Clustering analysis of lipoprotein profiles to identify subtypes of hypertriglyceridemia in Miniature Schnauzers

Abstract

Background

Hypertriglyceridemia (HTG) is prevalent in Miniature Schnauzers, predisposing them to life‐threatening diseases. Varied responses to management strategies suggest the possibility of multiple subtypes.

Hypothesis/Objective

To identify and characterize HTG subtypes in Miniature Schnauzers through cluster analysis of lipoprotein profiles. We hypothesize that multiple phenotypes of primary HTG exist in this breed.

Animals

Twenty Miniature Schnauzers with normal serum triglyceride concentration (NTG), 25 with primary HTG, and 5 with secondary HTG.

Methods

Cross‐sectional study using archived samples. Lipoprotein profiles, generated using continuous lipoprotein density profiling, were clustered with hierarchical cluster analysis. Clinical data (age, sex, body condition score, and dietary fat content) was compared between clusters.

Results

Six clusters were identified. Dogs with primary HTG were dispersed among 4 clusters. One cluster showed the highest intensities for triglyceride‐rich lipoprotein (TRL) and low‐density lipoprotein (LDL) fractions and also included 4 dogs with secondary HTG. Two clusters had moderately high TRL fraction intensities and low‐to‐intermediate LDL intensities. The fourth cluster had high LDL but variable TRL fraction intensities with equal numbers of NTG and mild HTG dogs. The final 2 clusters comprised only NTG dogs with low TRL intensities and low‐to‐intermediate LDL intensities. The clusters did not appear to be driven by differences in the clinical data.

Conclusions and Clinical Importance

The results of this study support a spectrum of lipoprotein phenotypes within Miniature Schnauzers that cannot be predicted by triglyceride concentration alone. Lipoprotein profiling might be useful to determine if subtypes have different origins, clinical consequences, and response to treatment.

Abbreviations

BCS

body condition score

HDL

high‐density lipoprotein

HTG

hypertriglyceridemia

LDL

low‐density lipoprotein

NTG

normal serum triglyceride concentration

PC

principal component

PCA

principal component analysis

TRL

triglyceride‐rich lipoprotein

1. INTRODUCTION

More than 75% of Miniature Schnauzers develop hypertriglyceridemia (HTG) by the age of 10 years old, predisposing them to pancreatitis, gallbladder mucoceles, glomerular proteinuria, and other diseases. ¹ , ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ When HTG is present without an identifiable underlying cause, HTG is classified as primary. Primary HTG is believed to have a genetic origin in Miniature Schnauzers, though the contributing genetic risk factors are unresolved. ¹² , ¹³

Historically, primary HTG in Miniatures Schnauzers was thought to be a single condition characterized by higher triglyceride‐rich lipoproteins (TRL) and lower low density lipoproteins (LDL). ¹⁴ However, Miniature Schnauzers with presumed primary HTG have varied responses to management strategies, such as feeding a low‐fat diet. ¹⁵ , ¹⁶ Further, lipoprotein profiles before dietary intervention can distinguish Miniature Schnauzers that will have HTG resolution when fed a low‐fat diet from those that will not respond. ¹⁵ This suggests that there might be multiple subtypes of HTG that contribute to clinical variability in treatment response. These subtypes could be the result of (a) distinct forms of primary HTG, (b) the presence of undiagnosed secondary contributors, or (c) a combination of both. Identifying and characterizing the spectrum of HTG subtypes in Miniature Schnauzers has potential implications for clinical management, research on HTG complications, and genetic approaches to discover risk variants.

Cluster analysis can identify phenotypic subtypes of disease. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² The objective of this study was to identify potential subtypes of HTG in Miniature Schnauzers through clustering analysis of lipoprotein profiles. We hypothesized that more than 1 phenotype of primary HTG exists within the Miniature Schnauzer breed.

