imaging findings and cytologic evidence. Dogs with immune-mediated hematologic diseases were excluded. Full data sets including CBCs, serum biochemical profiles, RBC osmotic fragility tests, and lipoprotein electrophoresis were obtained for all these dogs. Reticulocytosis and hemolysis were again exclusion criteria. Control animals in both rounds of selection were designated as healthy based on results of a CBC and serum biochemical profile, and no evidence of systemic illness as determined by physical examination and additional laboratory testing, including variables relevant for thyroid function. All animals were fasted overnight based on owner reports. Serum samples and whole blood anticoagulated in EDTA were obtained from the Cornell Clinical Pathology Department.

Serum lipoprotein agarose electrophoresis

Serum samples were prepared for electrophoresis by addition of 5 μL of loading buffer (60% sucrose, 0.1% bromophenol blue in double-distilled [dd] H2o) to 15 μL serum. Diluted serum samples were loaded by volume (15 μL) into a 1% agarose gel (Agarose Unlimited, Gainsville, FL, USA) prepared in 60 mM sodium-barbital buffer (Sigma-Aldrich, St. Louis, MO, USA). Lipoproteins were separated by horizontal electrophoresis at 80V for 55 minutes in 60 mM sodium barbital buffer running buffer (BioRad gel systems, Hercules, CA, USA). Gels were stained overnight at room temperature (20–22°C) in 0.18% Sudan black B (Sigma-Aldrich) in 70% ethanol. Gels were destained in a 15% acetic acid/20% acetone solution in dH2O for 2–3 hours, until the background was light gray to clear and the bands were clearly visible. Gels were scanned (Epson Perfection V500, Long Beach, CA, USA) as negative images, and lipoprotein bands were quantified by converting the pixel intensity of the scanned lane to a linear peak plot followed by calculation of the area under the curve (AUC) for each defined peak using NIH Image J software (free imaging analysis software, The HDL:non-HDL ratio was calculated using the following equation: HDL AUC/(LDL AUC + VLDL AUC).

RBC osmotic fragility

The method utilized was a modification of a previously published method.[14] Whole blood anticoagulated with EDTA (15 μL) was added to 1 mL of saline solutions of 0.1, 0.2, 0.3, 0.36, 0.38, 0.4, 0.42, 0.44, 0.46, 0.48, 0.5, 0.52, 0.54, 0.56, 0.58, 0.6, 0.7, and 0.9% NaCl. Following 45 min of incubation at room temperature (20–22°C), samples were centrifuged for 10 min at 3500g (Eppendorf 5430R, Eppendorf, Hauppauge, NY, USA). This centrifugation step caused no RBC lysis in isotonic saline in optimization experiments or in analysis of RBCs from control dogs. Optical densities (at 540 nm) of supernatants (200 μL per sample) were measured by spectrophotometry (Spectra Max M3, Molecular Devices, Sunnyvale, CA, USA) and converted to percent hemolysis using the 0.9% saline solution as a blank and ddH2O as the full-lysis control. Percent hemolysis was plotted against the saline concentration to determine the median corpuscular fragility (MCF), which is equal to the saline concentration causing 50% hemolysis. Each osmotic fragility curve was visually examined for irregularities, such as obvious pipetting errors or nonsigmoidal shape. When technical errors were identified, these data were eliminated from the data set. As samples from hyperlipidemic or diseased dogs and healthy control dogs could not always be collected at the same time, we performed initial experiments to determine the effects of RBC storage on the fragility assay.

