Tissue-Specific Knockout of the Insulin Receptor in Pancreatic Beta Cells
Creates an Insulin Secretory Defect Similar to that in Type 2 Diabetes
Rohuit N. Kulkarni, Jens C. Bruning, Jonathan N. Winnay, Catherine Postic,
Mark A. Magnuson, and C. Ronald Kahn
                         Reviewed by Michael Perraut
    The authors of this paper sought to determine the function of the
insulin receptor protein on pancreatic beta cells to see if it plays a role
in the development of Type 2 diabetes.  To test this, they knocked out the
gene which encodes the insulin receptor protein using a Cre-loxP system
which can select to delete a protein found in specific cells.  This is a
very well designed paper.  It is very critical of itself and takes many
steps to back up its claims.  However, the conclusions drawn from some of
the data is suspect and may require further study.

Type 2 diabetes in humans is characterized by insulin resistance and failure
of pancreatic beta cells to release insulin in response to glucose as
opposed to other signals like arginine.  The authors sought to discover
whether knocking out the insulin receptor on mouse pancreatic beta cells
would cause Type 2 diabetes-like symptoms.

The Cre-loxP system allows molecular biologists to specifically remove the
portions of DNA from cells of living animals and observe the effects.  The
authors inserted "a selection cassette flanked by loxP sites...upstream of
exon 4 [of the mouse insulin receptor gene] with a third loxP site
downstream of exon 4." (329)  This procedure transfected this insertion into
mouse embryonic stem cells to create heterozygous mice with flanked IRlox
genes (insulin receptor).  "In the presence of Cre recombinase, exon 4 of
the IRlox allele would be deleted, thereby causing a frameshift mutation and
an immediate stop of translation" of the insulin receptor gene (329).  Mice
with the Cre gene were created "using a 668bp beta cell-specific rat insulin
2 promoter" (329-330).  This would ensure that Cre would be expressed only
by "insulin producing beta cells." (330).  The Cre gene was transfected into
the IRlox stem cells using a plasmid and these mice were bred with
heterozygous IRlox mice to create the Type II deletion which knocked-out the
insulin receptor gene.  These beta cell-specific insulin receptor knockout
mice (BIRKO) were then tested to see the effects of removing the gene.  The
breeding also produced wild type mice, homozygous IRlox mice without the Cre
gene, and mice with only the Cre gene insertion.  These mice were used as
controls.  The authors explain that the "animals were born normally at the
expected Mendelian frequency, and no significant differences in body weight
were observed between newborn BIRKO mice and homozygous IRlox, Rip-Cre, and
WT littermates up to 6 months" (330).  Because no developmental differences
were observable the authors assume that any differences observed in insulin
production are the results of the knockout deletion as opposed to some birth
defect caused by the genetic alteration.  Figure 1a is a cartoon showing how
the authors created the BIRKO mice.  It shows schematic representations of
the endogenous allele for the insulin receptor, the targeting vector which
inserted the flanking loxP sites, the homologous recombinant, and the Type
II deletion or BIRKO mice.  Figure 1b is a flourograph of [pancreatic beta
cells with the Cre gene.  The red dots are immunoflouresced Cre proteins,
proving that the recombination was successful.