2. MATERIALS AND METHODS

2.1. Samples

This was a cross‐sectional study using samples selected from 95 Miniature Schnauzer dogs with serum biobanked (−80°C) at the University of Minnesota Canine Genetics Laboratory from completed and ongoing research projects. ⁴ , ⁵ , ²³ The banked serum samples were collected between 2011 and 2019 with owner consent under study protocols previously approved by the University of Minnesota Institutional Animal Care and Use Committee (protocols #1207A17243 and #1509‐33019A). Samples were selected if a serum triglyceride concentration (measured by either a Roche/Hitachi Modular Analytics D2400 Module, Roche Diagnostics, Indianapolis, Indiana, or a Beckman Coulter AU480 Chemistry Analyzer, Beckman Coulter, Brea, California) was available and had been collected after asking owners to withhold food for 12‐18 hours. Samples were only included if they were obtained from dogs that met the below criteria for primary HTG, secondary HTG, or normal serum triglyceride concentration (NTG). Samples were excluded if the dog was receiving glucocorticoids, fibrates, or statins at the time of serum collection. Samples from dogs receiving omega‐3 fatty acid supplementation were permitted, and this information was recorded for reporting. None of banked samples available were from dogs receiving niacin, chitosan, or other lipid‐lowering medications.

2.2. Clinical categorization of hypertriglyceridemia

Data extracted from medical records included sex, age, body condition score (BCS, 1‐9 scale), fat content for the primary diet fed (g/100 kcal), medications, diagnoses, and results of endocrine testing performed as part of previous research or clinical evaluation. Primary HTG was defined as a serum triglyceride concentration greater than 108 mg/dL (1.2 mmol/L) in a dog without an identifiable cause. The 108 mg/dL threshold was selected based on the upper limit of the reference interval for 1 of the analyzers (Roche/Hitachi Modular Analytics D2400 Module, Roche Diagnostics, Indianapolis, Indiana); this threshold has also been used to define HTG in previous studies. ¹ , ² , ⁷ , ¹⁰ , ¹¹ , ¹⁴ , ¹⁵ Exclusion criteria for the primary HTG group included a diagnosis or clinical suspicion of an underlying condition that can cause HTG (eg, diabetes mellitus, hypothyroidism, hyperadrenocorticism, nephrotic syndrome) at the time of sampling and up to 6 months thereafter. Dogs with proteinuria were permitted in the primary HTG category if they had no other evidence of renal dysfunction (non‐azotemic and normoalbuminemic). ⁵ Secondary HTG was defined as a serum triglyceride concentration greater than 108 mg/dL (1.2 mmol/L) in a dog with a diagnosis of an endocrinopathy known to cause hyperlipidemia (eg, diabetes mellitus, hypothyroidism, hyperadrenocorticism). For both primary and secondary groups, the severity of HTG was characterized as mild for dogs with triglyceride concentrations of 109‐400 mg/dL (1.2‐4.5 mmol/L) and moderate‐to‐severe for concentrations >400 mg/dL (>4.5 mmol/L). ¹ Dogs with a serum triglyceride concentration less than or equal to 108 mg/dL (1.2 mmol/L) at 8 years of age or older were categorized as having NTG.

2.3. Lipoprotein profile analysis

Lipoprotein profiles were generated as described using a continuous lipoprotein density profiling method that uses bismuth sodium ethylenediaminetetraacetric acid as a self‐generating density gradient solution. ²⁴ Lipoprotein fractions were imaged as described using a custom fluorescence imaging system consisting of a digital camera and a metal halide continuous light source. ²⁴ This method identifies 11 distinct density lipoprotein fractions in dogs based solely on density characteristics (Table 1). ¹⁴ While functional classes (ie, TRL, LDL, etc.) are used to name the subfractions, these are strictly nominal and indicative of what lipoproteins are most likely to separate into these density ranges. The true composition and functional classes of the lipoproteins in each subfraction has yet to be determined in dogs.

TABLE 1.

Eleven density lipoprotein fractions identified in dogs using a continuous lipoprotein density profiling method that uses bismuth sodium ethylenediaminetetraacetric acid as a self‐generating density gradient solution.

Fraction	Density (g/mL)	Classification a
R1	<1.017	TRL
R2	1.019‐1.023	LDL1
R3	1.023‐1.029	LDL2
R4	1.029‐1.039	LDL3
R5	1.039‐1.050	LDL4
R6	1.050‐1.063	LDL5
R7	1.063‐1.091	HDL2b
R8	1.091‐1.110	HDL2a
R9	1.110‐1.133	HDL3a
R10	1.133‐1.156	HDL3b
R11	1.156‐1.179	HDL3c

Abbreviations: HDL, high‐density lipoproteins; LDL, low‐density lipoproteins; TRL, triglyceride‐rich lipoproteins (including chylomicrons and very‐low density lipoproteins).

^a The functional characteristics and composition of most lipoprotein density subfractions in dogs are currently unknown. Thus, all density subfractions can only be nominally assigned to traditional functional classes.