Preparation of RBC membrane ghosts

One mL of EDTA anticoagulated whole blood was placed in a 1.7-ml centrifuge tube (Eppendorf) and centrifuged for 10 min at 1500g. The plasma and buffy coat were removed by aspiration; 500 μL of the remaining RBC suspension was placed in a new tube containing 1 mL ice-cold 0.45% saline. The suspension was flash-frozen by placing the tube in a −80°C freezer for 30 min and then rapidly thawed by immersion in a 37°C water bath. The lysed RBC membranes were washed extensively with phosphate-buffered saline (137 mM NaCl, 2.7 mM KCL, 10 mM NaHPO4, 1.8 mM KH2PO4, pH 7.4) with repeated centrifugations for 10 min at 3000g until the pellet was white. The membrane pellet was then resuspended in 250 μL solubilization buffer (20 mM Tris-HCl, pH 7.5,150 mM NaCl, 1 mM Na2EDTA, 1 mM EDTA, 1% Triton) and frozen at −20°C for 2–9 months for later analysis.

Measurement of RBC membrane cholesterol, phospholipid, and protein content

Cholesterol content was measured in 50 μL of the thawed RBC membrane suspensions using the Amplex Red Cholesterol Assay (Invitrogen, catalog number A12216, Carlsbad, CA, USA) according to the manufacturer's instructions. Assays were performed in duplicate. In this assay, cholesterol esterase hydrolyzes all cholesterol that is then oxidized by cholesterol oxidase to yield H2O2 and the corresponding ketone product. The H2O2 is detected via a 1:1 stoichiometric reaction with 10-acetyl-3,7-dihydroxyphenoxazine in the presence of horseradish peroxidase to produce resorufin, which is measured fluorimetrically (emission and excitation at 571 nm and 585 nm, respectively; Spectra Max M3; Molecular Devices, Sunnyvale, CA, USA). Fluorescent units were converted to cholesterol concentration by extrapolation from a standard curve.

Phospholipids are the sole source of phosphate in RBC membranes, and measuring phosphate is an accepted means of quantifying phospholipids, which are 4% phosphate by weight.[15] RBC phosphate was measured using the Phosphate Colorimetric Assay Kit (catalog number MAK 030, Sigma-Aldrich) according to the manufacturer's instructions. In this assay, phosphate reacts with a proprietary chromogenic complex producing a colorimetric product that is quantified at 650 nm. The optical density is converted to phosphate concentration by extrapolation from a standard curve. Fifty microliters of the RBC membrane suspensions were assayed in duplicate.

Protein in 10 μL of the RBC membrane suspension was quantified in duplicate using a standard bicinchoninic acid assay (catalog number 23227, Pierce, Rockford, IL, USA). Cholesterol and phosphate concentrations were converted to total content per 50 μL of suspension and normalized to mg/mL protein in the suspension. Normalized values were used to compare measurements across samples and used to calculate the serum cholesterol:phosphate ratio.

Serum delipidation

Lipid Removal Agent (LRA, Sigma/Supelco, St. Louis, MO, USA), a calcium silica hydrate compound designed for clearing pharmaceutical agents of lipid-based solvents, was used to adsorb lipoproteins from plasma. Solid LRA was packed to the 250 μL mark in a 1.5-mL centrifuge tube (Eppendorf) to which 750 μL plasma was added. The mixture was rotated at room temperature (20–22°C) for 12 hours. The slurry was then centrifuged at 6500g and the lipid-depleted plasma in the supernatant was used in subsequent experiments. Removal of lipoproteins and retention of remaining proteins were verified by lipoprotein electrophoresis as described earlier and automated serum protein electrophoresis (SPIFE 3000, Helena Laboratories, Beaumont, TX, USA). Samples that clotted during processing or samples not cleared of lipid as demonstrated by electrophoresis were not used.

Treatment of whole blood with delipidated plasma

EDTA-anticoagulated whole blood from a healthy control dog was divided into 4 separate 400-μL aliquots and centrifuged at 1500g for 10 minutes; 150 μL of autologous plasma was removed from each tube and replaced by an equivalent volume of (1) autologous plasma, (2) PBS, (3) plasma from a diabetic dog, or (4) delipidated plasma from the same diabetic dog. Tubes were incubated overnight at room temperature (20–22°C) with rotation, and an osmotic fragility assay was performed the following morning. Samples from diabetic dogs were selected for these experiments because of the frequency of large dogs with diabetes in our patient population, permitting collection of larger blood volumes.

