You use a signed integer type, so you just store a negative number.

You can think of fixed point as equivalent to ieee754 floats with a fixed exponent and a two’s complement mantissa instead of a sign bit.