2.4. Data analysis

All statistical analyses were performed using R statistical software (v4.1.2, www.r-project.org). ²⁵ Data normality was evaluated using the Shapiro‐Wilks test and quantile‐quantile plot graphs generated using the “lattice” package (v0.20‐45). ²⁶ Clinical data, including age, sex, dietary fat, and BCS were compared between dogs with primary HTG and NTG using a Welch Two Sample t‐test (dietary fat), the Wilcoxon rank sum test for nonparametric data (age and BCS), and the Fisher's exact test for count data (sex and number of dogs fed a low‐fat diet, defined as a fat content <3 g/100 kcal).

The 11 fractions of the lipoprotein profile data (described in Table 1) were log transformed and Pareto scaled before analysis. Principal component analysis (PCA) was performed to visually assess lipoprotein profile data clusters. The clustering tendency of the data was also assessed using the Hopkins statistic, a statistic that measures the probability that a given dataset is uniformly distributed. ²⁷ The Hopkins statistic was calculated using get_clust_tendency() from the “factoextra” R package (v1.0.7). ²⁸ Hopkins statistics can take on values from 0 to 1, with those greater than 0.75 indicating a clustering tendency at the 90% confidence level. ²⁷ Variables were clustered using unsupervised (ie, agnostic to triglyceride concentration), agglomerative, hierarchical clustering. The Ward linkage method was applied on Spearman correlation distances between variables. The number of optimal clusters was determined using the “optCluster” R package (v1.3.2). ²⁹ A heatmap was created with the dendrogram to visually represent relative differences in lipoprotein fractions between clusters; the heatmap color key was used to refer to these visual intensities with values near zero (yellow) considered intermediate, negative values (blue shades) considered lower, and positive values (orange/red shades) considered higher.

Clinical data, including age, sex, dietary fat, HTG classification, and BCS were compared between each cluster to determine potential sources of differentiation. Clusters were compared using a 1‐way analysis of variance for parametric variables (dietary fat), the Kruskal‐Wallis test for nonparametric variables (age and BCS), and the Pearson's Chi‐squared test for count data (sex, HTG classification, number of dogs fed a low‐fat diet). Since triglyceride concentrations were not used to generate the clusters, the median and range of triglyceride concentrations were determined for each cluster; however, no statistical comparison was performed since triglycerides contribute to the lipoprotein profiles which are the variable use to measure similarity and dissimilarity in cluster analysis. Similarly, the proportion of dogs classified as NTG, primary HTG, and secondary HTG were determined for each cluster, but no statistical comparison was performed.

P values are presented without specifying a threshold for significance, consistent with the American Statistical Association's Statement on Statistical Significance of P‐values. ³⁰

3. RESULTS

3.1. Samples

A total of 50 samples from Miniature Schnauzers were included in the study. Of these, 20 were categorized as NTG, 25 as primary HTG (19 mild and 6 moderate‐to‐severe), and 5 as secondary HTG (2 mild and 3 moderate‐to‐severe). Signalment, BCS, triglyceride concentration, and dietary fat content are summarized in Table 2. Serum cholesterol concentrations were not measured as part of this study but were available for 27 dogs, including 8/20 NTG dogs (all were within the laboratory's reference interval), 14/25 primary HTG dogs (1 above the reference interval), and 5/5 secondary HTG dogs (3/5 above the reference interval). Of the 27 dogs with cholesterol concentrations available, 21 were measured on the same serum sample as the triglyceride concentration, 4 were measured within 30 days of the triglyceride concentration measurement, and 2 (both NTG dogs) were measured greater than 6 months after the triglyceride concentration measurement. As per the exclusion criteria, no dogs were receiving glucocorticoids, fibrates, or statins at the time of serum collection. Four dogs were receiving omega 3 fatty acid supplements, including 1 NTG dog, 2 primary HTG dogs (both mild), and 1 secondary HTG dog. No dogs were receiving niacin or chitosan.

TABLE 2.

Clinical characteristics of 50 Miniature Schnauzers with normal serum triglyceride concentration (NTG), primary hypertriglyceridemia (HTG), or secondary HTG.