Statistical analysis

Mean MCF values for hyperlipidemic and control dogs were compared using a Wilcoxon rank-sum test. Correlation analyses were performed using a Spearman regression for nonparametric data. Comparisons between defined patient groups and the control group were performed using Dunnett's multiple comparison test. The treatment groups and control group in the diabetic plasma-treatment experiments were compared using a one-way ANOVA followed by a Tukey pair-wise comparison. For all analyses, P < .05 was considered significant. All analyses were performed using Prism GraphPad software, v. 5 (GraphPad Software, Inc, La Jolla, CA, USA).


Study population

In the first round of experiments, osmotic fragility tests were performed on 20 control dogs and 21 hyperlipidemic dogs. The hypercholesteremic group included dogs with idiopathic epilepsy (2), mast cell neoplasia (1), protein-losing nephropathy (1), hypothyroidism (3), glaucoma (1), Shetland Sheepdogs with familial hypercholesterolemia (2), pancreatitits (2), chronic vomiting (1), glaucoma (1), hyperadrenocorticism (2), diabetes mellitus (3), and lymphoma (2). There were 30 dogs in the second group, including dogs with diabetes mellitus (9), hyperadrenocorticism (5), and cholestatic liver disease (2 with a mucocoele and 2 with cytologic evidence of marked cholestasis and imaging evidence of an extrahepatic mass), as well as dogs undergoing treatment with corticosteroids (6) and an additional 7 healthy dogs serving as controls. In the control group, all CBC and serum biochemical results were within RIs established at the Cornell University Hospital for Animals, with the exception of 3 dogs with mild hyperproteinemia attributed to dehydration. In the diabetic group, 2 dogs had mildly decreased Na and Cl concentrations, all 9 had mildly increased in ALP activities, all but one had moderately increased glucose concentrations (mean 363.46 mg/dL; RI 63–118 mg/dL), 4 had mild increases in GGT activity, and cholesterol concentration was > the RI in 7 dogs (mean 577.15 mg/dL; RI 138–332 mg/dL). Hyperglycemia was not present in the dogs being treated with glucocorticoids, or in dogs with hyperadrenocorticism; otherwise, these dogs had biochemical changes similar to the diabetic dogs, with 3 of the 6 dogs treated with glucocorticoids having hypercholesterolemia (mean 313.75 mg/dL; RI 138–332 mg/dL). One diabetic dog and one dog with hyperadrenocorticism had a mild decrease in MCHC, and one diabetic dog was leukopenic (WBC 2.2 × 103/μL; RI 5.7–14.2 × 103/μL).

RBC osmotic fragility and total serum cholesterol

To determine effects of RBC storage on fragility assays, osmotic fragility curves were obtained using RBCs from both healthy dogs (n = 3) and a diabetic dog with hypercholesterolemia (n = 1) on the day the samples were collected, and after storage for 24 and 48 hours at 4°C. RBCs from the hyperlipidemic dog initially had increased osmotic fragility compared with RBCs from control dogs (Figure 1A). Minimal shifts in fragility of RBCs from healthy dogs were noted after 24 hours of storage, with slightly larger shifts evident after 48 hours. As the osmotic fragility in all the samples shifted to the same relative degree over time, the difference between the fragility of RBCs from healthy dogs and RBCs from the hyperlipidemic dog at any time remained constant (Figure 1B). Therefore, we elected to run all osmotic fragility assays on matched samples that had been drawn within 2–3 hours of each other and stored between 18 and 24 hours at 4°C, allowing us to perform the assay the day after the patient was initially seen when all clinical data were available. This is similar to how the assay is performed at human medical centers, which allow testing of samples up to 48 hours after collection, provided the sample has been refrigerated and a matched control is provided (Mayo Clinics,

Figure 1.