To ensure that the recombination was successful the authors used PCR to
amplify the DNA taken from pancreatic islet cells from the mice.  Figure 2a
shows a schematic representation of the IRlox allele to show the primers
used in the PCR analysis.  The knockout allele is also represented.  This is
a good method because it allows scientists to follow the work closely and
hopefully reproduce the same results.  Figure 2b is the result of RT-PCR
analysis on RNA from one homozygous IRlox mouse (negative control) and two
BIRKO mice.  As a positive control actin RNA was loaded to show that equal
amounts of RNA were loaded.  It appears that a little more RNA was loaded
for the homozygous IRlox mouse lane that for the two BIRKO mice lanes.
There is a large 480bp stain in all three lanes, a 300bp band for the
homozygous IRlox lane and a 220bp band for the two BIRKO mice.  The 300bp is
apparently the full insulin receptor while the 220bp band is the receptor
missing exon 4 which would indicate that the recombination was successful
and that these receptors are inactive.  However during the introduction they
claimed that the deletion would immediately stop translation, but for these
results they state that the 220bp indicates "the deletion of exon 4 and
splicing of exons 3 and 5 as would occur following Cre-mediated
recombination" (330).  If translation was stopped then exon 5 would never
have been produced.  This brings the results under suspicion since the
receptor is present, although missing 80bp on its RNA.  The authors do say
that the result should be a "308 amino acid fragment of the N terminus of
the insulin receptor alpha subunit, lacking a high-affinity binding site,
and the transmembrane and kinase domains" (329).  Therefore, since an
important part of the protein was deleted, the protein may still be
essentially "knocked out" even though it is translated by the pancreatic
beta cells.  To the authors credit they admit that a faint 300bp band is
found in the BIRKO mice's lanes.  They explain that this might be because
the DNA was taken from pancreatic islet cells that contain both beta and
non-beta cells.  Therefore, since the knockout was specific to beta cells,
RNA from non-beta cells would still be wild type and stain at the 300bp
height.  To test this the authors performed PCR analysis of beta cells from
an IRlox mouse (negative control), non-beta cells from a BIRKO mouse and
beta cells from a BIRKO mouse.  This well designed experiment showed a 300bp
band for the IRlox lane and non-beta cell lane and a 220bp band for the beat
cell lane.  These results are shown in Figure 2c and prove that the faint
300bp band in 2b is the result of non-beta cells from the BIRKO mice.  This
also reinforces the authors claims that the recombination was successful and
the IR gene was knocked out in the BIRKO mice.  The authors tested whether
the recombination was specific for pancreatic beta cells so the performed a
Western blot of protein from the muscle, liver, heart and brain from all
four strains of mice.  The results shown in Figure 2d show no difference in
expression.  The authors failed to include the pancreas cells as a negative
control.  Although all 4 lanes stained at the same height there is no way of
knowing whether BIRKO mice pancreas insulin receptors are the same size as
those of other cells.

The authors analyzed the BIRKO mice's glucose and insulin levels at 6 months
of age to see any difference when compare to the other three strains of wild
type, Rip-Cre, and hom. IRlox.  Figure 3 a-f show the results.  Overall,
there were no recognizable differences in glucose levels.  Scatter plots are
used in Fig 3a-d which analyze the glucose levels of male and female mice
when starved or randomly fed.  The glucose levels, of course are higher for
the randomly fed mice, but uniformly so.  Bar graphs show the insulin levels
of male and female mice for Fig 2e-f.  Insulin levels are higher for the
BIRKO mice.  The authors say the females showed mild hyperinsulemia and the
males showed a 2-fold increase as compared to IRlox and wild type controls.
This was not found at 2 months of age showing that the condition developed
over time just like Type 2 diabetes.  This is a well controlled experiment.
Male and female groups were separated.  Three controls were used each time
to rule out interference by either the Cre or lox insertions.

The authors then tested the different group's' ability to secrete insulin
when induced by two signals: glucose and arginine.  Figure 4a-b shows the
results.  Fig 4a-b show how the insulin level reacted after glucose
injection.  This that the insulin levels of the control groups jumped up
sharply less than 5 minutes after the injection while the BIRKO mice insulin
levels rose slowly, never reaching the control groups' level after 30
minutes.  The slow rise is explained as a secondary glucose response while
the sharp increase is a first phase response.  However, Fig 4c-d show no
difference in insulin levels when stimulated by arginine showing that the
BIRKO mice have a decreased ability to release insulin only when stimulated
by glucose as a result of their deletion of the pancreatic beta-cell's
insulin receptor.  These results are very convincing.

To test if the condition was age-dependent as in Type 2 diabetes, the
authors performed a similar test by injecting glucose into mice, but this
time tracking their glucose levels.  They performed this experiment
separately on male and female mice of ages 2, 4, and 6 months.  Fig5 a,c,
and e show the results for the males and Fig 5b, d, and f for the females.
These line graphs show that glucose levels remain higher at all ages for the
BIRKO mice compared to the three controls and that the gap worsens the older
the mouse (especially in females).  The authors claim, "the BIRKO mice
exhibited an age-dependednt progressive impairment in their ability to
dispose of a glucose lad" 332).  Apparently this impairment is due to the
mice's inability to secrete an adequate amount of insulin.  It would be
helpful if insulin levels were shown as opposed to glucose levels so the
readers could compare the results of Figures 4 and 5, but the authors do an
adequate job of connecting the knockout with insulin secretion impairment.