Group characteristics	NTG (<108 mg/dL)	Primary HTG (>108 mg/dL)	P value for NTG vs Primary HTG	Secondary HTG
Total number, n	20	25	‐	5
Age years, median (range)	9 (8‐13)	10 (6‐14)	.39	12 (8‐14)
Sex a male, female	15, 5	15, 10	.35	3, 1
BCS, b median (range)	6 (3‐8)	6 (3‐7)	.40	5 (5)
Serum TG mg/dL, median (range)	58 (35‐102)	266 (110‐2821)	‐	772 (399‐1848)
Dietary fat c g/100 kcal, median (range)	4.0 (2.8‐4.8)	3.8 (2.1‐5.3)	.22	2.8 (2.8‐3.4)
Proportion of dogs fed a low‐fat diet	1/18 (0.06)	6/21 (0.29)	.10	3/5 (0.60)

Abbreviations: BCS, body condition score; HTG, hypertriglyceridemia; NTG, normal serum triglyceride concentration; TG, triglyceride.

^a All dogs were spayed or neutered except for 1 intact male with secondary HTG.

^b BCS unknown for 1 primary and 2 secondary HTG dogs.

^c Dietary fat content unknown for 2 NTG and 4 primary HTG dogs.

As per the exclusion criteria, none of the dogs with primary HTG had clinical suspicion for an endocrinopathy. However, the extent of diagnostic screening varied between dogs. All 25 primary HTG dogs had blood or serum glucose concentrations measured (median 100 mg/dL, range 67‐129 mg/dL [5.6 mmol/L, range 3.7‐7.2 mmol/L]) after food was withheld for 12‐18 hours. Sixteen primary HTG dogs were screened for hyperadrenocorticism (14 urine cortisol: creatinine ratios and 2 low dose dexamethasone suppression tests) with results within reference intervals. Five of the 9 dogs not screened for hyperadrenocorticism had undergone abdominal imaging (4 assessed by ultrasonography and 1 by computed tomography) that showed normal adrenal gland size and appearance. Twenty‐two primary HTG dogs were screened for hypothyroidism (serum total thyroxine concentration), all with results within laboratory reference intervals.

The secondary HTG group included 2 dogs with untreated hypothyroidism and 3 dogs with diabetes mellitus (1 diagnosed the day the serum sample was obtained, 1 diagnosed 9 days prior and unregulated, and 1 diagnosed 2 years prior and described as well‐regulated in the medical records). Three of the dogs in the secondary HTG group had no triglyceride concentration available before the diagnosis of the endocrinopathy. The other 2 had mild elevations at 172 and 261 mg/dL (1.9 and 2.9 mmol/L) noted before the diagnosis of diabetes mellitus and hypothyroidism, respectively; the dog with hypothyroidism was not screened for thyroid function at the time the HTG was first noted.

3.2. Data analysis

Clustering tendency was evaluated visually with PCA. The PCA score and loading plot for the first and second principal component (PC) are shown in Figure 1. Clustering of NTG, primary HTG, and secondary HTG dogs can be seen. The first PC was influenced by the TRL and LDL fractions. The second PC was influenced by the TRL, LDL_{4 and 5} fractions, and some of the high‐density lipoprotein (HDL) fractions (HDL_{2b, 3b, and 3c}). The third PC was strongly influenced by the HDL fractions. Score and loading plots for the first and third and second and third PCs are included in Figure S1. A list of the individual contributions of each lipoprotein fraction to PC 1, PC 2, and PC 3 is provided in Table S1. The Hopkins statistic also indicated a clustering tendency (H = 0.75). ²⁷

FIGURE 1.

Principal component analysis biplot to visualize clusters in lipoprotein profile data from 20 Miniature Schnauzers with normal serum triglyceride concentrations (NTG), 25 with primary hypertriglyceridemia (HTG), and 5 with secondary HTG. The first 2 principal components are plotted with loading vectors and ellipses drawn at 95% confidence intervals around the mean data points within NTG (yellow diamonds), primary HTG (dark gray circles), and secondary HTG (blue squares) groups. HDL, high‐density lipoproteins; LDL, low‐density lipoproteins; TRL, triglyceride‐rich lipoproteins.

Hierarchical cluster analysis identified 6 clusters. These clusters and a heatmap corresponding to the distribution of 11 lipoprotein fractions in the clusters are shown in Figure 2. Figure S2 includes overlaid lipoprotein profiles from the dogs in each cluster. Clinical data (ie, age, sex, dietary fat, and BCS) were compared between clusters with results reported in Table 3.

FIGURE 2.