Storage of blood induced a minimal and proportionate increase in median fragility of RBCs from healthy and hypercholesterolemic dogs over time. (A) Representative plots of RBC fragility using samples from a control dog (Dog 1 from panel B) and a diabetic dog with hyperlipidemia (Dog 4 from panel B) after sample storage for 24 hours at 4°C. Dashed lines indicate mean corpuscular fragility (MCF), which is the NaCl concentration in which 50% of RBCs are lysed. (B) Median corpuscular fragility (MCF) of RBCs from 3 healthy dogs (Dogs 1, 2, and 3) and one diabetic dog with hyperlipidemia (Dog 4) after 0, 24, and 48 hours of storage of whole blood in EDTA at 4°C. Storage caused minimal increases in osmotic fragility over time, and the relative difference between the control and hypercholesterolemic samples was consistent.

Osmotic fragility of samples was compared using the median corpuscular fragility (MCF), which is equivalent to the saline concentration in which 50% of RBCs are lysed. In the first round of experiments comparing fragility of RBCs from 20 control dogs and 21 hypercholesterolemic dogs, RBCs from hypercholesterolemic dogs had significantly higher MCF (P < .001) compared with RBCs from normolipidemic dogs (Figure 2A). Some of the most profound shifts in MCF occurred in a dog with diabetes mellitus (MCF of 0.71%), and the 2 dogs with hyperadrenocorticism (MCF of 0.49% and 0.47%), compared with the mean (± SD) MCF for control dogs (0.42 ± 0.032%). In the second round of experiments, fragility of RBCs from dogs with diseases or treatment conditions associated with hyperlipidemia or dyslipidemia regardless of serum cholesterol concentration was evaluated. Dogs with diabetes mellitus and dogs being treated with glucocorticoids had MCF values that were significantly higher than MCF values of control dogs (Figure 2B, Table 1). Interestingly, some of these dogs had increased osmotic fragility despite having a serum cholesterol concentration within the RI. RBCs from dogs with hyperadrenocorticism tended to be more fragile, but this did not reach statistical significance. In some dogs with marked increases in serum cholesterol concentration, RBCs tended to have increased resistance to hypotonic lysis; these dogs were hyperlipidemic due to cholestasis. When data from the first round of 41 dogs and the second round of 30 dogs were compiled, serum cholesterol concentration and RBC fragility were weakly correlated (Figure 3). As the diseases and treatment conditions of the dogs recruited for this study were likely to result in increased concentrations in apolipoprotein B-containing lipoproteins or beta-lipoproteins[16, 17], samples from a subset of these dogs were investigated in greater detail with respect to RBC fragility, relative proportions of different lipoprotein classes in serum, and composition of RBC membranes.

Table 1. Osmotic fragility of RBCs from healthy control dogs and dogs with diseases or treatment conditions associated with hyperlipidemia or dyslipidemia.
  1. DM indicates diabetes mellitus; HAC, hyperadrenocorticism; GC, glucocorticoid administration; CS, cholestasis; MCF, median corpuscular fragility; SEM, standard error of the mean.

  2. a

    Data are mean values.

  3. b

    Statistically different from control values (P < .05, one-way ANOVA, Dunnett's Multiple Comparison Posttest).

Number of Dogs179564
MCF (% NaCl)a0.41630.4658b0.4450.4874b0.3875
Figure 2.

Box and whisker plots of median corpuscular fragility (MCF) of RBCs from hyperlipidemic dogs that were osmotically fragile, but specific hyperlipidemic disease states had disparate effects on RBC osmotic fragility. (A) Control dogs with serum cholesterol concentration within the reference interval had lower MCFs than hyperlipidemic dogs (**P < .01, Wilcoxon rank-sum test). Boxes represent the interquartile intervals from the 25th to the 75th percentiles; solid horizontal lines through the boxes represent the means, and the minimum and maximum values are represented by the capped vertical bars. (B) When dogs were grouped by disease state or treatment condition, RBCs from dogs with diabetes and from dogs treated with glucocorticoids were more osmotically fragile than RBCs from control dogs (*P < .001 ANOVA, Tukey). RBCs from dogs with hyperadrenocorticism tended to be more osmotically fragile and RBCs from cholestatic dogs tended to be less osmotically fragile, but neither of these relationships was significant. DM indicates diabetes mellitus; HAC, hyperadrenocorticism; CG, dogs treated with glucocorticoids; CS, cholestasis.

