Culture and detection of primary cilia in endothelial cell models

Background The primary cilium is a sensor of blood-induced forces in endothelial cells (ECs). Studies that have examined EC primary cilia have reported a wide range of cilia incidence (percentage of ciliated cells). We hypothesise that this variation is due to the diversity in culture conditions in which the cells are grown. We studied two EC types: human umbilical vein endothelial cells (HUVECs) and human microvascular endothelial cells (HMEC-1s). Both cell types were grown in media containing foetal bovine serum (FBS) at high (20 % FBS and 10 % FBS for HUVECs and HMEC-1s, respectively) or low (2 % FBS) concentrations. Cells were then either fixed at confluence, serum-starved or grown post-confluence for 5 days in corresponding expansion media (cobblestone treatment). For each culture condition, we quantified cilia incidence and length. Results HUVEC ciliogenesis is dependent on serum concentration during the growth phase; low serum (2 % FBS) HUVECs were not ciliated, whereas high serum (20 % FBS) confluent HUVECs have a cilia incidence of 2.1 ± 2.2 % (median ± interquartile range). We report, for the first time, the presence of cilia in the HMEC-1 cell type. HMEC-1s have between 2.2 and 3.5 times greater cilia incidence than HUVECs (p < 0.001). HMEC-1s also have shorter cilia compared to HUVECs (3.0 ± 1.0 μm versus 5.1 ± 2.4 μm, at confluence, p = 0.003). Conclusions We demonstrate that FBS plays a role in determining the prevalence of cilia in HUVECs. In doing so, we highlight the importance of considering a commonly varied parameter (% FBS), in the experimental design. We recommend that future studies examining large blood vessel EC primary cilia use confluent HUVECs grown in high serum medium, as we found these cells to have a higher cilia incidence than low serum media HUVECs. For studies interested in microvasculature EC primary cilia, we recommend using cobblestone HMEC-1s grown in high serum medium, as these cells have a 19.5 ± 6.2 % cilia incidence. Electronic supplementary material The online version of this article (doi:10.1186/s13630-015-0020-2) contains supplementary material, which is available to authorized users.


Results
Table S1 Endothelial cell primary cilia incidence of HUVEC and HMEC-1s in various cell culture conditions. HUVECs grown in high serum (20% FBS) have higher cilia incidence than HUVECs grown in low serum (2% FBS). Post-confluence treatment did not significantly affect HUVEC cilia incidence. HMEC-1s have higher cilia incidence than HUVECs. Following cobblestone treatment high serum (10% FBS) HMEC-1s showed the highest cilia incidence of any of the assessed conditions. Data represents the median ± quartile range from 3-5 experiments.

Cell type
Expansion media Cilia incidence (%)

Statistical analysis
Multiple comparison analysis was performed on cilia length and cilia incidence data. This study had 2x2x3 factor setup, where factors were cell type (HUVEC, HMEC-1), serum levels during expansion (low, high) and post-confluent condition (confluent, serum starvation, cobblestone).

Primary cilia incidence
Primary cilia incidence data was right-skewed ( Figure S1). Cilia incidence was defined as n cilia /n cells ×100%, where n cilia is the number of cilia and n cells is the number of cells. Hence, number of cilia is a count variable (integer) that is dependent on the number of cells counted (extensive data). Poisson regression is the most appropriate choice for analysing this type of data [1] [2] . Cilia incidence was analysed using Poisson regression with a log link function offset by the number of cells. The fitted main effect Poisson regression model is described by the following equation: log number of cilia number of cells For the 3 factors, the reference levels are HUVEC, low serum during expansion and confluence. The coefficients B 0−n refer to the difference in log(n cilia /n cells ) changing that particular categorical variable with respect to the reference variable. For instance, if B 1 = 3, we would expect log(n cilia /n cells ) to increase by 3 if cell type is changed from HUVEC to HMEC-1 (all other factors being held constant). Table S4 shows the main effect model estimates for B 0−n . From these estimates and the standard error, 95% confidence intervals can be calculated. These are presented in the results section of the main paper.
This model has the same reference levels as main effect model. Inclusion of two-way interactions changes interpretation of B 1 . B 1 now refers to effect of cell type at the reference case for serum and condition (i.e. when serum and condition both equal zero) . Hence the two-way interaction model is used to compare individual populations, by altering the reference levels (see Figure 4 of main paper, and Table ??).  Table S6 shows how well each of the models fits the data. Of the three models examined, the two-way interaction model has the best fit. Three-way interaction between cell type, serum level and post-confluent condition was not significant, and this model did not improve the model fit to the observed data, hence was discarded.

Cilia length
Cilia length is not normally distributed, and is right-skewed ( Figure S2). A log transform was applied to our data, which resulted in a normal distribution of aggregate log(cilia length) as well as normal distribution in each of the 9 ciliated populations (HUVEC high confluent, serum starved, cobblestone; HMEC-1 low confluent, serum starved, cobblestone; HMEC-1 high confluent, serum starved, cobblestone).
A three-way ANOVA model with interaction was used to analyse log(cilia length). Analysis was performed using R software (version 3.1.2).
Serum levels during expansion have no significant effect on cilium length (Serum, p = 0.35). Cell type has an effect (p = 2.4e-5), as does post-confluence condition (p = 1.19e-3). There is no significant interaction between cell type and condition

Frequency Frequency
Length log 10 (Length) Figure S2 Histogram of cilia length prior and post log transform. Non-transformed data is right-skewed (Shapiro-Wilk test, p < 2.2e-16). Transformed data is normally distributed (Shapiro-Wilk test, p = 0.0516). (p = 0.074), nor between serum and condition (p = 0.28). Tukey honest significant difference post-hoc tests were used to determine where the significant interactions occurred, and the results of this analysis are presented in the main paper.  Figure S3 Additional 2 pairs of cilia-cilia contact observed in adjacent ciliated high serum (10% FBS) HMEC-1s, at confluence. From 39 cilia, 6 formed contacts (3 pairs). The intensity profile shows a reduction in intensity at the same location in both arl13b and 611b channels, indicating two separate cilia. The black line in RHS indicates the path of the intensity profile.