Hierarchical cluster analysis and heatmap of lipoprotein profiling data from 20 Miniature Schnauzer dogs with normal serum triglyceride concentrations (NTG), 25 with primary hypertriglyceridemia (1 HTG), and 5 with secondary HTG (2 HTG). The heatmap corresponds to the intensity of the 11 lipoprotein fractions across dogs. For each dog, the clinical classification and triglyceride concentration (in parentheses) are included along the bottom. Clusters are separated by black lines. Cluster analysis was performed with Ward's method using the Spearman correlation distance for samples and Euclidean distance for lipoprotein fractions. 1 HTG, primary HTG; 2 HTG, secondary HTG; HDL, high‐density lipoproteins; LDL, low‐density lipoproteins; NTG, normal serum triglyceride concentrations; TRL, triglyceride‐rich lipoproteins.

TABLE 3.

Comparison of clinical variables across 6 clusters of lipoprotein profiles from 50 Miniature Schnauzers, including 20 with a normal serum triglyceride concentration (NTG), 25 with primary hypertriglyceridemia (1 HTG), and 5 with secondary hypertriglyceridemia (2 HTG).

Cluster	A1	A2	A3	A4	B1	B2	P value
Variables	n = 10	n = 8	n = 3	n = 7	n = 10	n = 12
Age years, median (range)	10 (8‐14)	10 (9‐13)	9 (9‐10)	10 (8‐14)	9 (8‐11)	11 (6‐14)	.50
Sex a male, female	4, 6	5, 3	3, 0	6, 1	7, 3	9, 3	.31
BCS, b median (range)	6 (4‐7)	5 (3‐6)	6 (6‐7)	5 (3‐7)	6 (5‐8)	6 (3‐7)	.14
Dietary fat c g/100 kcal, median (range)	3.4 (2.1‐4.3)	4.0 (3.6‐4.8)	3.9 (3.5‐4.0)	3.3 (2.8‐5.3)	4.0 (2.5‐4.3)	4.0 (2.8‐4.8)	.36
Proportion of dogs fed a low‐fat diet	3/9 (0.33)	1/8 (0.13)	0/3 (0)	3/7 (0.43)	1/7 (0.14)	2/8 (0.25)	.59
TG (mg/dL) (median, range)	179 (35‐392)	42 (35‐91)	61 (54‐65)	334 (85‐720)	136 (42‐338)	643 (41‐2821)	‐
Number of NTG	2	8 d	3	1	5	1	‐
Number of 1 HTG	8	0	0	5	5	7	‐
Mild	8 d	‐	‐	4	5	2 d
Mod/severe	0	‐	‐	1	0	5
Number of 2 HTG	0	0	0	1	0	4	‐
Mild	‐	‐	‐	0	‐	2
Mod/severe	‐	‐	‐	1 d	‐	2

Note: P values are for comparisons to determine if the clinical variable differs by cluster. Mild HTG is defined as a serum triglyceride concentration of 109‐400 mg/dL, and moderate‐to‐severe (mod/severe) HTG is defined as >400 mg/dL.

Abbreviations: BCS, body condition score; HTG, hypertriglyceridemia; NTG, normal serum triglyceride concentration; TG, triglyceride.

^a All dogs were spayed or neutered except for 1 intact male with secondary HTG.

^b BCS unknown for 1 primary and 2 secondary HTG dogs.

^c Dietary fat content unknown for 2 NTG and 4 primary HTG dogs.

^d Four dogs were receiving omega‐3 fatty acid supplementation, including 1 NTG, 2 primary HTG, and 1 secondary HTG dog; each of these dogs fell into different clusters.

The most dissimilar nodes (separated by the first branch of the dendrogram and referred to as A and B), predominantly differed in the LDL₄ and LDL₅ fractions, with dogs in the B clusters having a higher intensity across these 2 fractions compared to those in A clusters. The second major separation occurred between clusters B1 and B2. The lipoprotein profiles for dogs in cluster B1 were characterized by lower intensities over the HDL_2a fraction. The dogs in cluster B1 also had low to intermediate intensities across the TRL fraction and higher LDL_{4 and 5} fraction intensities. Cluster B1 comprised 5 NTG and 4 primary HTG dogs (all mild). Cluster B2 was characterized by the highest intensities across the LDL_1‐3 and TRL fractions. Cluster B2 included 1 NTG and 11 HTG dogs. Of those with HTG, 7 had primary and 4 had secondary (2 with untreated hypothyroidism, 1 with recently diagnosed, unregulated diabetes mellitus, and 1 with well‐regulated diabetes mellitus). This cluster comprised 37% (11/30) of the dogs with any degree of HTG and 78% (7/9) of the dogs with moderate‐to‐severe HTG. This cluster also included 80% (4/5) of the secondary HTG dogs. Due to the relatively high number of dogs with secondary HTG in the B2 cluster, endocrine testing results were reviewed for the 7 dogs with primary HTG to determine the comprehensiveness of their evaluations: 6 were screened for both hypothyroidism and hyperadrenocorticism, and 1 was only screened for hypothyroidism but had normal size and appearance of adrenal glands on abdominal ultrasonography.