Figure 3.

Serum cholesterol concentration plotted against median corpuscular fragility (MCF) (n = 71 dogs). Cholesterol concentration and RBC osmotic fragility were weakly correlated (Spearman r = .3274, P < .005).

Lipoprotein profiles and RBC fragility

In dogs with diabetes or hyperadrenocorticism, and dogs being treated with glucocorticoids, marked increases in fractions of low-density lipoprotein (LDL) and very low-density lipoprotein (VLDL) and decreases in the high-density lipoprotein (HDL) fraction were found. This shift was reflected numerically by a significant decrease in the ratio of HDL to non-HDL (the sum of LDL and VLDL fractions and any retained chylomicrons) in all 3 groups compared with ratios in control dogs (Figures 4A and B). Dogs with cholestatic disease had either a proportional increase in both HDL and non-HDL lipoproteins (n = 2, Figure 4B, part E), or a predominant single band migrating between the alpha and beta regions (n = 1, Figure 4B, part F). Lipoprotein analysis was not obtained for one of the dogs with cholestasis. The correlation between the fraction of lipoproteins composed of non-HDL lipoproteins (beta-lipoproteins) and the MCF was much higher than the correlation with serum cholesterol concentration (Figure 5).

Figure 4.

Dogs with diabetes mellitus (DM, n = 6) or hyperadrenocorticism (HAC, n = 5), and dogs treated with glucocorticoid (n = 7) had similar dyslipidemias; higher levels of low-density lipoprotein (LDL) and very low-density lipoprotein (VLDL), and lower levels of high-density lipoprotein (HDL) (*< .001 ANOVA, Tukey) compared with values in control dogs. (A) Box and whisker plot of mean HDL:non-HDL ratio [HDL area under the curve (AUC)/(LDL AUC + VLDL AUC)] for control dogs, dogs with diabetes mellitus (DM) or hyperadrenocorticism (HAC), and dogs treated with glucocorticoids (GC). Boxes represent the interquartile intervals from the 25th to the 75th percentiles; solid horizontal lines through the boxes represent the means, and the minimum and maximum values are represented by the capped vertical bars. (B) Lipoprotein densitometry scans of serum for one representative dog from the following populations: (A) Control; (B) DM; (C) HAC; and (D) GC. (E, F) Dogs with cholestasis had either an increase in both the alpha (presumptively HDL, but migrating slightly more toward the origin) and beta (LDL) peaks (E), or a large band that migrated between the alpha and beta regions (F).

Figure 5.

RBC osmotic fragility is correlated with non-high-density lipoproteins (HDL) (beta)-lipoproteins. The area under the curve (AUC) for the low-density lipoprotein (LDL) and very low-density lipoprotein (VLDL) bands (non-HDL lipoproteins) was plotted against the median corpuscular fragility (MCF) for the 30 dogs in the second round of patient selection. The percentage of lipoproteins present in the non-HDL fraction had a moderate positive correlation with RBC fragility (Spearman r = .579, P = .0015).

Serum cholesterol, phosphorus, and lipoproteins and RBC membrane components

To further characterize the relationships among serum constituents and RBC membrane components, we measured the cholesterol, phosphate, and protein concentrations in RBC membrane ghosts and performed additional correlational analyses. Serum cholesterol concentration was positively, but weakly, correlated with RBC membrane cholesterol (Figure 6A), and no correlation was found between serum phosphorus concentration and phosphate, and hence phospholipid content, in the RBC membrane (Figure 6B). A moderate positive correlation was found between the phosphate:cholesterol ratio in serum and that in the RBC membrane (Figure 6C). The highest correlation was found between the proportion of non-HDL lipoproteins in serum and cholesterol content in the RBC membrane (Figure 6D).