To ensure that the observed impairment found in the BIRKO mice was due to
the loss of insulin receptor function and not some other structural reason,
the authors compared the physiology of the wild type cells and BIRKO mice
cells.  Using immunohistochemical studies on pancreas cells the authors
tried to show that there is no difference morphologically between the wild
type and BIRKO cells.  Figure 6a shows immunostained pancreatic cells from
wild type and BIRKO mice at 2 and 4 months.  There is a 20-40% reduction is
size of cells in 4-month old BIRKO mice.  IT is plainly seen in the figure.
Figure 6b shows a bar graph comparing insulin levels of pancreatic cells in
control and BIRKO mice at 2 and 4 months of age.  The insulin levels of
BIRKO mice remains unchanged at the two ages, but is 35% lower than control
mice at 4 months of age.  The authors claim that "the decrease in BIRKO mice
cannot be explained simply by a decrease in islet size or insulin content"
(333).  This claim seems unfounded.  Perhaps the BIRKO mice's inability to
process glucose is because there is simply not as much insulin somehow due
to the knockout of the insulin receptor.  Figure 7a shows electron
microscopic pictures of IRlox and BIRKO mice.  There is no discernible
difference between the two, showing that the cells organelles were not
adversely affected by the genetic manipulation.  Figure 7b shows
immunoflourescence of IRlox, BIRKO, and ob/ob (negative control) cells with
Glut2 antibody.  While there is significantly less staining of the ob/ob
cells as should be expected, there is also less staining in the BIRKO
cells.  This is contrary to the authors' observation that "the distribution
of Glut2 appeared to be comparable with that in the controls" (333).  Not
only are the pancreatic cells smaller but there also appears to be less
paraffin (Glut2's substrate) in BIRKO cells.  It seems that the BIRKO mice
have more differences than just the knockout.  This hurts the authors'
claims and the fact that they have ignored these results further weakens
their stance.

Overall this is a very good paper based on sound theory, well designed and
controlled experiments, and detailed analysis.  Several suggestions have
been made as to how the paper could be improved, but by in large the authors
were successful in showing that the removal of the pancreatic beta cells'
insulin receptor lead to a decrease in the ability to process glucose.
However, the morphological data show obvious differences that could account
for this result, and the authors did little to explain it.  I do not accept
that the "altered glucose sensing is not the result of a developmental
abnormality, but rather the result of alterations in the intracellular
mechanisms of glucose sensing or the specific secretory machinery involved
in glucose stimulated insulin release", as the authors claim (336).  More
study into how the insulin receptor may affect cell growth is needed before
I can believe the authors' proposed mechanism.

What to do next

It would seem obvious that the next step in research is to isolate whether
the impairment was due to a decrease in cell size.  To test this, I believe
a knockout of the insulin receptor in other cell lines is necessary in order
to see if growth is affected there as well.  Since all cells have insulin
receptors, although not all secrete insulin, it would be simple to isolate
whether it is a cell's ability to secrete of it's actual size which is
impairing the BIRKO mice's ability to process glucose.  If the size of other
cells was found to decrease after deletion of their insulin receptors, the
work of these authors might be brought under question.

As mentioned above though, the paper still is a good one with loads of fine
work.  As good papers do, the authors have suggested areas of further
study.  They propose to follow the mechanism of glucose processing to
isolate exactly what role the insulin receptors of the pancreatic beta cells
play.  They hypothesize that glucose either acts on proteins which stimulate
the receptor or stimulate it directly which in turn stimulates itself "in an
autocrine manner" (337).  A two-hybrid system would be useful in following
the interaction of these proteins; however it is difficult to say how good
the results might be since hormones are involved.

The authors also mention that "the reduced glucose-stimulated insulin
release and low insulin content in BIRKO mice are in part the result of an
alteration at the level of insulin-regulated gene transcription.  If this is
the case, perhaps insulin may be involved in promoting the transcription of
certain growth factors that could account for the BIRKO mice's smaller size
and would vindicate the authors.  To test this scientists could test how
well certain cells grew in media with or without insulin.  These cells could
have either their insulin receptors knocked out or growth factors of
interest to isolate the role insulin might play in growth.

No matter what the end result.  There is no doubt that this paper is an
example of good science.  It asks a question, proposed a n answer, followed
reasonable experiments to prove itself, and has produced more questions to
be answered.

References:
Kulkarni, Rohit N., Jens C. Bruning, Jonathan N. Winnay, Catherine Postic,
Mark A. Magnuson, and C. Ronald Kahn. (1999). Tissue- Specific Knockout of
the Insulin Receptor in Pancreatic Beta Cells Creates an Insulin Secretory
Defect Similar to that in Type 2 Diabetes. Cell. 96. pp. 329-339.
  ------------------------------------------------------------------------

Link back to Mike's home page
Link to Davidson's Molecular Biology page
Send all questions comments or suggestions to: miperraut@davidson.edu