Within the A node, clusters were less distinct. Clusters A2 and A3 had the lowest intensities for the TRL fraction but differed in their intensity over the LDL fractions. Dogs in cluster A2 had lower intensities over these fractions. These 2 clusters contained only NTG dogs, altogether including 55% (11/20) of the NTG dogs in the study. Dogs in cluster A1 had low to intermediate intensities in the TRL fraction and the lowest intensities across the LDL fractions. This cluster included 2 NTG and 8 primary HTG (all mild) dogs. Most dogs in cluster A4 had mildly higher TRL intensities; the intensities of other fractions were intermediate. Six of 7 dogs in cluster A4 had HTG, including 4 mild (all primary) and 2 moderate‐to‐severe (1 primary HTG and 1 secondary HTG, diagnosed with diabetes mellitus on the day the serum was collected).

4. DISCUSSION

In this study, hierarchical cluster analysis of lipoprotein profiles in 50 Miniature Schnauzers with NTG, primary HTG, or secondary HTG identified 6 clusters that might represent different lipid phenotypes in this breed. Three clusters were composed almost entirely of HTG dogs, 2 of only NTG dogs, and 1 with a 50:50 mix of dogs with NTG and mild HTG. These clusters appeared to be driven by differences in intensities across LDL and HDL fractions in addition to TRL. The clinical data evaluated (age, sex, BCS, and dietary fat) were not identified as a source of differences between clusters. Overall, the study results suggest the possibility of a spectrum of lipoprotein phenotypes within the breed that cannot be predicted by triglyceride concentration alone.

The first major separation between the lipoprotein profiles (A vs B node), was predominantly driven by differences in the LDL₄ and LDL₅ fractions, with dogs in the B node having a higher intensity across these 2 fractions. There were clusters of dogs with primary HTG within both of these nodes. In healthy dogs, VLDL and LDL contribute equally to triglyceride concentrations. ³¹ This explains why triglyceride concentration alone cannot reliably predict whether a dog with HTG has elevations in TRL, LDL, or both.

The largest cluster in the present study, B2, comprised 28% (7/25) of the dogs with primary HTG, and 80% (4/5) of the secondary HTG dogs. The inclusion of both primary and secondary dogs in the same cluster could indicate that the number of dogs with endocrinopathies was too small for the clustering analysis to capture unique phenotypes of these disorders and segregate them into their own cluster. Lipoprotein profiles of dogs with hypothyroidism and diabetes mellitus are generally characterized by increases across all fractions (TRL, LDL, and HDL), whereas hyperadrenocorticism primarily increases LDL fractions. ³² , ³³ , ³⁴ However, methodologic differences hamper direct comparison of subfractions between the present study and past studies. ³² , ³³ , ³⁴ Three of the dogs with secondary HTG (1 with hypothyroidism and 2 with diabetes mellitus) clustered side‐by‐side within the B2 cluster, meaning that they were more similar to each other than most other dogs within the cluster. These dogs had high intensities across TRL and LDL fractions but not HDL. Of note, another dog with secondary HTG (diabetes mellitus) had a different lipoprotein profile pattern that was not even within the B node. It is possible that the endocrinopathy was not the source of HTG in that dog and that genetic risk factors were instead the major underlying cause (ie, the dog had primary HTG, despite the concurrent endocrinopathy). Even in the dogs with secondary HTG that clustered together, it is likely that the HTG is not solely from the endocrinopathy but rather the sum of multiple risk factors that affect lipid metabolism. In support of this theory, 2 of the dogs with secondary HTG had mild HTG documented before their diagnosis of an endocrinopathy. Triglyceride concentrations are inconsistently increased in dogs with diabetes mellitus, hypothyroidism or hyperadrenocorticism, suggesting that the development of HTG is not an assured outcome of those disorders. ³² , ³³ , ³⁵ In humans, HTG is viewed as a continuum, with various degrees of genetic and environmental factors contributing to disease. ³⁶ Another possible explanation for the clustering of primary and secondary HTG cases in the B2 cluster is that some of the dogs categorized as having primary HTG had an undiagnosed subclinical endocrinopathy. However, all 7 primary HTG dogs in this cluster were screened for hypothyroidism, and 6 were screened for hyperadrenocorticism. The dog that was not screened for hyperadrenocorticism had normal appearance of adrenal glands on ultrasound. The lack of comprehensive endocrine screening for all HTG dogs is a limitation of the current study.