Figure 6.

Correlation between RBC membrane and serum lipid components based on analysis of 28 of 30 dogs in the second round of patient selection. (A) Serum cholesterol concentration was weakly correlated with cholesterol per milligram of protein in the RBC membrane (Spearman r = .35, P = .02). (B) No correlation was found between serum phosphorus concentration and RBC membrane phosphate content. (C) Serum cholesterol:phosphate (C:P) ratio and RBC membrane C:P ratio had a moderate positive correlation (Spearman r = .58, P = .01). (D) Non-HDL lipoproteins in serum (AUC for LDL and VLDL bands) had a strong positive correlation with RBC cholesterol content (Spearman r = .72, P = .001).

Effects of plasma from diabetic dogs on osmotic fragility of RBCs from healthy dogs

To address the direct role of lipoproteins in inducing changes in RBC osmotic fragility, we tested the capacity of plasma from diabetic dogs to increase RBC fragility and the dependence of effects on lipids by chemical absorption of lipids. First, selective lipid depletion was verified (Figure 7A). Incubation of RBCs from a healthy dog in intact diabetic plasma caused a 5–6% increase in MCF, whereas incubation in lipid-depleted plasma had no effect (Figure 7B). In 2 additional experiments, incubation in diabetic plasma caused approximately 7% and 20% hemolysis of RBCs from a healthy dog, and this effect was also lipid-dependent (data not shown). We repeated the same experiments using RBCs from control dogs and intact and lipid-depleted autologous plasma; fragility of RBCs was not changed by the treatment (Table 2).

Table 2. Osmotic fragility (expressed in % NaCl) of normal canine RBCs incubated in intact or lipid-depleted autologous plasma.
TreatmentDog 1Dog 2Dog 3Dog 4
  1. Data are median corpuscular fragility values for RBCs collected from healthy dogs and incubated for 24 hours at room temperature with PBS, intact autologous plasma, or autologous plasma depleted of lipids. Each dog represents an independent experiment.

Intact plasma0.490.510.500.48
Lipid-depleted plasma0.480.500.510.50
Figure 7.

Intact plasma from diabetic dogs increased osmotic fragility of RBCs from a healthy dog in a lipid-dependent manner. (A) Representative validation of lipid depletion. Intact plasma sample (1) and plasma from the same dog cleared of lipoproteins by incubation with a lipid absorbent (2) were run in agarose electrophoresis (upper panel) and polyacrylamide electrophoresis for proteins (lower panel). The lipid gel was stained using Sudan black B and the protein gel using Coomassie Blue. Upper: Nearly all lipoproteins were removed by treatment, as indicated by the loss of bands in lane 2. Lower: The banding pattern is visually indistinguishable before (1) and after (2) lipid removal, indicating no or minimal loss of protein. (B) Increased osmotic fragility, as indicated by percent increase in median corpuscular fragility (MCF), of RBCs from a healthy dog that were incubated with PBS, whole (intact) plasma from a diabetic dog, and plasma from the same diabetic dog, but depleted of lipids. Baseline MCF (0 on the y axis) was set as the value obtained when RBCs were incubated with autologous plasma condition, similar to values obtained with incubation in PBS. Incubation in intact diabetic plasma caused an approximate 6% increase in MCF, which was dependent on the types of lipids in the plasma (n = 3, P < .0015, one-way ANOVA, Tukey posttest).


