You cannot simply multiply the probabilities as the events are clearly not independent.
The hypergeometric pmf measures the probability for obtaining $x$ from $a$ favoured and $r-x$ from $n-a$ unfavoured when selecting $r$ from all $n$ items. Now, if we let $y=r-x$ and $b=n-a$ then we get the probability for obtaining $x$ from $a$ in category-1 and $y$ from $b$ in category-2 when selecting $x+y$ from all $a+b$ items:
$$\dfrac{\dbinom ax\dbinom{n-a}{r-x}}{\dbinom{n}{r}}=\dfrac{\dbinom ax\dbinom{b}{y}}{\dbinom{a+b}{x+y}}$$
Extending this to three (or more) categories is simply a matter of matching the number of binomial coefficients in the numerator to the number of categories. (ie Use three of them.)
You seek the probability for obtaining 0 from 3 apples, 2 from 5 bananas, and 2 from 7 oranges when selecting 4 from all 15 fruit.
So...
$$\mathsf P(A=0,B=2,O=2)=\dfrac{\dbinom 30\dbinom 52\dbinom72}{\dbinom{3+5+7}{0+2+2}}$$