A pattern of higher TRLs without elevations in LDL fractions was observed in 2 clusters, A1 and A4, composed primarily of HTG dogs. The dogs in cluster A1 had low to intermediate intensities in the TRL fraction and low intensities across LDL fractions; all dogs with HTG in this cluster were mild cases. The A4 cluster had low to intermediate intensities in LDL fractions; this cluster included a range of HTG severity. The pattern observed in the dogs in the A1 and A4 clusters are most similar to what has previously been described in Miniature Schnauzers with NTG and HTG, respectively. ¹⁴ Using a similar method, a 2013 study determined that Miniature Schnauzers with HTG typically have higher TRL and lower LDL fractions relative to dogs of other breeds with NTG. ¹⁴ The previous study also determined that lipoprotein profiles in Miniature Schnauzers with NTG, have similar, albeit less pronounced, changes to their lipoprotein profiles. ¹⁴ The results of the present study also demonstrate that triglyceride concentrations are an incomplete method to detect alterations in lipoprotein metabolism, as each main HTG cluster (A1, A4, and B4) contained 1‐2 dogs out of 20 dogs with NTG.

Dogs in the B2 cluster, vs A1 and A4, might have distinct mechanisms contributing to HTG development that determine whether there is an increase only in TRLs or whether LDL increases also occur. The mechanisms that contribute to these different lipoprotein profile patterns could be genetic, environmental, or both. For example, lipoprotein lipase deficiency, which could be because of a loss of function mutation in the lipoprotein lipase gene, is associated with lower levels of LDL and HDL. ³⁷ In contrast, hepatic lipase deficiency, which could result from a loss of function mutation of the hepatic lipase C gene, is characterized by higher concentrations of triglycerides, LDL, and HDL. ³⁷ In terms of environmental contributors to dyslipidemia, none of the clinical variables tested (age, sex, BCS, or dietary fat) differed by cluster. Other than the 4 dogs in the B2 cluster with endocrinopathies, there was no identifiable clinical factor shared by other HTG dogs that clustered together. However, there are important contributing factors to dyslipidemia that were not evaluated for the purpose of this study, including insulin resistance, central obesity (vs overall BCS), pancreatitis, and exercise. ¹¹ , ³⁶ , ³⁷ Insulin resistance is of particular interest as a source of variation in lipoprotein profiles, as nearly 1 in 3 Miniature Schnauzers with HTG have an elevated serum insulin concentration. ¹⁰ Pancreatitis is another important consideration given the link between HTG and pancreatitis in Miniature Schnauzers. ² , ³ Pancreatitis is characterized by increases of the fractions corresponding to LDLs, mainly LDL_2‐4. ¹¹ However, lipoprotein profile patterns of dogs with pancreatitis differ from those identified in this study as dogs with pancreatitis also have decreases in the fractions corresponding to TRL, HDL_2a, and HDL_3c. ¹¹ Another important environmental factor could be different types of dietary fat, which can alter lipoprotein metabolism. ³⁸

In addition to identifying multiple lipoprotein profile patterns in dogs with HTG, we found 2 clusters, A2 and A3, which comprised only NTG dogs. Both clusters overall had lower intensities of the TRL fraction but differed in intensity of the LDL fractions, with most dogs in cluster A2 having lower intensities of these fractions. Low intensities across LDL fractions are a previously reported feature of the Miniature Schnauzer breed pattern. ¹⁴ We did not include non‐Schnauzer breeds in this study, which makes it difficult to know whether the A2 and A3 clusters are normal variations of lipoprotein profiles in healthy dogs or whether cluster A2 represents a breed‐specific dyslipidemia.