Our data demonstrate that RBCs from canine patients with increased beta-lipoproteins, both in the context of hyperlipidemia and dyslipidemia, are osmotically fragile. This has potential clinical significance because RBC fragility could contribute to intravascular hemolysis or increased splenic clearance of RBCs. In murine models of hyperlipidemia, increased RBC osmotic fragility is associated with increased splenic clearance and shortened RBC survival time. However, these studies also demonstrated a defect in RBC maturation characterized, in part, by organelle retention that may contribute to increased clearance, independent from membrane alterations.[18, 19] Organelle retention was not addressed in our study. The consequences of lipid-induced hemolysis or increased RBC clearance include subclinical to overt anemia, potentially unrecognized vascular inflammation, and increased risk of thrombosis. Additionally, because lipoprotein profiling is not routinely performed in veterinary medicine, a number of the dogs in this study would not have been recognized as having a defect in their plasma lipid compartment, and potentially their RBCs, using current clinical standards.

Most studies investigating which classes of lipoproteins are associated with RBC osmotic fragility have been carried out in rodent models or people. Our data are in concordance with data generated in these studies. In one study, RBCs isolated from human patients with a mixed-type hyperlipidemia, characterized by increased beta-lipoproteins, were osmotically fragile and the fragility was reversed by treatment with the lipid-lowering drug atrovastatin.[20] Others demonstrated a correlation between LDL concentrations and decreased fluidity of the superficial layer of the human RBC membrane.[21] In another study, it was demonstrated that purified human LDL causes an increase in human RBC rigidity as shown by electron paramagnetic resonance.[22] Interestingly, in the same study, HDL did not render the RBC rigid. Another study demonstrated that incubation with LDL caused decreased deformability of RBC concurrent with formation of high molecular weight protein complexes and evidence of spectrin degradation in the RBC membrane.[23] Our data indicate that lower density lipoproteins are more tightly associated with increased cholesterol content of RBC membranes and increased RBC osmotic fragility than total cholesterol. In clinical blood samples, lipemia and what is interpreted as artifactual hemolysis are frequently recognized concurrently, but the in vivo significance of this is not clear. The association we detected between RBC fragility and beta-lipoprotein concentrations suggests that hemolysis clinically recognized in lipemic samples may be occurring in vivo as well as in vitro.

In a recent report, administration of test compound SCH 900875 to Sprague–Dawley rats elicited hypercholesterolemia, and a decrease in concentrations of serum triglycerides that correlated with hemolytic anemia, acanthocyte formation, and increased RBC osmotic fragility.[24] Decreased concentration of triglycerides in the context of hypercholesterolemia may suggest an increase in triglyceride-poor, cholesterol-rich lipoproteins, which are typically high-density lipoproteins. However, because these results were found in the context of a drug that may also be altering hepatic synthesis and composition of lipoproteins, and individual lipoprotein fractions were not analyzed in the study, the relationship between various lipoprotein classes and the drug-induced anemia is unknown. None of the dogs we examined had morphologic alterations, detectable by light microscopy, in their RBCs. In a canine study, a high-fat diet induced hypercholesterolemia, moderate anemia, and cholesterol enrichment of high-density lipoproteins.[10] However, even with pooling multiple canine samples, the lower density fractions were too sparse in the control dogs to permit biochemical analysis, and the authors did not quantify relative proportions of the lipoprotein classes. In our study, we selected patients that were not anemic to avoid the confounding variables that would be introduced by the presence of reticulocytes when measuring RBC osmotic fragility. Thus, we can only speculate that alterations in serum lipoproteins may be contributing to increased RBC turnover and various degrees of anemia in our patients. Mechanistic links between hyperlipidemic state, increased osmotic fragility, and development of clinical anemia in dogs are not clear at this time, but warrant further investigation.

