Pilot 1: size selection

Aim 1: what selects for cell size?

Action: compare cell size distributions across samples subjected to different centrifugation efforts.

Experimental design:

Centrifugation efforts: 4 times x 6 speeds

Load libraries:

library(ggplot2)
library(dplyr)
library(rstatix)

data = data.frame(x1 = rep(c(1,2,3,4,5,6),4), 
                  x2 = rep(c(2,3,4,5,6,7),4), 
                  y1 = c(rep(1,6), rep(2,6), rep(3,6),rep(4,6)), 
                  y2 = c(rep(2,6), rep(3,6), rep(4,6), rep(5,6)),
                  ctr = c("D1", "D2", "D3", "D4", "D5", "D6",
                          "C1", "C2", "C3", "C4", "C5", "C6",
                          "B1", "B2", "B3", "B4", "B5", "B6",
                          "A1", "A2", "A3", "A4", "A5", "A6"),
                  count = c("analysed", rep("not analysed", 4), "analysed",
                            rep("not analysed", 6),
                            rep("not analysed", 2), "analysed", rep("not analysed", 3),
                            "analysed", rep("not analysed", 4), "analysed"))


ggplot(data=data) + 
  geom_rect(mapping=aes(xmin=x1, xmax=x2, ymin=y1, ymax=y2, fill = count), color="black", alpha=0.5) +
  geom_text(aes(x=x1+(x2-x1)/2, y=y1+(y2-y1)/2, label=ctr), size=4) +
  scale_x_continuous(breaks = c(1.5, 2.5, 3.5, 4.5, 5.5, 6.5), labels = c(500, 700, 900, 1100, 1300, 1500)) +
  scale_y_continuous(breaks = c(1.5, 2.5, 3.5, 4.5), labels = c(6, 5, 4, 3)) +
  xlab("rpm") +
  ylab("min")

Small sampled from supernatant and large sampled from pellet

Note: so far, I have analysed 5 centrifugation efforts: A1 (3min, 500rpm), A6 (3min, 1500rpm), B3 (4min, 900rpm), D1 (6min, 500rpm) and D6 (6min, 1500rpm)

Import the data:

data <- read.table("size_data.txt", header = T)
data$treatment <- factor(data$treatment, levels = c("L", "CC", "S"),
                         labels = c("large", "control", "small"))

# Calculate the average cell size by sample 
mean_data <- data %>%
  group_by(treatment, centrifugation) %>%
  summarise(avg_size = mean(size), sd_size = sd(size), na.rm = T)

Cell size distributions

Small was significantly different from control (~ 8% of differential selection), small was significantly different from large (6-16% of differential selection), but large was not significantly different from control.

ggplot(data, aes(x = size, fill = treatment)) + 
  geom_density(alpha = 0.3) +
  geom_vline(data = mean_data, aes(xintercept = avg_size, color = treatment),
             linetype = "dashed", size = 1) +
  xlab (expression(paste("Cell size (",mu,"m)", sep=""))) +
  ylab("Density") +
  facet_wrap(~ centrifugation)

ggplot(data, aes(x = size, fill = treatment)) + 
  geom_histogram(alpha = 0.7) +
  xlab (expression(paste("Cell size (",mu,"m)", sep=""))) +
  ylab("Count") +
  facet_wrap(~ centrifugation)

ggplot(data, aes(x = size, y = centrifugation, fill = treatment, by = treatment)) + 
  geom_boxplot() + 
  theme(legend.position="top")

mean_data

## # A tibble: 13 x 5
## # Groups:   treatment [3]
##    treatment centrifugation avg_size sd_size na.rm
##    <fct>     <chr>             <dbl>   <dbl> <lgl>
##  1 large     A1                 4.13   0.468 TRUE 
##  2 large     A6                 4.28   0.422 TRUE 
##  3 large     B3                 4.16   0.388 TRUE 
##  4 large     D1                 4.20   0.483 TRUE 
##  5 large     D6                 4.35   0.441 TRUE 
##  6 control   A1                 4.22   0.487 TRUE 
##  7 control   A6                 4.27   0.567 TRUE 
##  8 control   D1                 4.21   0.496 TRUE 
##  9 small     A1                 3.85   0.568 TRUE 
## 10 small     A6                 3.92   0.629 TRUE 
## 11 small     B3                 3.73   0.749 TRUE 
## 12 small     D1                 3.83   0.463 TRUE 
## 13 small     D6                 3.66   0.684 TRUE

pwc <- data %>%
  group_by(centrifugation) %>%
  pairwise_t_test(size ~ treatment, p.adjust.method = "bonferroni")
pwc

## # A tibble: 11 x 10
##    centrifugation .y.   group1  group2     n1    n2           p p.signif   p.adj
##  * <chr>          <chr> <chr>   <chr>   <int> <int>       <dbl> <chr>      <dbl>
##  1 A1             size  large   control    84    48 0.309       ns       9.27e-1
##  2 A1             size  large   small      84    60 0.00129     **       3.86e-3
##  3 A1             size  control small      48    60 0.000191    ***      5.72e-4
##  4 A6             size  large   control    65    48 0.987       ns       1   e+0
##  5 A6             size  large   small      65    68 0.000212    ***      6.37e-4
##  6 A6             size  control small      48    68 0.000666    ***      2   e-3
##  7 B3             size  large   small      62    55 0.000138    ***      1.38e-4
##  8 D1             size  large   control    98    57 0.83        ns       1   e+0
##  9 D1             size  large   small      98    51 0.0000127   ****     3.8 e-5
## 10 D1             size  control small      57    51 0.0000407   ****     1.22e-4
## 11 D6             size  large   small      58    28 0.000000222 ****     2.22e-7
## # ... with 1 more variable: p.adj.signif <chr>