The remaining cluster, B1, comprised an equal proportion of dogs with NTG and mild HTG. The dogs in this cluster had variable intensities across the TRL fraction and higher LDL_{4 and 5} fractions and, less consistently, LDL_1‐3. One possible explanation is that the dogs in cluster B1 have a mild/early form of the dyslipidemia phenotype present in cluster B2. A limitation of this study is that not all dogs had cholesterol concentrations determined, and the proportion with this measurement was particularly low for the NTG group. Thus, there might be NTG dogs with hypercholesterolemia within the B1 cluster or other clusters. It is also possible that the mild HTG in dogs in this cluster is not because of genetic risk factors, but rather an effect of another patient or dietary factor that was not measured in this study.

In this study group of Miniature Schnauzers, dietary fat did not differ by cluster. Feeding a low‐fat diet resolves HTG in approximately 1 in 3 Miniature Schnauzers and can alter lipoprotein profiles. ¹⁵ There were 10 dogs included in this study on low‐fat diets (<3 g/100 kcal). It is unknown whether these dogs would have clustered in different groups if fed a higher fat diet, but they appeared to cluster with other dogs with similar triglyceride concentrations. We also were not able to identify an effect of age, sex, or BCS as a source of differences between clusters. Mild elevations in triglyceride concentrations occur in dogs in association with aging, ³⁹ , ⁴⁰ and aging is associated with mild increases in the TRL and decreases in the LDL and HDL fractions. ⁴¹ An overweight or obese condition is also reported to affect lipoprotein profiles in dogs; when compared to dogs with ideal body condition, overweight and obese dogs have higher TRL and HDL. ⁴¹ The absence of differences in dietary fat, age, and BCS between clusters might be related to the size of the study and relatively low numbers of dogs representing extreme ends of these variables (eg, relatively few dogs were fed low‐fat diets, all dogs were between 8 and 12 years of age, and most dogs in this study had BCS between 5 and 6).

Our study is limited by a small sample size, especially for dogs with moderate‐to‐severe HTG. Analysis of additional dogs could reveal further separations and better capture the true range of HTG subtypes in Miniature Schnauzers. Similarly, increasing the number of dogs could increase our power to detect subgroups that are present, particularly those with large effect, using clustering techniques. ⁴² Also, in dogs, the composition of lipoprotein density subfractions and their functional characteristics are currently unknown and assignment to traditional functional classes, such as LDL and HDL, can only be done nominally. Furthermore, we are limited in our ability to interpret what is driving the differences in lipoprotein profiles. Correlation of genetic variants with specific clusters might reveal genetic drivers of HTG subtypes, but analysis of genomic data was not within the scope of this study. Inclusion of additional dogs with secondary HTG of both the Miniature Schnauzer breed and other breeds not reported to have primary HTG could help separate disturbances that are directly associated with an endocrinopathy from those because of underlying genetic risk factors or a combination of both. Finally, it is also important to note, that while all the serum samples in this study were collected from dogs after withholding food for 12‐18 hours, these lipoprotein profiles only represent a “snapshot” from a single moment in a dog's life. Triglyceride concentrations can vary considerably in Miniature Schnauzers with HTG and range from mild to severe on different sample dates after food being withheld for 12 or more hours, without discernible changes in environment between dates. ¹⁵

In conclusion, using a hierarchical clustering technique, we identified 6 clusters of lipoprotein profiles in our study of Miniature Schnauzer dogs, 3 of which primarily included dogs with HTG. The data supports the hypothesis that more than 1 HTG phenotype might exist in Miniature Schnauzers, and differences in LDL fractions might be the major distinguishing factor. Further investigation is warranted to confirm the range and number of distinct lipoprotein profiles within this breed. Lipoprotein profiling might be a useful tool for future research to determine if subtypes of HTG in Miniature Schnauzers have different origins or clinical consequences.

CONFLICT OF INTEREST DECLARATION

Authors declare no conflict of interest.

OFF‐LABEL ANTIMICROBIAL DECLARATION

Authors declare no off‐label use of antimicrobials.

INSTITUTIONAL ANIMAL CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION

Samples used in this study were previously collected with IACUC approval (protocols #1207A17243 and #1509‐33019A).

HUMAN ETHICS APPROVAL DECLARATION

Authors declare human ethics approval was not needed for this study.

Supporting information

Data S1: Supporting Information.

Tate NM, Yamkate P, Xenoulis PG, et al. Clustering analysis of lipoprotein profiles to identify subtypes of hypertriglyceridemia in Miniature Schnauzers . J Vet Intern Med. 2024;38(2):971‐979. doi: 10.1111/jvim.17010