RBC fragility in canine diabetes has been previously evaluated.[25] In that study, no significant difference in RBC osmotic fragility between diabetic dogs and healthy controls was determined. In our study, the RBCs from diabetic dogs were consistently more fragile than those from healthy dogs. Both well-controlled and poorly regulated diabetic dogs were included in that earlier study. At our institution, poorly regulated diabetics are presented much more frequently, and therefore our samples may represent a selection bias toward more severely affected animals. Studies in people have shown an increase in osmotic and mechanical fragility of RBCs from diabetic patients.[26, 27] Our data show that even diabetics with total cholesterol concentrations within the RI have osmotically fragile RBCs, possibly due to the increased proportion of low-density lipoproteins, as our data suggest. However, diabetes is an oxidative disease and additional mechanisms may also be involved. RBC fragility in human diabetic patients has been shown to be highly correlated with concentrations of glycosylated hemoglobin.[28] Although in our study whole diabetic plasma with its lipid component intact enhanced RBC osmotic fragility, the increase in MCF was only approximately half the difference in MCF between RBCs from diabetic patients and healthy control dogs. Although lipoproteins in diabetic plasma are required to cause the plasma-induced changes in osmotic fragility, there are other factors inherent in RBCs of diabetic patients that also contribute to increased fragility. Oxidation of proteins within RBCs may be one of those factors that was not represented by incubation of normal RBCs in diabetic plasma. We investigated the effects of the full complement of plasma lipoproteins, and not individual lipoproteins, on fragility by evaluating the impact of their removal. Therefore, in line with what has been demonstrated in studies of human RBC, we hypothesize, based on consistent increases in LDL ± VLDL and increased osmotic fragility in the same canine patient samples, that lipoproteins and RBC fragility are linked mechanistically.

Human patients with hereditary spherocytosis frequently experience severe cholestasis, which rescues the inherent RBC membrane defect and returns the cells to a normal degree of fragility. Furthermore, RBCs from people with hereditary spherocytosis become less spheroidal and less osmotically fragile when they are transferred into serum from patients with obstructive cholestasis.[29] In our study, RBCs from dogs with cholestasis were at least as stable as RBCs from healthy controls and tended to have increased osmotic resistance. Studies of the effects of plasma from dogs with mechanical cholestasis on RBCs from normal dogs are warranted, but unfortunately, dogs with documented obstructive cholestatic disease and no additional disease processes were not presented to our hospital during the appropriate phase of the study period.

Osmotic fragility curves are also used to detect membrane defects in dogs. Most laboratories run osmotic fragility tests by comparing a potentially affected animal with a healthy control animal, and the lipid content of plasma from the affected animal or the healthy control may not always be known. Owing to the increase in RBC fragility detected in hyperlipidemic and dyslipidemic dogs, osmotic fragility curves could be misinterpreted if either the affected animal or healthy control has an abnormal lipid profile. Thus, increased RBC fragility may be attributed to inherited membrane defects in the affected animal when the increase actually reflects hyperlipidemia; conversely, increased RBC fragility in a normolipidemic dog may not be recognized when comparing the curve with one from a hyperlipidemic control animal.

The pathologic significance of increased RBC fragility in the context of the hyperlipidemic states we investigated is currently unknown. Hyperlipidemia is a well-known risk factor for thrombotic events in people. Although dogs with a number of endocrine and metabolic diseases have hyperlipidemia and concurrent increased risk for thrombosis, links between hyperlipidemia and thrombosis are understudied in this species. Hemolysis results in release of hemoglobin, which has been shown to scavenge nitric oxide with resultant promotion of platelet reactivity.[30] Furthermore, endothelial cells exposed to heme can activate the alternative complement pathway.[31] Low-grade hemolysis may be inciting unrecognized chronic vascular inflammation in hyperlipidemic and dyslipidemic dogs, as is widely recognized in people. The potential contribution of lipid-induced fragility and subclinical hemolysis to thrombotic risk warrants further investigation.


The authors would like to thank Dr. Karen Young, University of Wisconsin, Madison, for her critical review and editorial support in drafting this manuscript.


The authors have indicated that they have no affiliations or financial involvement with any organization or entity with a financial interest in, or in financial competition with, the subject matter or materials discussed in this article.