Differential selection

data_small <- mean_data %>%
  filter(treatment == "small")

data_large <- mean_data %>%
  filter(treatment == "large")

data_control <- mean_data %>%
  filter(treatment == "control")

data_small$diff <- (data_large$avg_size - data_small$avg_size)*100/data_large$avg_size

ggplot(data = data_small, aes(x = diff, y = centrifugation)) +
  geom_point(size = 3) +
  xlab("Differential selection (%)") +
  ylab("Centrifugation effort") +
  ggtitle("Large versus small") +
  xlim(0,16)

cc_small <- data_small %>%
  filter(centrifugation == "A1" | centrifugation == "A6" | centrifugation == "D1")

cc_large <- data_large %>%
  filter(centrifugation == "A1" | centrifugation == "A6" | centrifugation == "D1")

cc_small$diff <- (data_control$avg_size - cc_small$avg_size)*100/data_control$avg_size
cc_large$diff <- (data_control$avg_size - cc_large$avg_size)*100/data_control$avg_size

ggplot(data = cc_small, aes(x = diff, y = centrifugation)) +
  geom_point(size = 3) +
  xlab("Differential selection (%)") +
  ylab("Centrifugation effort") +
  ggtitle("Control versus small") +
  xlim(0,16)

ggplot(data = cc_large, aes(x = diff, y = centrifugation)) +
  geom_point(size = 3) +
  xlab("Differential selection (%)") +
  ylab("Centrifugation effort") +
  ggtitle("Control versus large") +
  xlim(0,16)

Cell abundance after centrifugation and size selection at T0

Cell abundance was always higher in the large compared to small (3 to 53 times higher). Cell abundance in control samples were higher than small but lower than large. It seems that almost all cells were in the pellet for D6 (and they survived centrifugation, which suggests this time/speed could be a good candidate for washing the sample by the end of selection, following Dustin’s protocol).

data_count <- read.table("count_data.txt", header = T)
data_count$treatment <- factor(data_count$treatment, levels = c("L", "CC", "S"),
                         labels = c("large", "control", "small"))

ggplot(data_count, aes(x = cell_ml/10^5, y = centrifugation, color = treatment, by = treatment)) + 
  geom_point(size = 3) + 
  theme(legend.position="top") +
  ylab("Centrifugation effort") +
  xlab (expression(paste("Cells m",L^-1,"x ", 10^5, sep="")))

count_small <- data_count %>%
  filter(treatment == "small")
count_large <- data_count %>%
  filter(treatment == "large")
count_control <- data_count %>%
  filter(treatment == "control")

count_small$diff <- (count_control$cell_ml - count_small$cell_ml)*100/count_control$cell_ml
count_large$diff <- (count_control$cell_ml - count_large$cell_ml)*100/count_control$cell_ml

ggplot(data = count_small, aes(x = diff, y = centrifugation)) +
  geom_point(size = 3) +
  xlab("Differential selection (%)") +
  ylab("Centrifugation effort") +
  ggtitle("Control versus small") +
  xlim(0,100)

ggplot(data = count_large, aes(x = diff, y = centrifugation)) +
  geom_point(size = 3) +
  xlab("Differential selection (%)") +
  ylab("Centrifugation effort") +
  ggtitle("Control versus large") +
  xlim(-500,100)

Aim 2: do Ochromonas survive centrifugation?

Action: track growth over 7 days and compare a control sample not centrifuged with a control sample centrifuged.

Data here are shown only for not centrifuged control. I am still counting the centrifuged control samples.

Note: I believe cell abundance goes down from T0 to T2 because abundance was still pretty high after dilution AND I could not add yeast to fresh media (the volume was so low that I could not pipette it out of the tip into the well!) while the mother culture was growing in media with the addition of yeast.

data_cn <- read.table("cn.txt", header = T)

ggplot(data_cn, aes(x = day, y = cell_ml, col = centrifugation)) +
  geom_point(size = 3)

Next steps:

Following Dustin’s protocol, I understood we want to:

1) Aim to dilute biomass (after seletion) to ~ 20% of the initial biomass for both small and large.

2) Get a 10% differential selection in cell size for both small and large relative to the control.

If the above is correct (probably worthy checking with Dustin), I should run additional pilots to see a difference between large and control and to decrease the biomass of the large samples (from my results I think most cells are going to the pellet). This will also mean that I will have different optimal centrifugation efforts to select for small and large, the former being faster and the latter being slower (just like Dustin describes in the paper).

Suggestions for next pilot:

1) Run lower centrifugation efforts to see if I can pellet a smaller amount of cells which will be larger than the control.

2) Run a slightlty different protocol. To evaluate for size differential selection, I think I should sample (from the same tube) before and after centrifugation and measure the cells. I think that by doing this I will have a “true” control to evaluate changes in cell size distributions, what do you think?

3) Should I try to implement Dustin’s protocol and evaluate survival after carrying out centrifugation twice + a final fast centrifugation to pellet all cells and replace supernatant with fresh media?

From the results we already have, what about picking a centrifugation effort for the small cells? For example, test D1 and B3 following the above? Or should I do more tests?

Any suggestions for other samples that I should analyse from Pilot 1? (except from analysing survival over time, which is work in progress!). Any other cell measurements that could perhaps be useful?